mechanical behavior of the timberâ€“terrazzo composite floor

Construction and Building Materials 80 (2015) 295–314

Contents lists available at ScienceDirect

Construction and Building Materials

journal homepage: www.elsevier .com/locate /conbui ldmat

Mechanical behavior of the timber–terrazzo composite floor

http://dx.doi.org/10.1016/j.conbuildmat.2015.01.0680950-0618/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +39 041 2571289.E-mail address: [email protected] (P. Foraboschi).

Paolo Foraboschi ⇑, Alessia VaninUniversità IUAV di Venezia, Dipartimento di Architettura Costruzione Conservazione, Convento delle Terese, Dorsoduro, 2206, 30123 Venice, Italy

h i g h l i g h t s

�Mechanical behavior of the terrazzo flooring.� Experimental results from tests performed on real scale structures differing for the Venetian floor.� Experimental data used to develop a predictive analytical model.� Peculiarities of the traditional structural system are described.� Terrazzo provides the floor with significant extra-stiffness and extra-strength.

a r t i c l e i n f o

Article history:Received 2 October 2014Received in revised form 18 January 2015Accepted 20 January 2015

Keywords:Composite floorsHistorical buildingsHistorical flooringsTimber-slab floorVenetian flooringVenetian terrazzo

a b s t r a c t

The mechanical behavior of the timber–terrazzo floor is investigated in this paper. The presentationrefers to the traditional Venetian floor; however, the research was comprehensive and hence the resultsof this paper include any timber floor finished with terrazzo flooring.

The first part of the paper presents experimental results from laboratory tests performed at differentload stages on two real scale masonry buildings, differing only for the presence of the Venetian floor.The experiments also included inducing damage artificially.

The experimental data were then elaborated to describe peculiarities of the traditional structural sys-tem. The test results presented in the paper prove that the terrazzo is an important load-bearing compo-nent, since it provides the floor with substantial extra-stiffness and extra-strength. The test results alsoprove that even a high level of damage does not reduce significantly the stiffness of this type of floor.

Research activity was then directed toward analytical modeling. The second part of the paper presentshow the experimental data were used to construct a predictive theoretical model, whose reliability isproven by the good agreement between the results from the model and the findings from the tests.The model can be used to assess the stiffness and load-carrying capacity of timber–terrazzo floors andas a basis for floor rehabilitation.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction Today, the chips may also be of onyx, glass, smalti or synthetic

This paper deals with the terrazzo flooring that rests on timberfloors, and analyzes the mechanical behavior of the structural sys-tem that derives from the composition of terrazzo, beams, andboards.

The term ‘‘terrazzo’’ derives from the Italian ‘‘Terrazza’’. The def-inition of terrazzo over the centuries is: a form of mosaic flooringmade by embedding chips in a matrix and polishing.

In a terrazzo floor, the chip is the ingredient that acts as aggre-gate (i.e., inert materials). The chips are screened to various sizes.Traditionally, the chips were made of small pieces of marble, gran-ite or other minerals such as quartz, quartzite, and river gravel.

types.In a terrazzo floor, the matrix is the ingredient that acts as a bin-

der to hold the chips in position. Traditionally, the matrix wasmade of mortar (either hydraulic or non-hydraulic lime mortar).Nowadays, the matrix may also be made of cementitious, modi-fied-cementitious or resinous materials.

The terrazzo is poured in place, cured, ground, and polished toachieve a smooth surface.

The terrazzo was invented by Venetian construction workers.Nevertheless, it was the evolution of patient works of art attributedto the first Roman age, where the floors were cemented with lime,according to a Greek procedure that had been then perfected inItaly. The Venetian terrazzo was introduced as a low cost durableflooring system using marble chips from upscale jobs and goat milkas the sealer, which originally were usually set in clay to surfacethe patios around the living quarters.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.conbuildmat.2015.01.068&domain=pdf

http://dx.doi.org/10.1016/j.conbuildmat.2015.01.068

mailto:[email protected]

http://dx.doi.org/10.1016/j.conbuildmat.2015.01.068

http://www.sciencedirect.com/science/journal/09500618

http://www.elsevier.com/locate/conbuildmat

296 P. Foraboschi, A. Vanin / Construction and Building Materials 80 (2015) 295–314

The use of the terrazzo changed over the years and after fewcenturies it was used as high-grade flooring, widely specified byVenetian architects for its aesthetic and decorative characteristicsas well as for its durability. Thus, we find the terrazzo in the vastmajority of the Venetian buildings. Nevertheless, this type of flooris also used nowadays all over the world.

In historical buildings [1], the slab that forms the terrazzo floor-ing rests on timber beams and boards. According to the technicalhistory of Venetian architecture, in the designers’ intent the timberbeams and boards were the structure of the floor, while the ter-razzo slab was a non-load-bearing element devoted to obtainingthe flat surface as well as to providing the flooring with aestheticquality. However, the comparison between historical timber floorswith and without [2] the terrazzo suggests that not only has theterrazzo a structural role in the floor structure, but that the design-ers knew and considered that structural role. The Venetian build-ings, where the terrazzo is widely used, provide significantexamples of the structural role of the terrazzo.

1.1. Venetian buildings structurally considered

Venetian buildings have always been designed and constructedconsidering three peculiar characteristics — namely, the brackishwater and tides of the lagoon, the shortage of available space,and the soft superficial soil layers (silt and clay). The Venetianfoundation system was designed and constructed so as to ade-quately cope with such adverse environment [3].

The soft soil entailed that the masonry walls of the buildingshad to be supported on pile groups embedded into the soil. Eachpile was driven vertically into the ground up to reaching the deepsoil layers, which are relatively strong (Fig. 1). Then, a pile cap wasconstructed on the top of each group of foundation piles to evenlyaccommodate the building that the piles had to carry. The pile capsconsisted of timber boards, called ‘‘madieri’’ (Fig. 1 shows a single-layer-board system, i.e. one layer of "madieri").

Bricks were then laid onto the madieri and covered with specialwaterproof clay, called ‘‘tera da savon’’ (in Venetian vernacular), soas to obtain the basis of each wall. Moreover, a vertical layer madeof Istrian stones was placed on the side of the walls facing the canal(Fig. 1). In so doing, contact between the masonry and the water ofthe lagoon was prevented.

Some continuous courses of Istrian stones, called ‘‘cadene’’, wereplaced above the middle see level, in order to create waterprooflayers that protected the upper parts of the wall from the risingmoisture (Fig. 1). Moreover, the face of each wall in contact withthe lagoon water was protected from the aggressive atmosphericagents by spreading hydraulic lime plaster onto (Fig. 1). However,the above-described system improved the performance of thefoundation system but not enough to provide a building with ade-quate capacity to cope with the soft soil.

In order to allow the walls to withstand the differential settle-ments of the ground, special techniques had also to be used forthe superstructure. In some cases, the longitudinal walls (parallelto the canal faced by the building) were not connected to the trans-verse walls (Fig. 2a), in order to allow the movements of the wallsto occur without causing any cracks. In these cases, tie-rods wereplaced, in order to compensate for the lack of connection betweenlongitudinal and cross walls (Fig. 2b). The tie-rods prevented theout-of-plane collapse mechanism of the walls from occurring andtherefore allowed the walls to be slender although without con-nections at the vertical edges [4,5].

One technique that was always used in Venetian buildings towithstand soil settlements was to use weight-saving constructionelements. To this end, the walls were relatively thin (high slender-ness ratio) or perforated, and bricks were preferred over stones.Where possible, moreover, the walls were replaced by columns.

In so doing, the vertical structures were space-saving as well,which fulfilled another requirement — namely, to occupy as littlespace as possible [3].

In order to obtain weight-saving floors, timber beams andboards [6–18] were preferred to masonry vaults and spandrel fill(Fig. 3). The floor-system also included a timber beam, called‘‘rema’’, which was placed onto the top of the wall (transverse tothe floor beams) to protect from the rising damp the floor beamsthat rested on (Fig. 3). Conversely, except for the churches, themasonry vaults were used in few Venetian buildings only. There-fore, the curved components of the Venetian buildings consist ofmembers having the appearance of an arch or a vault though notof arch or vault construction (false vaults). A vault-like was madeof wattle and daub hung to timber curved beams, which, in turn,were hung to the timber beams of the floor (Fig. 3b). The wattleand daub surface was plastered. Contrary to a masonry vault[19], this system neither is heavy nor transmits any horizontalthrust to the walls.

While the floor’s structure of the Venetian buildings met theweight-saving criterion, conversely, the flooring did not meet thiscriterion. In fact, the Venetians architects and builders usually usedthe terrazzo for the flooring of both rich and ordinary buildings,which is a particularly heavy type of flooring. The terrazzo was laidonto the timber boards, which had been nailed to the timberbeams; its thickness was never less than 130 mm and often sur-passed 200 mm, even up to 250 mm (sometimes more). It is to notethat those values refer to the average thicknesses of the terrazzoslabs, since the timber floors that the terrazzo was poured ontowere not flat but warped (i.e., cupped).

The typical Venetian Terrazzo floorings can be grouped into threetypes — namely, lime terrazzo, pastellone, and cement terrazzo.

The lime terrazzo was composed of a first layer called ‘‘sotto-fondo’’, made of lime and stone or pieces of bricks, whose thicknessranged from approximately 80 to 110 mm. Hence this flooring wasnamed after the first layer. Over this, there was a second layer,called ‘‘coprifondo’’, made of lime and dust of bricks, whose thick-ness was from 20 to 50 mm. Over this, there was the final layer,called stabilitura, made of pieces of marble; this third layer wasthin (10–20 mm).

The pastellone differed from the lime terrazzo for the secondlayer, which was made of lime together with powder of brickand marble and had thickness from 10 to 150 mm, and for the thirdlayer, called ‘‘pasta’’, which was made of lime together with pow-der of brick or/and marble and was thin (10–20 mm). Hence, thisflooring was named after the third layer.

The cement terrazzo differed from the lime terrazzo only for thebinder, i.e., cement in lieu of lime. Hence, it was named after thebinder.

According to the building materials that typify the layers com-posing the terrazzo and to the way it is composed, the unit weightof a terrazzo ranges from 3.2 kN/m2 to 4.5 kN/m2. Thus, not only isthe weight of the terrazzo high, but above all the terrazzo is dras-tically heavier than all the other types of non-load-bearing compo-nents of the floor as well as of every other building component,either load-bearing or non-load-bearing [20,21].

Hence, the terrazzo seems to be in flat contradiction to all theother construction elements of the Venetian buildings [3]. Thatanalysis suggested that the terrazzo was also uses as load-bearingcomponent of the floor.

2. Research significance

The research presented in this paper aimed at identifying thestructural role of the terrazzo flooring in the load-carrying capacityand service–load deflection of floors. Activity was directed at ana-lyzing if the terrazzo provides the timber structures that it rests on

TIMBER PILES

TIMBER BOARDS("madieri")

INTERIOROF THE BUILDING

BRICK MASONRY

PROTECTION PLASTER

ISTRIAN STONE BLOCKS

LAGOON WATER

WATERPROOF CLAY("tera da savon")

MEAN SEA LEVEL

SIDEOF THE SEACOURSES OF

ISTRIAN STONE BLOCKS("cadene")

Fig. 1. Diagram of the typical foundation system of Venetian buildings, with the components (some of which are called in Venetian vernacular).

TIMBER BOARDS

TIE ROD

TIMBER BEAMSISTRIAN STONE BLOCK

"TERRAZZO"

(b)

(a)

Fig. 2. Relationships between the longitudinal and transverse walls of Venetian buildings. (a) In some buildings, no significant connection existed between orthogonal load-bearing walls. (b) Structure used in buildings whose orthogonal walls were not connected to each other: tie-rod beneath the terrazzo.

P. Foraboschi, A. Vanin / Construction and Building Materials 80 (2015) 295–314 297

with composite behavior [16,21–26] which guarantees extra-stiff-ness and extra-strength [5,6,8,10], as well as at carrying outresearch to help reduce the incidences of structural failure[1,5,7,15,27–35] of timber–terrazzo floors, so as to extend theoperating horizons of this flooring system.

Those objectives were pursued by means of laboratory tests onfull-scale specimens. The experimental observation and the elabo-ration of the test results provided the main mechanisms that gov-ern the behavior of the timber–terrazzo floor. Since those findingsare comprehensive, they can be generalized and applied to anytimber–terrazzo floor, including to interpret the mechanicalbehavior of this type of floor.

Those findings were also used to develop a predictive theoreti-cal model whose assumptions were derived directly from theexperimental data. This analytical model takes into account thecontributions that the terrazzo provides the floor with, and there-fore the model predictions are reliable.

The model, which this paper presents from the assumptions tothe final equations, can be used to provide existing timber–ter-razzo floors with adequate behavioral characteristics (stiffnessand strength), using the minimal amount of structural work andwithout altering the way the building reacts to applied loads. Ulti-mately, the paper provides the means to assess the structural con-ditions of this type of floors and helps understand how to

"TERRAZZO"

TIMBER BOARDS

TIMBER BEAMS

TIMBER BOARDS

TIMBER BEAMS

WATTLE AND DAUB

PLASTER

"TERRAZZO"

(a) (b)

REMA

Fig. 3. Diagram of timber floors (a) with plane (b) and vaulted intrados. From the structural perspective, both the floors are not-curved (no thrusting action transmitted to thewalls).


rehabilitate existing timber–terrazzo floors simultaneously guar-anteeing safeguarding and conservation [1,6–10,24,27,29–38].

3. Experimental tests

This research was based on laboratory tests devoted to detecting experimen-tally the contribution that the terrazzo provides a timber floor with. The laboratorytests were also devoted to detecting how damage to the timber affects the behaviorof the floor.

In order to identify the structural role of the terrazzo, the experiments wereperformed on two real scale structures differing only for the presence of this Vene-tian flooring system (Fig. 4). More specifically, the first structure included a floormade of timber beams with transverse timber boards nailed onto (single layer);the second structure included the same timber beams and boards plus the terrazzoon the boards. These floors were subjected to distributed and concentrated down-ward loads; the differences in behavior were only due to the terrazzo.

3.1. Specimens

The behavior of a timber floor also depends on the supporting vertical structures,since the restraint conditions are dictated by the masonry walls that the floor rests onand is fixed into. In particular, the end-restraint of a timber beam depends on thethickness and brick pattern of the wall underneath the beam as well as on the stressesthat the wall above transmits to the end of the beam. Therefore, in order to reproducethe real behavior, the specimens had to avoid scale effects and had to include themasonry walls that support the floor as well as the walls above, including the com-pression transmitted to the floor’s edges and walls underneath.

Fig. 4. Photo of the two specimens prepared for the laboratory tests: full-scale two-story one-bay specimens.

Therefore, two full-scale two-story one-bay specimens were designed and fab-ricated (Fig. 4); the two specimens had the same overall dimensions while they dif-fered from one another (Fig. 5) only in the presence of the terrazzo on the first floor.The total span was 3.24 m, the width 2.85 m, the total height 4.23 m (the height ofthe first floor was 2.28 m).

The second floor of each specimen was only meant to provide the walls abovethe first floor with weight, so that the end-restraints of each first floor reproducedthe real restraints of historical buildings.

The first level of each test specimen was made up of the floor (first floor) andtwo supporting masonry walls (first-story walls) having thickness of 250 mm(Fig. 5). The second level was made up of the floor (second floor) and two support-ing masonry walls (second-story walls) having thickness of 125 mm (Fig. 5).

All the masonry assemblages were made of brick units and mortar (Fig. 4). Thegeometry of the walls, the brick pattern, and the mortar composition reproducedthe load-bearing masonry walls of historical buildings (Fig. 5).

Timber beams were placed along the top of the first-story masonry walls (i.e., atthe intrados of each first floor; Figs. 5 and 6). These transverse timber beams (i.e.,directed perpendicular to the span), which collected the ends of the longitudinaltimber beams, were the ‘‘rema’’, which is a recurring component in historical build-ings. The width of each rema was 125 mm, i.e. half the thickness of the masonrywall it was inserted into (Fig. 7).

The first floor of each specimen (Fig. 6) was composed of eight parallel timberbeams placed 0.35 m apart (one-directional floor), labeled from B1 to B8. Each tim-ber beam had square cross-section with side of 100 mm (Fig. 6).

Each timber beam was accommodated into the entire thickness of the support-ing wall. Hence, each timber beam rested for 125 mm onto the rema and for125 mm onto the masonry wall (Figs. 4, 5 and 7).

A single-layer timber board was placed onto the beams in the transverse direc-tion. Hence, the boards were perpendicular to the beams they rested on (i.e., normalto the span; Fig. 5). Every timber board had thickness of 30 mm and was connectedto the beams with numerous nails, which prevented the former from sliding rela-tively to the latter (Fig. 6).

The two free edges of the boards (i.e., the two edges perpendicular to the walls)were not in correspondence with the edges of the two perimeter beams (i.e., B1 andB8), but the boards overhung the beams (Fig. 6).

The second-story masonry walls were built after having finished the first floor.In so doing, the masonry filled the gaps between the timber beams of the first floor,so that the beams resulted to be fixed into the walls.

The second floor of each specimen was identical to the timber system of the firstfloor (Fig. 8), i.e., beams and boards. However, the thickness of the second-storymasonry walls was 125 mm. Consequently, the second floor was provided with asupport different from the first floor, since the latter rested on 250-mm-thicknesswalls. For that reason, the rema was not placed into the second-story walls. Ulti-mately, the second floor rested directly on the masonry for a length of 125 mm(Fig. 9).

Once completed the second floor, timber wedges were driven between theextrados of the rema and the intrados of four first-floor-beams, on one sideof the floor of each specimen (Fig. 7). These wedges were meant to be progres-sively removed, so as to progressively modify the end-restraint of the beams.

2.08

1.82

0.19

SPECIMEN 1

0.19

2.08

1.82

SPECIMEN 2

2.85 2.85

0.250.13

0.13

2.482.74

2.482.74

0.250.13

0.130.25

0.13

0.13

0.250.13

0.13

Fig. 5. Geometric dimensions of the specimens, in m, and of the individual components. The representation also shows the terrazzo and the boards, including their direction(transverse to the span).

0.100.10

0.35B1

0.100.10

0.35

0.130.03

0.03

SPECIMEN 2

TIMBER BOARDS

TIMBER BEAMS

TIMBER BEAMS

VENETIAN TERRAZZO

(a)

(b)

SPECIMEN 1

B2 B3 B4 B5 B6 B7 B8

B1 B2 B3 B4 B5 B6 B7 B8

Fig. 6. Structure of the first floor of (a) specimen 1 (b) and 2. Geometric dimensions of the two floors, in m, and labels.


Therefore, the wedges made it possible to reduce the degree of restraint thatthe wall provided the beam ends with, which is the main effect of damageon a timber beam. In so doing, damage and its evolution could be reproducedin some tests.

At this point, while the first specimen was completed, the second specimen wasfinished pouring the terrazzo onto the first floor (Figs. 5 and 6). It is to note that theterrazzo was not poured onto the second floor (Figs. 8 and 9).

Ultimately, the two specimens had the same masonry structure, second floor,and timber structure, while they differed from one another in the presence of theterrazzo on the first floor (Figs. 5, 6 and 8).

The dimensions of each brick that composed all the walls were 250, 125, and55 mm. The courses of the brickwork that composed the first-story walls weremade of two wythes of bricks with the units running horizontally (i.e., stretcherbricks), bound together with bricks running transverse to the wall (i.e., headerbricks). The courses of the brickwork that composed the second-story walls weremade of one wythe of brick only (stretcher bricks).

The mortar of all the walls was composed of lime (3/12), Portland cement (1/12), sand (1/3) and water (1/3). The thickness of the horizontal and vertical mortarjoints was approximately 10 mm. They were made flush the brick’s face.

Masonry elasticity modulus was determined, although it was not involved inthe interpretation of the test results; it resulted to be 1767 N/mm2.

The timber elements were made of fir-tree. The elasticity modulus of the timberwas determined as well, since it is essential to interpret the results from the tests[2,7–9,16–18,26,38–40]. The elasticity modulus of the timber was obtained by per-forming compression and tension tests on six samples made of the same fir-treeused for the floors. The average value of the elasticity modulus obtained from thetests resulted to be 8454 N/mm2.

The terrazzo was built pouring three layers, one onto another. The lower layerwas the ‘‘sottofondo’’, whose thickness was 80 mm and whose unit weight was2.02 kN/m2. This layer rested onto the timber boards. The layer midway was the‘‘coprifondo’’, whose thickness was 35 mm and whose unit weight was 0.44 kN/m2. The upper layer was the ‘‘stabilitura’’, whose thickness was 15 mm and whoseunit weight was 0.26 kN/m2. Hence, the total thickness of the terrazzo was 130 mm(Fig. 6b), which is the typical terrazzo’s thickness of Venetian buildings, especiallywhen the timber system is flat (not warped). The composition of each layer and theunit weight of the components are summarized in Table 1.

The test included the determination of the terrazzo elasticity modulus too,which is essential in order to interpret the test results [41–44]. The elasticity mod-ulus of the terrazzo was obtained by performing compression tests on four cylinderspecimens with 200 � 500 mm dimensions, which were made and cast in the sameway as the terrazzo poured onto the floor. The elasticity modulus of the terrazzoresulted to be 1213 N/mm2.

Fig. 7. First floor: view of the end-restraint of the beams, including the rema (on thetop of the first-story masonry wall). Timber wedges inserted between the beams ofthe first floor and the rema.

0.100.10

0.35

0.03

SPECIMENS 1 AND 2

TIMBER BOARDS

TIMBER BEAMS

Fig. 8. Structure of the second floor of specimens 1 and 2, with the dimensions, inm. No differences existed between the two second floors; i.e., the specimensdiffered only in the first floors.

Fig. 9. View of the second floor: timber floor supported by the 125-mm-masonry-wall. The second floor of both the specimens was identical to the first floor ofspecimen 1.


3.2. Test set-up

Basically, the tests consisted of loading the first-story floor of the two specimensin the same way as one other and measuring the displacements of the structures.

Two types of loads were applied. The first load consisted of sacks of sand, whichwere laid onto the entire extrados of the first floor (Fig. 10), so as to obtain a uni-formly distributed load. The sacks were placed in three layers, which were laid ontothe floor in three steps; in so doing, three levels of loadings were obtained.

The second load consisted of a steel cable (tendon), which was applied at a dis-tance of 0.33 m from one of the supporting masonry wall (Fig. 11a) and at a pointlocated on the longitudinal axis of symmetry of the first floor, i.e., the axis midwaythe two central first-floor beams (i.e., B4 and B5; Fig. 11b). The tendon was ten-sioned by a hydraulic jack, so as to induce a concentrated load applied to the beamsB4 and B5. The bottom of the tendon was fixed to a steel beam, which, in turn, wasanchored to the foundation system of the laboratory (Fig. 11b). The top of the ten-don was fixed to a steel beam placed onto the extrados of the first floor; a hole hadbeen drilled through the timber boards, to let the tendon pass through. In order toobtain two localized loads, two bricks were placed between the steel beam and thetimber boards (Fig. 11). The steel beam was then fixed to the boards with fourscrews. In so doing, the test set-up consisted of a stiff beam that converted the forceof the tendon into two forces applied to the beams B4 and B5, respectively.

No loads were applied to the second floor, which hence contributed to the testswith its own weight only.

Displacement transducers were mounted on telescoping poles and placed at theintrados of the floors, to measure the deflection profile due to the loadings (Figs. 11and 12). Displacement transducers were installed at the common boundarybetween terrazzo and timber boards, to measure the relative displacements (slips)under load. Strain gages were installed on the end of the timber beams, to measurethe curvature of the beams at the end-restraints. Displacement, slip, and strainmeasures were taken during the entire tests.

The measuring equipment was connected to a data acquisition system, whichcontrolled the jack and recorded the measures from the displacement transducersand strain gages.

The data that were collected included the maximum vertical displacement(deflections) as well as the deflections of numerous points of the floors, which pro-vided many load–deflection relationships, the slips, which showed whether the ter-razzo and the timber system exchanged in-plane shear force or were independentof one another, and the curvature of the ends of the beams, which revealed theactual restraint that the walls provided the beams with.

The test included the survey of the cracks in the slab of terrazzo at each loadinglevel. Crack survey determined the density of cracks and the locations of cracks dueto each load [39–48].

3.3. Test methodology

As above-mentioned, the test methodology consisted of subjecting the twospecimens to the same test program. In so doing, the structural role of the terrazzocould be observed. The test program was composed of five stages, which differen-tiated from each other in the loading and/or the presence of the timber wedges. Thesame measures were taken during the five stages for the two specimens (Fig. 12).

Stage 1: The uniformly distributed loads (Subsection 3.2) were applied onto theentire first floor of each specimen (Fig. 10). The sacks of sand were laid so as to forma layer, then a second layer, finally a third layer. In so doing, three levels of uni-formly distributed load were applied to the first floor at stage 1 — namely, 1.0,2.0, and 3.0 kN/m2. This stage did not surpass the theoretical elastic limit of the tim-ber material.

Stage 2: At the end of stage 1, the concentrated load was applied to the first floor(Subsection 3.2). During the second stage, hence, the first floor of each specimenwas subjected to the action of the downwards force induced by the jack plus theuniform load of 3.0 kN/m2. The concentrated load was gradually increased fromzero up to 20.0 kN. This stage slightly surpassed the elastic limit of the timber,which theoretically was reached for 18.8 kN.

Stage 3: The first floor was unloaded; i.e., first the jack was unloaded and thenthe three layers of sacks of sand were removed from the extrados of the floor. Thisstage aimed at detecting the residual deflection (Fig. 12).

Stage 4: This stage and the last one were devoted to damage. Two of the timberwedges that had been placed between the beams and the rema were removed(Fig. 7) — namely, the wedges at one side of the two central beams of the first floor(i.e., under the beams B4 and B5). In so doing, the effects due to decay were simu-lated, since timber decay usually results in a reduction of the cross-section of theends of beams. The first floor was then reloaded, in the same was as stage 2.

Stage 5: The first floor was unloaded and two other wedges were removed fromthe ends of the beams — namely, the wedges at one side of the span of the beamsnext to the central ones (i.e., B3 and B6). The first floor was then loaded in the sameway as stage 2 (i.e., as stage 4).

3.4. Experimental results

The maximum deflections of a beam due to the loadings of stage 1 occurred atmidspan. The maximum deflection of a beam due to the loadings of stages 2, 4, and5 occurred at a point located between the midspan and the distance of 0.33 m fromthe right supporting wall (0.33 m was the distance of the force from the wall; Figs. 11and 12). In order to provide a synthetic representation of the floor’s behavior, the fig-ures presented herein do not show the maximum deflections of each beam due toeach load, since these deflections occurred at different positions from each other.Conversely, the figures show the deflections of the beams at a distance of 0.36 m fromthe midspan. In fact, it was proven that the transverse axis at a distance of 0.36 m fromthe midspan collects the deflections that are closer to the maximum for each beamunder each load than any other transverse axis. Ultimately, all the figures show thedeflections of the ‘‘reference points’’, which are the points of the beams that lie on atransverse axis passing at a distance of 0.36 m from the midspan.

Some of the experimental results are shown in Figs. 13–20. More specifically,Figs. 13–16 show the deflections of the representative points at each stage (i.e.,deflection versus position), while Figs. 17–20 for all the stages (load versus deflec-tion). In each figure, the graph ‘‘a’’ refers to specimen 1 and ‘‘b’’ to specimen 2; thecomparisons between the graphs ‘‘a’’ and ‘‘b’’ show directly the contribution of theterrazzo to the service–load deflections.

During stage 1, the beams exhibited only marginal differences in behavior fromeach other. During stages 2, 4, and 5, the deflections were reasonably symmetricwith respect to the longitudinal axis (the axis passing midway the beams B4 andB5). Therefore, the test results were deemed to be accurate.

Table 1Composition of the three layers of the terrazzo that was built onto the first floor of specimen 2 (the names of the layers are in Venetian vernacular). Unit weights of thecomponents of each layer, in kN/m2.

Sottofondo Coprifondo Stabilitura

Slaked lime Crushed bricks and tiles Slaked lime Powdered bricks Slaked lime Marble chips

0.42 1.62 0.11 0.33 0.12 0.14

Fig. 10. Uniformly distributed load on the first floor. The load was obtained withsacks of sand. Three layers of sacks were laid in three subsequent steps, so as toobtain three levels of load.


The measures of the slips collected for specimen 2 (with terrazzo) were elabo-rated to obtain information about the composite behavior of the structure. Theseelaborations proved that the slab of the terrazzo shifted significantly with respectto the timber boards, which suggested that no significant composite action existedin this floor.

Crack survey showed that, during all the stages, the terrazzo exhibited only fewand short cracks, which opened at the bottom of the terrazzo slab, whose depth wasmuch lower than half the thickness of the slab. This result suggested that the ter-razzo slab did not reduce its stiffness during loadings.

1.00

0.33

(a)

SECT

ION

HYDRAULIC JACK

TRANSDUCER

Fig. 11. Test set-up: (a) frontal view, (b) and cross-section. The representation also showthe top steel beam, and the bottom anchorage steel beam.

The experimental strain showed that, during all the stages, the ends of thebeams did not exhibit any considerable strain, which suggested that the beamsrotated freely relatively to the masonry walls they rested on. Accordingly, theend-restraint was always a hinge.

No substantial residual deformations were observed. In particular, the deflec-tions measured at stage 3 were practically zero, which suggested that the loadingshad induced marginal or no inelasticity. Also the deflections measured during stage5, before applying the new load, were substantially zero.

The comparisons of the deflection profiles due to the concentrated loads only(obtained by subtracting the contributions of the uniform loads from the experi-mental deflections measured at stages 2, 4, and 5) showed that the curvatures inthe transverse direction of the floor without terrazzo were significantly greater thanthose of the floor with it. This result suggested that the terrazzo significantlyincreased the load-sharing mechanism between the timber beams.

3.5. Discussion

Under the same load, the floor’s deflections of specimen 1 (that without ter-razzo) drastically surpassed those of specimen 2 (that with terrazzo). Since thetwo specimens differed from one another only for the terrazzo, this comparisonproved that the terrazzo had a significant structural role.

More specifically, the deflections of the representative point due to (stage 1) anuniformly distributed load of 1.00, 2.00, and 3.00 kN/m2, respectively, were 0.98,2.11, and 3.38 mm for specimen 2, and 3.83, 7.61, and 11.58 mm for specimen 1.Thus, the terrazzo enabled the first floor to reduce the deflections due to a uniformload by approximately 3.5 times.

Moreover, the deflections of the representative point due to (stage 2) the uni-formly distributed load of 3.00 kN/m2 and a concentrated load of 5.00, 10.00, and20.00 kN, respectively, were 3.81, 4.54, 6.02 mm for specimen 2, and 13.51, 16.23,22.84 mm for specimen 1. Thus, the terrazzo enabled the floor to reduce the deflec-tion due to a uniform load plus a concentrated load by more than 3.5 times.

0.350.20 0.20

1.25 1.25

SECT

ION

STEEL BEAMBRICKS

LOAD CELL

TRANSDUCER

HYDRAULIC JACK

STEEL BEAM

TELESCOPING ROD

T1 T2 T3 T4 T5 T6 T7 T8

(b)

B2 B3 B4 B5 B6 B7 B8B1

0.35 0.35 0.35 0.35 0.35 0.35

s the jack, including its bottom and top anchorages, some displacement transduces,

Fig. 12. Displacement transducers used to measure the deflections of the timberbeams and of the floor.


This result was expected from the qualitative perspective only, while from thequantitative perspective the differences in deflections were much more thanexpected. In fact, on one hand, the stiffness and strength of the portion of ter-razzo included into the tributary area of a timber beam were considerable. Onthe other hand, however, the terrazzo was not connected to the timber structureby any mechanical element. In particular, no stud shear connectors or bars hadbeen used, and thus the connection between the terrazzo and the timber boardswas provided only by interlocking and friction, which were marginal[11,12,21,24–26,36,49]. Moreover, the tensile strength of the terrazzo was low[41–48,50].

The deflection data collected during the load tests were also analyzed in orderto evaluate the deviations from linearity of the load–deflection relationship. Thisanalysis showed that both the floors exhibited substantial linear behavior, whichwas consistent with the marginal residual deflections that had been found at stage3. More specifically, the deflections of specimen 2 (with terrazzo) deviated from lin-earity more than those of specimen 1, but these deviations were however minor.Considering that specimen 1 was without terrazzo, non-linearity of specimen 2was attributed to some inelasticity exhibited by the terrazzo. Considering that toexhibit the greatest deviations from linearity were the maximum deflections whilethe other deflections exhibited deviations much lower, nonlinearity was attributedto local effects because of imperfections related to load distribution, shape of thestructure, and support conditions [50–52].

In order to gain insight into the composite behavior, the experimental slipswere compared to the slips derived from modeling the structure as layered withoutconnections [49,53]. The comparison showed that the interface between terrazzoand timber reduced the relative displacements only marginally, i.e., the slipsapproached the layered limit. Therefore, no considerable in-plane shear stresseswere transferred at the common boundary between terrazzo and boards.

Ultimately, no composite action existed in the floor and therefore the behaviorof the floor was the result of the bending stiffness of the timber system and the ter-razzo only, but not of the coupling between them. That result was consistent withthe cracks observed in the terrazzo, which were few, short, and not deep; i.e., thecracks did not reduce the stiffness of the slab.

With concentrated load (stage 2) applied to the two central beams, the differ-ences in deflections between the two central beams and the other beams of speci-men 1 (Fig. 14a) were much greater than those of specimen 2 (Fig. 14b), at anygiven distance from the wall. For instance, under a force of 20.0 kN (plus the uni-form load), the difference in maximum deflections between a central beam and aperimeter beam was 12.6 mm for specimen 1, while it was 1.85 mm for specimen 2.

That result proved that the terrazzo had another structural role — namely, itprovided the floor with significant two-dimensional behavior. Accordingly, the ter-razzo increased the stiffness of the structure not only in the longitudinal directionof the floor but also in the transverse direction.

The deflection profiles from the tests were analyzed using the model of contin-uous elastic beam resting on elastic supports (i.e., linear springs). In that model, thebeam reproduced the transverse behavior of the floor, while the springs reproducedthe timber beams. Since all the geometric and mechanical parameters were knownexcept for the width of the ideal beam, this parameter was obtained by calibratingthe theoretical deflections against the test deflections. The best agreement betweenexperimental and theoretical deflection profiles was obtained for a width of theideal beam equal to half the span of the floor.

Ultimately, a point of the floor with terrazzo deflected much less than the samepoint of the floor without terrazzo; i.e., the ratio between the two deflections wasmuch less than unity for all the points of the floor. While under uniform load theratio between the two deflections did not depend on the point, under concentratedload the ratio strongly depended on the point; the greater its distance from theforce the lower the ratio, and vice versa. Hence, the timber–terrazzo floor had sig-nificant two-dimensional composite behavior, which was provided by the bendingstiffness of the terrazzo slab, although no in-plane (longitudinal) shear actions weretransferred between the terrazzo and the timber structure.

Since specimen 2 reproduced the typical terrazzo flooring, the above results canbe generalized to the historical floors with terrazzo.

The analysis of the test results obtained at stages 4 and 5 allowed the structuraleffects of damage to be detected.

The specimens after the removal of the timber wedges from the two centralbeams (stage 4) did not exhibit significant differences in behavior from the stagebefore (stage 2). More specifically, the deflections of stage 4 (Figs. 15 and 19) werelarger than those of stage 2, but the increases in deflection were moderate for spec-imen 1 (Figs. 14a, 15a, 18a and 19a) and minor for specimen 2 (Figs. 14a, 15b, 18aand 19b). The results obtained during stage 5 were almost the same as stage 4(Figs. 16 and 20).

Since the curvatures at the end of the timber beams were almost zero alreadyduring stage 1 and 2, the increase of deflections observed during stage 4 and 5was not due to any downgrading of the end-restraint. Thus, the deflectionsincreased because the reduction of the beam ends shifted the center of the contactstresses exchanged by wall and beam towards the external of the wall cross-sec-tion, which increased the span. This interpretation was supported by the data,which confirmed that the increase of the beam span caused the beam’s deflectionsto increase approximately of the values that had been measured.

The moderate to marginal increases in deflections observed at stages 4 and 5proved that the initial connection, i.e., at stage 1 and 2, already behaved as a hinge.Ultimately, the connection between the timber beams and the masonry walls werehinges both with and without the wedges.

That result was consistent with the fact that the ends of the timber beamsexhibited significant rotations during the loadings already at stages 1 and 2.

Considering that the end-restraints of the specimens reproduced the typicalconnection of a timber floor to the supporting walls in historical buildings, theexperimental results support two general conclusions.

The typical end-restraint of a timber floor, with or without terrazzo, can bemodeled as a hinge, which is simultaneously a conservative and a realisticassumption.

Damage to timber beams [2,7,9,11,17,18,40], whose main effect is a reduction ofthe end cross-section, reduces moderately the stiffness of the timber floors andmarginally the stiffness of timber–terrazzo floors. Nevertheless, the reduction incross-section due to damage implies a reduction in shear strength and in the capac-ity of the timber beam to connect the opposite walls it rests on; moreover, damagecan propagate towards the midspan [6,8,16,33,50,52]. Therefore, damage has to bedetected and eliminated [10,12,15,28–30]. However, according to the test results,the increase in deflection is not the way to detect damage to timber floors, espe-cially if they present the terrazzo flooring.

4. Analytical modeling

The test results were used to construct an analytical model thatpredicts the service–load deflection and stress behavior of timberfloors finished with terrazzo flooring. The mechanical assumptionsof the model were derived directly from the results of the labora-tory experiments (Section 3).

The model presented here was also used to predict the deflec-tions and stresses of timber floors without terrazzo, as specimen1, although the novelty of this model only lies in its ability todescribe the behavior of the timber–terrazzo floor. However, sim-ulating the two laboratory experiments with the same mathemat-ical model allowed unbiased interpretations of the observedbehaviors to be obtained.

The reference structure is composed of longitudinal timberbeams spaced a given distance apart from each other, transversetimber boards with no gap in between one another, and terrazzoslab onto the boards (Figs. 21 and 22). The reference structure is

timber beam

verti

cal d

ispl

acem

ent (

mm

)

timber beam

verti

cal d

ispl

acem

ent (

mm

)

Distributed load

(b)(a)

B1 B8B2 B3 B4 B5 B6 B7

-10

-12

-8

-6

-4

-2

0

-1

-3

-5

-7

-9

-11

SPECIMEN 1

B1 B8B2 B3 B4 B5 B6 B7

-10

-12

-8

-6

-4

-2

0

-1

-3

-5

-7

-9

-11

SPECIMEN 2

1 kN/m

2 kN/m

3 kN/m

2

2

2

Fig. 13. Stage 1: (a) specimen 1; (b) specimen 2. Vertical displacements (deflections) of the reference points (which lie on the transverse axis passing 0.36 m from themidspan), obtained from the tests. The reference points do not provide the strict specific maximum deflections of the first-floor beams under the load that is considered, butallow the figures of the various stages to provide the deflections of the same points of the first-floor beams (which however are very close to the maximum deflection of theconsidered beam and load).

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

7 kN6 kN5 kN4 kN3 kN2 kN1 kN0 kN

13 kN12 kN

10 kN11 kN

19 kN18 kN17 kN16 kN15 kN14 kN

20 kN

9 kN8 kN

B1 B8B2 B3 B4 B5 B6 B7timber beam

verti

cal d

ispl

acem

ent (

mm

)

SPECIMEN 2

3 kN/m2

Distributed load

0

-1

-2

-3

-4

-5

-6

-7


12 kN

10 kN11 kN

9 kN8 kN

13 kN


20 kN

(b)

-8

(a)Fig. 14. Deflections of the reference points from the tests, at stage 2: (a) specimen 1; (b) specimen 2.


symmetric with respect of the longitudinal axis passing throughthe center of the floor (i.e., to the axis midway the two centralbeams). Nails connect the boards to the beams, which preventany considerable sliding of the former relatively to the latter(monolithic behavior of the timber system). Conversely, no ele-ments connect the terrazzo to the timber system; i.e., the formeris not restrained to the latter. The reference structure is subjectedto vertical static load, which can be uniformly distributed and/orconcentrated (Figs. 21 and 22).

The diagram of the reference structure (Figs. 21 and 22) refersto specimens 1 and 2, and the formulation presented herein repro-duces directly the geometry of the floors that were tested —namely, eight timber beams, symmetric with respect to the axis

midway the beams B4 and B5, subjected to symmetric loads. Nev-ertheless, the reference structure is general and therefore the for-mulation can be extended to floors composed of a differentnumber of timber beams and to any type of loads.

Modeling the timber–terrazzo floor consists of addressing threebasic issues — namely, (1) how the timber floor and the terrazzoslab work together in the longitudinal and transverse directions,(2) whether a wall prevents the end rotation of the timber beamthat rests on it or not and, in case, up to what extent, (3) andhow the terrazzo provides the floor with stiffness.

Regarding the first issue, the experimental results showed thatno considerable in-plane shear action was transmitted through thecommon boundary between terrazzo and boards; i.e., no

verti

cal d

ispl

acem

ent (

mm

)

verti

cal d

ispl

acem

ent (

mm

)

3 kN/m2

Distributed load

(b)(a)

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

-24

SPECIMEN 1


13 kN12 kN

10 kN11 kN


9 kN8 kN

20 kN

B1 B8B2 B3 B4 B5 B6 B7

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7


12 kN

10 kN11 kN

9 kN8 kN

13 kN


20 kN

-8

timber beam timber beam

Fig. 15. Deflections of the reference points from the tests, at stage 4: (a) specimen 1; (b) specimen 2.

verti

cal d

ispl

acem

ent (

mm

)

3 kN/m2

Distributed load

(b)(a)

timber beam timber beamB1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

-24

SPECIMEN 1


13 kN12 kN

10 kN11 kN


9 kN8 kN

20 kN

B1 B8B2 B3 B4 B5 B6 B7

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7


12 kN

10 kN11 kN

9 kN8 kN

13 kN


20 kN

-8

verti

cal d

ispl

acem

ent (

mm

)

Fig. 16. Deflections of the reference points from the tests, at stage 5: (a) specimen 1; (b) specimen 2.


appreciable composite action existed. That finding, which held truefor both the longitudinal and transverse directions and for bothuniform and concentrated loads, was consistent with the fact thatno connectors joined the terrazzo to the timber system, since theformer had been simply poured onto the latter. Moreover, frictionstrength was almost zero, since the transverse forces were irrele-vant in the floor. Thus, the shear action was transmitted only bythe interlocking mechanism, whose capacity was marginal [51].

Since this finding is comprehensive, it can be generalized andapplied to any timber–terrazzo floor. Hence, the model has to ignoreany transmission of in-plane shear action between the boards andthe terrazzo (layered behavior of the timber–terrazzo compositefloor). Moreover, the model has to consider that, on one hand, timberbeams and timber boards are connected to each other, but on theother hand, beams and boards are perpendicular to one another.

Regarding the second issue, the experimental results showedthat the ends of the timber beams rotated freely with respect tothe supporting walls, since the fibers of the former did not exhibitany significant strain at the ends (i.e., the curvatures of the beamswere marginal at the supports). That finding was consistent withthe fact that the removal of the wedges from the supports didnot increase considerably the deflection of the two floors, whichproved that the end-restraints of the timber beams behaved ashinges already in the no-damage state. Ultimately, each connectionprovided the end of the timber beam with vertical force, but withno significant couple.

Since this finding is comprehensive, it can be generalized andapplied to any timber–terrazzo floor. Hence, the model has toreproduce the connection between the end of a beam and the sup-porting wall with a hinge.

0

1

2

3

0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13

SPECIMEN 1

0

1

2

3

0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13

SPECIMEN 2

(a) (b)

Beams

B2B3B4

B6B7B8

B5

B1

dist

ribut

ed lo

ad (k

N)

vertical displacement (mm) vertical displacement (mm)

dist

ribut

ed lo

ad (k

N)

Fig. 17. Load–deflection curves during stage 1. Abscissa: deflection of the reference point. Ordinate: uniform load applied to the first floor. (a) Specimen 1; (b) specimen 2.

conc

entra

ted

load

(kN

)

0

2

4

6

8

10

12

14

16

18

20

22

0 -2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26

SPECIMEN 1

vertical displacement (mm)

(a)

0

2

4

6

8

10

12

14

16

18

20

22

0 -8-1 -2 -3 -4 -5 -6 -7

SPECIMEN 2

(b)


Beams

B2B3B4

B6B7B8

B5

B1

B4 B5 B6 B7 B81B 2B 3BB4 B5 B6 B7 B81B 2B 3B

conc

entra

ted

load

(kN

)

Fig. 18. Load–deflection curves during stage 2. Abscissa: deflection of the reference point. Ordinate: concentrated load applied to the first floor. (a) Specimen 1; (b) specimen2. The uniform load at the end of stage 1 was maintained during stage 2. Therefore, the initial abscissa of each curve is equal to the displacement reached at the end of stage 1.


Regarding the third issue, the experimental results showed thatthe terrazzo followed the deflections of the timber floor it restedon without exhibiting any significant cracks. More specifically, thecracks were small even when the floor reached the timber elasticlimit. Hence, the structure found equilibrium before that the cracksin the terrazzo propagated significantly, and therefore cracking didnot reduce appreciably the bending stiffness of the terrazzo slab.Moreover, the elastic deflections of the timber floor induced com-pressive stresses in the terrazzo that were much lower not only thanits compressive strength but also than its elastic limit. Ultimately,the terrazzo provided the floor system with its full elastic bendingstiffness, in the longitudinal and transverse directions. That findingwas consistent with the deflections of the floor finished with ter-razzo, which were remarkably lower than those without it. More-over, that finding was also consistent with the marginal residualdeflection found at stage 3 and at the beginning of stage 5.

Regarding the third issue, moreover, the elaboration of theexperimental results showed that the transverse behavior of the

floor is governed by an imaginary strip whose width (tributaryarea) is half the floor span, which rests on linear springs that repro-duce the stiffness of the timber beams.

Since these findings are comprehensive, they can be generalizedand applied to any timber–terrazzo floor. During the elastic deflec-tion of the timber elements, hence, the model has to take intoaccount the full elastic bending stiffness of the tributary areas inboth the longitudinal and transverse directions of the terrazzo slab.

Ultimately, the terrazzo slab as a standalone building compo-nent would be inadequate; in particular, if it were used as a floor,it wouldn’t guarantee any stiffness and load-carrying capacity.Conversely, the terrazzo as construction material of the floor sys-tem helps the timber structure carry the loads, which increasesthe stiffness and the load-carrying capacity of the floor system.

Accordingly, the terrazzo slab can be envisioned as a grid ofadjacent and orthogonal beam strips interconnected continuouslyalong their length, which only bending occurs in, while twistingis minor. The width of the imaginary longitudinal strips is defined

conc

entra

ted

load

(kN

)


(a) (b)


Beams

B2B3B4

B6B7B8

B5

B1


0

2

4

6

8

10

12

14

16

18

20

22

0 -2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26

SPECIMEN 1

0

2

4

6

8

10

12

14

16

18

20

22

0 -8-1 -2 -3 -4 -5 -6 -7

SPECIMEN 2

Fig. 19. Load–deflection curves during stage 4. Abscissa: deflection of the reference point. Ordinate: concentrated load applied to the first floor. (a) Specimen 1; (b) specimen 2.

conc

entra

ted

load

(kN

)


(a) (b)


Beams

B2B3B4

B6B7B8

B5

B1


0

2

4

6

8

10

12

14

16

18

20

22

0 -2 -4 -6 -8 -10 -12 -14 -16 -18 -20 -22 -24 -26

SPECIMEN 1

0

2

4

6

8

10

12

14

16

18

20

22

0 -8-1 -2 -3 -4 -5 -6 -7

SPECIMEN 2

Fig. 20. Load–deflection curves during stage 5. Abscissa: deflection of the reference point. Ordinate: concentrated load applied to the first floor. (a) Specimen 1; (b) specimen2.


by the tributary area of each beam. The width of the imaginarytransverse strips is equal to half the span of the longitudinal strips;therefore, the transverse strips overlap each other.

Ultimately, the timber–terrazzo composite floor behaves as agrid of strips, interconnected to each other at fictitious nodes. Eachstrip, which is composed of timber and terrazzo with the two com-ponents positioned directly one above another (with their respec-tive centroids vertically above each other), has a certain elasticbending stiffness while its twisting stiffness is negligible. More-over, the two components do not transmit any in-plane shearstress to one another.

The formulation also includes the damage state of the floor, sothat the model possesses all the attributes necessary to providethe designers with the means of predicting the behavior of any his-torical timber–terrazzo floors. The experimental results showedthat damage did not reduce the ability of the end-restraint totransmit bending moment between the timber beam and the sup-porting masonry wall, since this end-restraint consisted of a hingealready in the undamaged state. Nonetheless, the experimentalresults showed that the reduction of cross-section at the end of abeam due to damage modified the mutual contact between thebeam and the supporting wall, which consequently caused the

q

VBn

L2_BnL1_Bn

x

LBn

q

VBn

L2_BnL1_Bn

x

LBn

q

x

q

x

L2_BnL1_Bn

LBn

VBn

VBn

L2_BnL1_Bn

LBn

VBn

SPECIMEN 1 SPECIMEN 2

(c) (d)

(e) (f)

q

x

LBn

q

x

LBn

VBn(a) (b)

Fig. 21. Reference structure: longitudinal section. Diagrams of the floor without terrazzo (timbers and boards; diagrams a, c, e) and with terrazzo (on the boards; diagrams b,d, f). The diagrams also show the walls that the beams rest on. The origin of the abscissa x is taken on the inner surface of the left masonry wall. The figure shows theapplication of the model to reproduce stage 1 (uniform load; diagrams a, b), stage 2 (uniform and concentrated loads; diagrams c, d), and stages 4 and 5 (the same loads asdiagram c, d, and the right support shifted towards the outer surface of the right wall, which increases the span; diagrams e, f).

q

q

q

q

B1 B2 B3 B4 B5 B6 B7 B8

B1 B2 B3 B4 B5 B6 B7 B8

B1 B2 B3 B4 B5 B6 B7 B8

B1 B2 B3 B4 B5 B6 B7 B8

q

(a) (b)

(c) (d)


VB VBVB VB

Fig. 22. Reference structure: transverse section. Diagrams of the floor without terrazzo (beams and boards; diagrams a, c) and with terrazzo (on the boards; diagrams b, d).The diagrams also show a wall that the beams rest on (shadowed horizontal line underneath the beams). The figure shows the application of the model to reproduce stage 1(diagrams a, b), and stages 2, 4, 5 (diagrams c, d).


vertical force that the wall provided the beam with to shift towardsthe outer edge of the wall cross-section. That finding was consis-tent with the deflections collected during the stages 4 and 5.

Since this finding is comprehensive, it can be generalized andapplied to any timber–terrazzo floor. Hence, the model has toreproduce the progression of damage in a timber beam by progres-sively increasing the span of the beam, from the initial value up to

reaching the outer span (i.e., the inner span plus twice the wallthickness).

The experimental results were reproduced by means of fourmodeling assumptions which allowed the structural behavior tobe reproduced — namely, (1) The floor is composed of longitudinalstrips, whose width is equal to the spacing of the beams, and trans-verse strips, whose width is equal to half the span of the beams,


which only elastic bending occurs in, while twisting is zero; (2) Thestrips of the grid, which are composed of timber and terrazzo, donot transmit any in-plane shear action between the timber systemand the terrazzo slab through the common boundary, so that nocomposite action exists in the structure; (3) The end-restraints ofthe timber beams are hinges; (4) Damage shifts the applicationpoint of each reaction force underneath the ends of the timberbeams toward the outer edge of the wall cross-section.

Those assumptions describe the behavior of the timber–terrazzofloor up to the limit of elasticity of the structure, which is dictated bythe timber system, since non-linearity due to the terrazzo is moder-ate, although it starts before than nonlinearity of the timber system.

The behavior of the nth-beam of the floor according to theassumptions is described in the diagrams of Figs. 21 and 22.

According to the assumptions, the damaged state of the nth beamis modeled by increasing its length LBn (Fig. 21e and f). The applica-tion of the model to the laboratory tests increased the span onlytowards what in Fig. 21 is the right support, since in the tests damagewas applied only to this side of the beams. In these applications,hence, the distance LBn 2 between the right support and the verticalforce of the nth damaged timber beam was increased by the length ofthe wedge (half thickness of the wall). Conversely, the distance LBn 1

(i.e., between the left support and the force) was not changed, sinceno wedge had been placed under the left support of the testedbeams. When the model is used to reproduce a floor whose timberbeams are affected by damage on both the sides, the span of thedamaged beams has to be increased towards both the supports.

The fraction of the distributed load that acts onto the nth beam,which depends on the tributary areas, is denoted by q (Figs. 21 and22). The fraction of the concentrated vertical load VB (Fig. 22c andd) that acts on the nth beam, which depends on the tributary areasand the stiffness of the components, is denoted by VBn (Fig. 21c–f);i.e., the force VBn is the fraction of the force VB that is applied to thenth beam.

According to assumption 1, the deflection of a generic point of thenth beam that belongs to the piece of beam of length LBn 1 (Fig. 21)can be calculated as the sum of the deflection due to the distributedload q and concentrated load VBn that act on the longitudinal strip:

gnxLBn 1¼

q � x4

24�q � LBn � x3

12þq � L3

Bn � x24

E � J þ

VBn � LBn 2

6 � LBn� L2

Bn � x� L2Bn 2 � x� x3

� �

E � Jð1Þ

Likewise, the deflection of a generic point of the nth beam thatbelongs to the piece of beam of length LBn 2 (Fig. 21) is given by:

gnxLBn 2¼

q � x4

24�q � LBn � x3

12þq � L3

Bn � x24

E � J

þ

VBn � LBn 2

6 � LBn� L2

Bn � x� L2Bn 2 � x� x3

� �þVBn

6� x� LBn 2ð Þ3

E � J ð2Þ

In Eqs. (1) and (2), x denotes the distance of the point that isconsidered from the wall that supports the left end of the nth beam(Fig. 21).

Moreover, E and J denote the elasticity modulus and the cross-section moment of inertia, respectively. The term E�J is equal toETJB + ETzJTz, where ET and ETz are the elasticity modulus of the tim-ber and terrazzo, respectively, and where JB and JTz are the cross-section inertia moment of the beam and terrazzo, respectively.More specifically, JB is the moment of inertia of the timber beamsection with respect to the centroidal transverse axis; it does notinclude the timber boards, since the boards are orthogonal to thebeams (i.e., the boards’ fibers are directed along the direction thatis orthogonal to the direction of the beams’ fibers). Moreover, the

inertia moment JTz uses the tributary area of the terrazzo; thus,JTz is the inertia moment of a rectangular section whose height isthat of the terrazzo and whose width is equal to the distance ofone beam from another, denoted by lwb (in the tests, the contiguousbeams were at the same distance from one another). In the partic-ular case of floor without terrazzo (specimen 1) E�J is equal to ETJB.

The force VBn is the unknown of Eqs. (1) and (2), whose obtain-ment allows the deflection g, in turn, to be obtained. According tothe grid behavior (assumptions 1), VBn can be derived referring to acontinuous elastic beam resting on elastic supports, i.e., linearsprings (Fig. 23). The springs reproduce the timber beams accord-ing to the longitudinal strip model; the elastic beam reproducesthe transverse behavior of the floor according to the transversestrip model too. The width of the beam cross-section is equal tohalf the span of the floor (assumption 1). According to assumption2, the beam is composed not only of terrazzo but also of timberboards. In fact, the boards are transverse and therefore the fibersof the boards are transverse too.

Ultimately, the moment of inertia of the elastic beam on springsis the sum of the moment of inertia of the terrazzo cross-sectionand of the board cross-section. The width of each cross-section isequal to half span; the thickness of each cross-section is equal tothe actual thickness of the component.

The beam resting on spring model involves two unknowns —namely, the moments M in the beams at the intersection withthe supports (springs), and the deflection d of the supports(Fig. 23a and b).

The symmetry exhibited by the structure and the absence ofmoment at the first and last intersections reduce the number ofunknowns from 16 to 7 (the formulation is developed for a floorwith 8 beams), which are the bending moments at the intersec-tions with the supports from 2 to 4, denoted by MB2, MB3, MB4,and the deflections of the supports from 1 to 4, denoted by dB1,dB2, dB3, dB4. More specifically, the symmetry condition implies thatdB1 ¼ dB8, dB2 ¼ dB7, dB3 ¼ dB6, dB4 ¼ dB5 and that kB1 ¼ kB8, kB2 ¼ kB7,kB3 ¼ kB6, kB4 ¼ kB5 (Fig. 23a and e; b and f). Moreover, the no-twisting assumption implies that MB1 = MB8 = 0.

The equilibrium and angular compatibility conditions allow the7 unknowns to be found. The 7 equations necessary to obtain the 7unknowns are herein formulated using the displacement method,where a vertical displacement is assumed to be positive if it isdirected downwards and a rotation is assumed to be positive if itis clockwise:

dB1 ¼MB2

kB1 � lwbð3Þ

dB2 ¼MB3 � 2 �MB2

kB2 � lwbð4Þ

dB3 ¼MB2 � 2 �MB3 þMB4

kB3 � lwbð5Þ

dB4 ¼VB

kB4þMB3 �MB4

kB4 � lwbð6Þ

dB2 � dB1

lwbþ lwb

6 � Ef � Jf� �2 �MB2ð Þ

¼ � dB2 � dB3

lwbþ lwb

6 � Ef � Jf� 2 �MB2 þMB3ð Þ ð7Þ

dB3 � dB2

lwbþ lwb

6 � Ef � Jf� �2 �MB3 �MB2ð Þ

¼ � dB3 � dB4

lwbþ lwb

6 � Ef � Jf� 2 �MB3 þMB4ð Þ ð8Þ

kB1

vB1

lwb

B8B7B6B5B4B3B2B1

MB2MB3 MB4 MB5 MB6

MB7 MB2MB3 MB4 MB5 MB6

MB7

B8B7B6B5B4B3B2B1

lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb

lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb lwb

kB2 kB3 kB4 kB5 kB6 kB7 kB8 kB1 kB2 kB3 kB4 kB5 kB6 kB7 kB8

vB2 vB3 vB4 vB5 vB6 vB7 vB8 vB1 vB2 vB3 vB4 vB5 vB6 vB7 vB8

(a) (b)

(c) (d)


(e) (f)

VB VBVB VB

VB VBVB VB

VB VBVB VB

Fig. 23. Reference structure: transverse structural behavior of the floor. The distribution of the concentrated loads VB to the beams provided by boards and terrazzo ismodeled as a transverse strip that intersects the longitudinal strips that model the beams. Floor without terrazzo (diagrams a, c, e) and with terrazzo (diagrams b, d, f).(Diagrams a, b): the unknowns are the deflections d and the bending moments M that occur at the intersections between the transverse strip and the longitudinal strips.(diagrams c, d): forces VB that the transverse strip distributes to the longitudinal strips according to the transverse behavior of the structure. (diagrams e, f): model of thetransverse behavior: continuous elastic beam (transverse strip) composed of either boards (diagram e) or both boards and terrazzo (diagram f) resting on elastic supports(linear springs) which model the beams (longitudinal strips).


dB4 � dB3

lwbþ lwb

6 � Ef � Jf� �2 �MB4 �MB3ð Þ ¼ lwb

6 � Ef � Jf� 3 �MB4ð Þ ð9Þ

In the Eqs. ((3)–(9)), lwb denotes the distance between two adja-cent beams (Fig. 23c–f), and kB1, kB2, kB3, kB4 are defined by the fol-lowing expressions (Fig. 23e and f), which take into account thatthe end-restraints of the beams are hinges (assumption 3):

kB1 ¼3 � ET � JB � LB1

L21 B1 � L

22 B1

ð10Þ

kB2 ¼3 � ET � JB � LB2

L21 B2 � L

22 B2

ð11Þ

kB3 ¼3 � ET � JB � LB3

L21 B3 � L

22 B3

ð12Þ

kB4 ¼3 � ET � JB � LB4

L21 B4 � L

22 B4

ð13Þ

According to assumption 4, the effects of damage are repro-duced by using the spans in the damaged condition, which islonger than those in the undamaged condition. Ultimately, usingthe original span, Eqs. ((3)–(13)) model the undamaged conditionof the floor, while using the span in the damaged conditions theequations model the effects of damage.

The vertical deflections dBn of the nth beam provided by Eqs.((3)–(9)) allows the concentrated vertical load VBn shared by thenth beam to be determined:

VBn ¼ kBn � dBn ð14Þ

where kBn is the stiffness of the beam provided by Eqs. ((10)–(13)).Finally, the values of VBn provided by (14) have to be plugged

into Eqs. (1) and (2). In so doing, the deflection g of any point ofthe floor with any pattern of loading can be obtained.

The internal actions can be obtained by calibrating a stress–strain model against the deflections obtained with this model.The model calibration guarantees that, when operating in the lin-ear range, the accuracy of such a new model is almost the sameas that of this model.

Since the model is analytical, it also shows the influences of theterrazzo on the deflections of the floor as well as the role of the ter-razzo in the distribution of concentrated loads between the timberbeams.

5. Application of the analytical model to the tests

The analytical model was used to run simulations of the labora-tory tests. The elasticity modula that had been obtained from thetests of timber specimens and terrazzo specimens (see Subsection3.1) were plugged into the model, which was then solved for thevalues of q and VB of stages 1 and 2 (see Subsection 3.3).

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(a)

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(b)

3 kN/m2

Distributed load

(experimental test)3 kN/m2

(analitical modeling)


13 kN12 kN

10 kN11 kN

18 kN17 kN16 kN15 kN14 kN

9 kN8 kN

19 kN20 kN

5 kN

10 kN

15 kN

20 kN

Fig. 24. Specimen 1; stage 2. (a) Deflections of the reference points obtained from the analytical model. (b) Comparison between the analytical predictions and theexperimental deflections measured for the reference points during stage 2.

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(a)

B1 B8B2 B3 B4 B5 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(b)

3 kN/m2

Distributed load




13 kN12 kN

10 kN11 kN

18 kN17 kN

15 kN14 kN

9 kN8 kN

19 kN20 kN

16 kN

5 kN

5 kN

5 kN

5 kN



The model was also used to run simulations of the damagedstate. In these simulations, the span of the timber beams thathad been subjected to artificial damage during stages 4 and 5(i.e., B3, B4, B5, and B6; see Subsection 3.3) was increased withrespect to the original value, according to the modeling assump-tions (see Section 4).

The analytical results are shown in Figs. 24–29. More specifi-cally, in their part a Figs. 24–26 reproduce the tests of specimen1, and Figs. 27–29 of specimen 2. In their part b, these figures pres-ent the analytical results together with the experimental resultsreproduced analytically, so as to provide directly the comparisonsbetween test and model. It is to note that the model is analyticaland thus the formulation depends only on the representativeness

of the assumptions. Thus, those comparisons only verify that theassumptions synthesize accurately and exhaustively the mechani-cal behavior of the timber–terrazzo floor.

The comparisons between the results from the analytical modeland from the tests show that the model reproduces accurately allthe displacements of the specimens in the undamaged condition.Not only is this true for the one-dimensional deflections inducedby uniformly distributed loads, but this is also true for the two-dimensional deflections induced by concentrated loads. More pre-cisely, the agreement between analytical and experimental curvesfor stages 1 and 2 is good for the points away from the maximumdeflection, whereas it is fair for the maximum deflection as well(Figs. 24–29).

B1 B8B2 B4 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(a)

-27

-28

B1 B8B2 B4 B6 B7

-20

-26

-16

-12

-8

-4

0

-2

-6

-10

-14

-18

-22

timber beam

verti

cal d

ispl

acem

ent (

mm

)

-24

SPECIMEN 1

(b)

-27

-28

3 kN/m2

Distributed load

(experimental test)

3 kN/m2



13 kN12 kN

10 kN11 kN


20 kN

9 kN8 kN

5 kN

5 kN

10 kN

15 kN

20 kN

15 kN

20 kN

10 kN

B3 B5 B3 B5



verti

cal d

ispl

acem

ent (

mm

)

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7

(a)

-8B1 B8B2 B3 B4 B5 B6 B7

timber beam

verti

cal d

ispl

acem

ent (

mm

)

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7

(b)

-8

13 kN12 kN

10 kN11 kN


9 kN8 kN

20 kN7 kN6 kN5 kN4 kN3 kN2 kN1 kN0 kN

5 kN

5 kN

10 kN

15 kN

20 kN

10 kN

20 kN

15 kN

3 kN/m2

Distributed load





From a qualitative to quantitative perspective, the deflections ofthe ‘‘representative point’’ (Section 3.4) predicted by the model ofspecimen 2 (that with terrazzo) for a distributed load of 3.00 kN/m2 and a concentrated load of 5.00, 10.00, and 20.00 kN (stage 2)are 4.31, 5.10, and 6.69 mm, respectively. Considering that the cor-responding experimental values were 3.81, 4.54, and 6.02 mm,respectively, the differences between the analytical predictionsand the measured deflections are of about 10%. This percentageconfirms that the model guarantees a reasonable level of accuracy.It is to note that the deflections predicted by the model ofspecimen 1 for the same loads were 12.41, 14.21, and 21.44 mm,

respectively, against experimental deflections of 13.51, 16.23,and 22.84 mm; thus also these differences were of about 10%. Itcan be concluded that the deviations between predictions andactual deflections are not due to how the terrazzo was modeled,but to some local effects of timber.

The accuracy of the model of the damaged state is lower thanthat of the undamaged state, although it is sufficient to providereliable estimations. For specimen 2 (Figs. 28 and 29), in fact, thedifferences between the model predictions and the test resultswere within 30%. For specimen 1, these differences were within40% (Figs. 25b and 26b); i.e., the model of the floor without


verti

cal d

ispl

acem

ent (

mm

) SPECIMEN 2

0

-1

-2

-3

-4

-5

-6

-7

(a)

-8

-9B1 B8B2 B3 B4 B5 B6 B7

timber beam

verti

cal d

ispl

acem

ent (

mm

)

0

-1

-2

-3

-4

-5

-6

-7

(b)

-8

-9

13 kN12 kN

10 kN11 kN


9 kN8 kN


10 kN

10 kN

5 kN

15 kN

5 kN

15 kN

20 kN

20 kN

3 kN/m2

Distributed load





verti

cal d

ispl

acem

ent (

mm

)

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7

(a)

-10

-8

-9

-11

-12

-13


verti

cal d

ispl

acem

ent (

mm

)

SPECIMEN 20

-1

-2

-3

-4

-5

-6

-7

(b)

-10

-8

-9

-11

-12

-13

13 kN12 kN

10 kN11 kN


9 kN8 kN


5 kN

10 kN

15 kN

20 kN

5 kN

10 kN

15 kN

20 kN

3 kN/m2

Distributed load



-14 -14




terrazzo was less accurate. The reason is that damage on the floorwithout terrazzo causes greater effects than to the floor withterrazzo.

The actual deflections of the floor in the damaged state werealways lower than the predicted deflections. Thus, the real effectsof damage to the timber beams are lower than what assumed inthe model. Consequently, the estimations of the model are system-atically conservative; hence, the proposed model can be used toassess safety of damaged timber floors.

6. Conclusions

The mechanical behavior of the timber–terrazzo compositefloor has been investigated. The first part of this paper has pre-sented experimental results from tests performed at different loadstages and with different loadings, on two real scale structures dif-fering only for the presence of the Venetian floor. These resultshave pointed out the differences in structural behavior betweentimber floors finished with terrazzo and without it. Some of thetests also induced damage to timber artificially; the results of thesetests have provided information about the structural effects ofdamage on this type of floor. In the second part of the paper, theexperimental data have been used to develop a predictive theoret-ical model, whose mechanical assumptions have been deriveddirectly from the results of the tests. Finally, the paper hasdescribed peculiarities of the traditional structural system.

The results of the tests have shown that the terrazzo slipsalmost freely with respect to the board that it rests on. Conse-quently, no significant shear action is transmitted at the commonboundary between the timber board and the terrazzo. Hence, nocomposite action exists in this type of floor.

Nevertheless, the results of the tests have also shown that theterrazzo shares a significant fraction of the load and that this con-tribution is provided for at least the entire elastic range of the tim-ber system.

Hence, the timber–terrazzo floor, on one hand, is neither a lam-inated nor a sandwich. In fact, no in-plane actions are transferredbetween the components. Therefore, the floor’s stiffness andload-carrying capacity are equal to the sum of the stiffness andload-carrying capacity of the components, and not greater, as itwould occur in an actual composite structure.

On the other hand, however, the terrazzo has been proven to bea load-bearing component. Therefore, this floor type consists of astructure whose components are both the timber system and theterrazzo slab, although no composite action is transmittedbetween the timber system and the terrazzo slab. As a result, theterrazzo create a load-sharing mechanism between the compo-nents, which makes the terrazzo share the loads applied to thefloor in proportion to its bending stiffness. Moreover, the terrazzoalso creates a load-sharing mechanism between the timber beams,which makes the terrazzo distribute part of the load to all the lon-gitudinal beams.

The reason why the terrazzo slab is a load-bearing componentof this floor is that the terrazzo does not crack appreciably duringthe elastic displacements of the timber system. In particular, evenfor the floor deflection that makes the timber attain (and surpass)its elastic limit, the terrazzo usually exhibits only marginal cracks.Hence, cracking does not reduce considerably the bending stiffnessof the terrazzo.

Since the thicknesses of the typical terrazzo floorings are great,the load fractions usually shared by the terrazzo floorings are sig-nificant. In particular, the load-sharing mechanisms provide thetimber floor of historical buildings finished with terrazzo withmuch greater stiffness and strength than the same floor withoutterrazzo. These mechanisms explain why many historical floorshave functioned well during the centuries although the timber

beams are small. In particular, these mechanisms prove that thepresence of the terrazzo drastically reduces the deflections of thefloor.

The experimental information from the full-scale structures hasallowed the behavior of the timber–terrazzo floor to be describedcomprehensively by few statements, which have been adopted asmechanical assumptions. Based on these assumptions, an analyti-cal model has been developed, which predicts the linear behaviorof timber floors with terrazzo as well as which provides mechani-cal interpretations of the behavior.

The good agreement between the results from the model andthe results from the tests has proven that the mechanical assump-tions reproduce closely and exhaustively the actual behavior of thistype of floor. In particular, the main assumptions whose validityhas been confirmed by the comparisons between theoretical andexperimental results are that the terrazzo and the timber are notcapable of transmitting to one another any in-plane shear actionthrough the common boundary, that the terrazzo provides thefloor with its entire bending stiffness, that in the transverse direc-tion the floor behaves as an elastic beam resting on elastic supportshaving a certain width, and that the twisting action in the terrazzoslab is negligible.

Another conclusion of this research is that damage to timberbeams does not reduce the stiffness of a floor. In fact, damageaffects mainly the end of the beams, since it is due to moistureand contact to outer environment. From the structural perspective,damage to a timber beam reduces the end cross-section, which, inturn, reduces the rotational stiffness of the connection, since theend is not fixed into the wall any longer. However, the reductionof the end cross-section has marginal effects on the stiffness, sincethe connection behaves as a hinge even when the beam is not dam-aged, and it has no effects on the strength, unless it is substantial.The main major effect of the reduction of the end cross-section isthat the contact stresses between a beam and the supporting wallshift from the inner edge to the outer edge of the wall cross-sec-tion. This behavior causes the span of the beam to increase. How-ever, the increase in deflection and the decrease in strength due tothis effect are not substantial.

That result shows that damage has opposite effects on thesafety of timber beams. On one hand, damage does not produceany significant extra deflection, since it usually affects the ends.On the other hand, the increase of deflection does not point outthe reductions of the shear capacity of the beam. As a result, if abeam is on the verge of collapsing since damage has reduced theshear-load-resisting capacity of the beam’s ends, its deflection doesnot exhibit this dangerous condition. However, the terrazzo miti-gates this risk, since it can share the load’s fraction not shared bythe beams.

Ultimately, the terrazzo slab that finishes the timber floors ofmany historical buildings has to be considered as a load-bearingelement of the construction, in particular of the floor.

Acknowledgments

The tests were sponsored by CO.RI.LA (Consortium for coordina-tion of research activities concerning the Venice lagoon system),within the program ‘‘Scientific Research and Safeguarding of Venice’’.The authors wish to acknowledge Prof. Mario Piana and Paolo Fac-cio, of the Università IUAV di Venezia (Italy), who organized andcoordinated the research ‘‘Protection of Venice from the lagoonwater’’, and supported the experiments presented in Section 3.

References

[1] Barbieri G, Biolzi L, Bocciarelli M, Fregonese L, Frigeri A. Assessing the seismicvulnerability of a historical building. Eng Struct 2013;57:523–35.

http://refhub.elsevier.com/S0950-0618(15)00093-8/h0005



[2] Franzoni E. Rising damp removal from historical masonries: a still openchallenge. Constr Build Mater 2014;54(15):123–36.

[3] Foraboschi P, Vanin A. Experimental investigation on bricks from historicalVenetian buildings subjected to moisture and salt crystallization. Eng Fail Anal2014;45:185–203.

[4] Biolzi L. Evaluation of compressive strength of masonry walls by limit analysis.J Struct Eng 1988;114(10):2179–89.

[5] Foraboschi P. Church of San Giuliano di Puglia: Seismic repair and upgrading.Eng Fail Anal 2013;33:281–314.

[6] Borri A, Corradi M, Speranzini E. Reinforcement of wood with natural fibers.Compos B Eng 2013;53:1–8.

[7] Brites RD, Neves LC, Saporiti Machado J, Lourenço PB, Sousa HS. Reliabilityanalysis of a timber truss system subjected to decay. Eng Struct2013;46:184–92.

[8] Calderoni C, De Matteis G, Giubileo C, Mazzolani FM. Flexural and shearbehaviour of ancient wooden beams: experimental and theoretical evaluation.Eng Struct 2006;28(5):729–44.

[9] Coggins CR. Timber preservation in building and construction. Constr BuildMater 1989;3(3):114–7.

[10] D’Ambrisi A, Focacci F, Luciano R. Experimental investigation on flexuralbehavior of timber beams repaired with CFRP plates. Compos Struct2014;108:720–8.

[11] Fragiacomo M, Lukaszewska E. Time-dependent behaviour of timber–concretecomposite floors with prefabricated concrete slabs. Eng Struct2013;52:687–96.

[12] Gattesco N, Macorini L. In-plane stiffening techniques with nail plates or CFRPstrips for timber floors in historical masonry buildings. Constr Build Mater2014;58:64–76.

[13] Mascia NT, Simoni RA. Analysis of failure criteria applied to wood. Eng FailAnal 2013;35:703–12.

[14] Moses DM, Prion HGL. Stress and failure analysis of wood composites: a newmodel. Compos B Eng 2004;35(3):251–61.

[15] Pajchrowski G, Noskowiak A, Lewandowska A, Strykowski W. Wood as abuilding material in the light of environmental assessment of full life cycle offour buildings. Constr Build Mater 2014;52:428–36.

[16] Premrov M, Dobrila P. Experimental analysis of timber-concrete compositebeam strengthened with carbon fibres. Constr Build Mater 2012;37:499–506.

[17] Vololonirina O, Coutand M, Perrin B. Characterization of hygrothermalproperties of wood-based products – Impact of moisture content andtemperature. Constr Build Mater 2014;63:223–33.

[18] Wang W, Eamon CD. Load path uncertainty in a wood structure and the effecton structural reliability. Eng Struct 2013;56:889–96.

[19] Foraboschi P. Resisting system and failure modes of masonry domes. Eng FailAnal 2014;44:315–37.

[20] Betti M, Orlando M, Vignoli A. Static behaviour of an Italian Medieval Castle:Damage assessment by numerical modelling. Comput Struct 2011;89(21–22):1956–70.

[21] Zabulionis D, Kizinievic O, Feo L. An analysis of the stress–strain state of atimber-concrete T cross section. Compos B Eng 2013;45(1):148–58.

[22] Dias AMPG, Jorge LFC. The effect of ductile connectors on the behaviour oftimber-concrete composite beams. Eng Struct 2011;33(11):3033–42.

[23] Gutkowski R, Brown K, Shigidi A, Natterer J. Laboratory tests of compositewood-concrete beams. Constr Build Mater 2008;22(6):1059–66.

[24] Mungwa MS, Jullien JF, Foudjet A, Hentges G. Experimental study of acomposite wood-concrete beam with the INSA–Hilti new flexible shearconnector. Constr Build Mater 1999;13(7):371–82.

[25] O’Loinsigh C, Oudjene M, Ait-Aider M, Fanning P, Pizzi A, Shotton E, et al.Experimental study of timber-to-timber composite beam using welded-through wood dowels. Constr Build Mater 2012;36:245–50.

[26] Yeoh D, Fragiacomo M, Carradine D. Fatigue behaviour of timber–concretecomposite connections and floor beams. Eng Struct 2013;56:2240–8.

[27] Alam P, Ansell MP, Smedley D. Mechanical repair of timber beams fractured inflexure using bonded-in reinforcements. Compos B Eng 2009;40(2):95–106.

[28] de Jesus AMP, Pinto JMT, Morais JJL. Analysis of solid wood beamsstrengthened with CFRP laminates of distinct lengths. Constr Build Mater2012;35:817–28.

[29] de la Rosa García P, Cobo Escamilla A, Nieves González García M. Bendingreinforcement of timber beams with composite carbon fiber and basalt fibermaterials. Compos B Eng 2013;55:528–36.

[30] Dourado N, Pereira FAM, de Moura MFSF, Morais JJL. Repairing wood beamsunder bending using carbon–epoxy composites. Eng Struct 2012;34:342–50.

[31] Khelifa M, Celzard A. Numerical analysis of flexural strengthening of timberbeams reinforced with CFRP strips. Compos Struct 2014;111:393–400.

[32] Jasienko J, Nowak TP. Solid timber beams strengthened with steel plates –Experimental studies. Constr Build Mater 2014;63:81–8.

[33] Li YF, Tsai MJ, Wei TF, Wang WC. A study on wood beams strengthened by FRPcomposite materials. Constr Build Mater 2014;62:118–25.

[34] Li YF, Xie YM, Tsai MJ. Enhancement of the flexural performance of retrofittedwood beams using CFRP composite sheets. Constr Build Mater2009;23(1):411–22.

[35] Yang YL, Liu JW, Xiong GJ. Flexural behavior of wood beams strengthened withHFRP. Constr Build Mater 2013;43:118–24.

[36] Corradi M, Speranzini E, Borri A, Vignoli A. In-plane shear reinforcement ofwood beam floors with FRP. Compos B Eng 2006;37(4–5):310–9.

[37] Foraboschi P, Vanin A. Non-linear static analysis of masonry buildings basedon a strut-and-tie modeling. Soil Dyn Earthq Eng 2013;55:44–58.

[38] Oudjene M, Meghlat EM, Ait-Aider H, Batoz JL. Non-linear finite elementmodelling of the structural behaviour of screwed timber-to-concretecomposite connections. Compos Struct 2013;102:20–8.

[39] Alves RJ, Magalhaes MDC, Carrasco EVM. Determination of the transverseYoung’s modulus (TYM) of wood by means of an input power technique.Constr Build Mater 2013;42:11–21.

[40] Benallal A, Bigoni D. Effects of temperature and thermo-mechanical couplingson material instabilities and strain localization of inelastic materials. J MechPhys Solids 2004;52(3):725–53.

[41] Biolzi L. Mixed mode fracture in concrete beams. Eng Fract Mech 1990;35(1–3):187–93.

[42] Biolzi L, Labuz JF, Muciaccia G. A problem of scaling in fracture of damagedrock. Int J Rock Mech Min Sci 2011;48(3):451–7.

[43] Lin Q, Yuan H, Biolzi L, Labuz JF. Opening and mixed mode fracture processesin a quasi-brittle material via digital imaging. Eng Fract Mech 2014;131:176–93.

[44] Piccolroaz A, Bigoni D. Yield criteria for quasibrittle and frictional materials: Ageneralization to surfaces with corners. Int J Solids Struct 2009;46(20):3587–96.

[45] Biolzi L, Labuz JF. Global instability and bifurcation in beams composed ofrock-like materials. Int J Solids Struct 1993;30(3):359–70.

[46] Gonzalez-Torre I, Calzada-Perez MA, Vega-Zamanillo A, Castro-Fresno D.Evaluation of reflective cracking in pavements using a new procedure thatcombine loads with different frequencies. Constr Build Mater2015;75(30):368–74.

[47] Grishin VI, Glebova MA, Bespalov VA, Gotselyuk TB. Investigation of fracturecriteria of composite samples with stress concentrators in compression. CompMech Comput Appl 2014;5(2):111–45.

[48] Qiu D. Industrial floor design, construction and testing. Struct Eng2008;86(8):29–37.

[49] Foraboschi P. Three-layered plate: elasticity solution. Compos B Eng2014;60:764–76.

[50] Bartoli G, Betti M, Facchini L, Orlando M. Non-destructive characterization ofstone columns by dynamic test: application to the lower colonnade of theDome of the Siena Cathedral. Eng Struct 2012;45:519–35.

[51] Foraboschi P. Coupling effect between masonry spandrels and piers. MaterStruct/Mater Constr 2009;42(3):279–300.

[52] Betti M, Bartoli G, Orlando M. Evaluation study on structural fault of aRenaissance Italian palace. Eng Struct 2010;32(7):1801–13.

[53] Foraboschi P. Layered plate with discontinuous connection: exactmathematical model. Compos B Eng 2013;47:365–78.

































































































































mechanical behavior of the timberâ€“terrazzo composite floor

Documents