You are on page 1of 12
Engineering Seturs 216 (2020) 110877 journal homepage: www-e!: Contents sts available at ScienceDirect Engineering Structures vier.com/locate/engstruct, System reliability-based limit state design of support scaffolding systems Cao Wang, Hao Zhang, Kim J.R, Rasmussen, James Reynolds, Shen Yan" Schou of Chil Enger The Unive of Syn, Srey, MSW 2006, Auta owed ‘The safety of seafold systems daring construction ir vital to prevent files with catastrophic consequences. Lungloeerng practice inthe seffelding industry, however, does nt have a atonal structural tellablity sis as that ued for buildnge and bridges. The implementation of design clauses for seaffeld sytem in the current state design approach s developed in his study, making fl use ofthe advanced finite element (FE) method ané ‘he secenly accumulated satstlal data of seaflding resistance and constuction loads. A stachaste finite clement method wed lo oblain the probsbilstie characteris ofthe ulimateloud-arying eapacies of sutiess properties. The latest construction loa survey data is ulizeé System reliably asessment is then Desformed to develop new design eitera which are consistent with generally accepted structural ibility targets 1. Introduction ‘The failure of temporary structures in construction such as scaffolds may lead to estastrophie consequences to workers and the general ‘publi, While the design of permanent structures such as buildings and bridges har been codified, limited attention hae been paid to the safety check of seaffold systems. It is offen the ease in practice that fast ‘erection and minimized costae the drivin factors for scaffold systems, ‘with insufficient attention paid to the overall system safety (1). Nu- merous severe accidents in the past due to scaffolding failure have clearly demonstrated the importance of establishing a reliable design practice for saffo systems fo achieve an acceptable safety level (2. Formwork-supportng scaffolding structures are wed in concrete ‘construction as temporary shoring systems to support the weight of formwork, poured concrete, workmen and equipments during the ‘construction. A typical scaffold system has the components of uprights (vertical member), horizontal ledgers, adjustable jacks, diagonal braces ‘and steel U-heads supporting horizontal formwork. Uprights, ledgers land braces are typically circular hollow sections. Uprights and ledgers ate connected by seafold joints, e.g, cuplock joints and wedgetype joints. The heights of stel jacks atthe top and bottom are adjustable "ig. 1 demonstrates atypical L-bay scaffold frame and its 3-dimensionel finite element model “The failure risk of a saffold system may arse from the uncertainties “Coresponding ator. sssociated with the structural resistances and loads, the geometric im- perfections due to multiple reuse, as well as human errors. I¢ was re ported that forthe period between 1961 and 1982, about 74% scaf- folding collapse occured during concrete pouring, mainly due overloading (2) ‘Traditionally, the design of temporary structures follows the al lowable sress design (ASD) procedure, with a design criterion given by Be EE > design tnd, ow in which Ry isthe nominal strength ofa scaffold system computed using ‘the nominal values of material properties and geometric properties, and SP isthe safety factor, typieally around 2.0 in Austeaia, In practice, [Ry/SP is often termed as “working load limit". The design load in ASD is ‘typically a combination of unfactored relevant loads, For example, in the case of gravity loads only, the design load is Dy + La ia Which Dy and Ly represent the dead load and live loa, respectively. It can be seen ‘thatthe ASD approach uses a single, overall safety factor (SF in Eg. ()) te control the risk arising from the uncertainties from both the re sistance and loads For stel building and bridge structures, current design specifica Lions and standards adopt the so-called lime state design (LSD) approach. In the North America and Australis, the LSD takes the format of mail adaresies:c20 wang’sydney cd au (C. Wang) hao zhangtésyéneyc@u.au CH, Zhang), him rasmussenisydney du.au ORR Rasmussen), James. teysoldsetsydsey edu. au G. Reynolds), shen yanlsydney eéu.au(S Ya). Inups//dol.o1g/10.1016/engstrve. 2020.110577 Received 25 September 2018; Received in revised form 15 Apel 2020; Acepted 16 Apel 2020 formwork adjustable jack == (100 to 600 mm) 1 adjustable jack (100 te 600 mm) egerig Src: 216 2020) 10677 Fig. 1 Atypleal bay selfld frame and 3D FE mocel i rien Sores ti ccna i rs 20> 3 10 he Be @ in which fy represent the reritanc ofthe system (le frame, ie the pen restance actor andy ate the nominal rca od. ‘an bescn that the DDM represents aye bed design nthe sense thatthe over este behaviour Calta strength) te deety nd plicy checked gains the scr oad. The key sipec ofthe DDM is thatthe lines strength of stratum eytn can be de termined sng vanced nonlinear fe clement anays rovding Scurate predictions sone would achlee in physica ts ofthe ‘uc. Ths bar been ade posible hy the rap development of Computing power and sophisticated montinenr tata! sna soe ware verte a hoe decades, numerous research prograte have Alononsted that ooninenr FE methods an sina the beh Sd cape al eevan Imi water ote scalding saces,n- Corpering tater and geome necinare, il pte Inperfctiony and semtigh jt stiffness [7-12] he design procedie ofthe DDM hasbeen gradaly permite in the codified design of steel structures, Including the Australian steel standard AS4100 (13), the American steel specification AISC 360-16 (14) (ermed as design by advanced analysis), the latest revision of ‘Australian colformed stel standard AS4600 (15), and the Australian ‘tel rack specification AS4084 (15). Theres also ongoing effort forthe urocode to further improve the DDM, which is termed as design by geometrically and material nonlinear analysis with imperfections (GMNIA) analysis. shall be noted that at this tage, the DDM is based fon static analyses, Le, push-down analysis for gravity loads and push- ‘over analysis for wind loads (17. The DDM using nonlinear time-his. tory analysis for dynamic loadings will be future study. ‘The resistance and structural loads of stel scafolding stvctuces Ihave different probabilistic characteristic than those of ordinary Duilding-type steel structures. For the DDM to be adopted in the cod fied design of stel scaffolds, in addition to the availabilty of advanced ronlinesr FE analysis techniques, the remaining challenge isto develop a LSD eriterion suitable for scaffolding structures to achieve an accep: table level of structural safety, ic, the appropriate values of system resistance factor and load combination rules, the objective of this paper isto analyze the system reliabilities of steel scaffolding systems in current practice, and to recommend the optimal system resistance factor and load combination rules for the DM of typical steel scaffolding systems. The present paper considers ‘only the gravity construction loads during conerece placement. Studies Ihave shown that most formwork-tupporting scaffolding failures oc ‘cured during this stage due to construction loads (3). While formwork supporting seaffold structures ean be subjected to wind and seismic Toading, the design seldom would be governed by wind or seismic lads ‘The remainder ofthis paper is organized as follows. Section 2 presents an overview of the firstorder reliability method (FORM). The prob- abilistic resistance and oad models for typical scaffold systems are ‘resented in Section 3. The development of new design criteria and the ‘comparison with the existing design criteria (in tems of strucaral re Liability) aze presented in Section 4, Concluding remarks ae formulated in Section 5, zs = egerig Src: 216 2020) 10677 De pp procing E E 3 2 | | a: ~ Elevation ~ ‘Plan eh hgh yh yh ye E Elevation (c) Plan MP PrP er Pe PP Pe i é a Elevation Fig. 2 Seafold sytem configurations, (4) 1 1 system, (9) 8% 3 yee, (2) 3 Gystem, and (d) 939 system. 2, Structural reliability analysis In the presence of resistance R and m load effects Qi, Qs, Qn, the limit sate function associated with the LSD eriterion of Eq. (2) i given by 3 re R= Qn Qy = Cy ~ Qe, = OO ? o where C= R/Ry, representing the actualto-nominal ratio of the re sistance, Structiral failure is deemed to occur if < 0, The limit state function can be reformulated at o-f- QQ am en Dar Que Oar Qa Ons OR Xe © in which each % ((= 1, 2,..m) isthe actual-to-nominal ratio of load ‘effect Q,. The probability of failure, denoted by f, is the probability that dhe limit state function i less than zero, ie, R= PKG <0) © Jn which Pr() denotes the probability ofthe event inthe bracket. Since probabilities of structral failure are typically very small numbers, the reliability index is often used as an alternative measurement of struc. tural safety in practice. The reliability index, 8, is related to 2) by p=ea-m) ” in which ¢-1() is the inverse function of the cumulative distribution function of the standard normal distribution, Ifthe probabilistic in formation (distribution type and parameters) of the random variables , Ni Xi, Bq. (5) known, several methods, either analytical or simulation-based, are available in the literature (o compute By oF 8 18,19). OF these methods, the fice-order reliability method (FORM) is adopted inthis papet, The FORM is a practical tool for evaluating the sk and reliabilities of civil facilities (18 20-22). Itean make fll use of the probabilistic information ofeach random input to yield a good er LUmate of structural faluze probability. Moreover, the FORM has been used inthe development of many desig criteria in structural standards and thus the establishment an calibration of new design criteria using FORM would be consistent with existing practice. The implementation of FORM is to, in 2 normalized probability space, find the shortest distance from the origin to the normalized limit state function. The reliabity index B is then equal to this shortest distance. The point on the limit state function which yields the shortest distance is called the checking point, denoted by (R, Qf, 3). Details on FORM ean be found in structural reliability textbooks (e-., (18). ‘Asan inverse problem, when developing a new design criterion, one needs to determine & set of resistance and load factors 4, YoY fy 8088 tw achieve a (predefined) target reliability index fj. AN iteration based approach can be used. Starting with two ‘itil estimates = (Yoo) and ZA (RE, Qi, Q5). she teracon-dased procedure is a fellows, (2) Set the intial realization of€ eg. € = (2,1, 1D) (@) For the jth iteration (= 1,2.) find the checking point using © (6) The procedure converges if such ai found that!240) — Where ¢ is a predefined threshold (eg, 0.001). In this case, @ cone vergee value of e® has been obtained. BHD} €, ‘Theoretically, the required resistance and load factors may vary with diferent load conditions, depending on the load ratios 2 in Eq (©). Ths, 8 imposible to achieve a constant reliability inde for all design situations using single set of (ys 7) Im practice, the interes ist find a eotmmon se of (9-7) Which can achieve a relatively uniform reliability sross the spectrum of design scenarios. Such sof patil factors canbe determined by using an optimisation proces [5] which minimis the diference between the resulting and the target elailty indices, c, minimizing ¥, (8, ~ ys)” over the ‘entice design parameter range a which and wate the acbeved re lability index and che weight ofthe th design parameter, respectively. In the cate of gravity loads only, one would choose the posible values of livetodead lod ratio= as the design parameter 5, Resistance and load models for scaffold systems 3.1. Configuration of representative saffld systems Four representative scaffold systems are chosen as the baseline structures to study the probabilistic characteristics of scaffolding te- sistances. Fig. 2 shows the configurations ofthe four systems, which are ‘assigned with diferent numbers of bays, Le, 1% 1,3%3,3%6, and 9x9, The width of each bay Is 1829 mm, All systems consis of three lifts (levels). For each baseline frame, three cases of lift height are ‘considered, ©, 1.0 m, 1.5 m and 2.0 m, The joints connesting the uprights and the ledgers are cuplock joints. This type of joint has a bottom cup and a movable upper cup. The blade ends ofthe ledgers are positioned into the bottom cup, then the top cup is moved down and rotated until a tight connection is achieved. Fig. 3 schematically de- monstrates a cuplock joint. The length of adjustable jack extension can take three representative values, {100 mm, 300 mm or 600 mm, The top and bottom jack extension lengths are set to be equal. As such, the four baseline frames are expanded into total of 4 x 3 x 3 = 36 cases. Spigot joints (column splices) are located atthe second and thie lits of all ysieme. This ie consistent with current construction practices where Spigot joints are not located atthe base lif. The bracing configuration _- Locking pia = Top cup Ledger blade J <— Bottom ou P Ledger |< — Upright Fig. 5. Schematic ofa exploc joint egerig Src: 216 2020) 10677 for each model can be seen in fig. 2. Details ofthe component prop: erties are summarized in Table 1 Advanced finite element models of steel seafolding systems shall take into account important nonlinear effects including geometric non linearity, material yielding, initial geometric imperfections, loading eccentricity, spigot joint modelling, and semi-rigid joint stiffness. The ‘modelling of these effects has been studied previously (8,12,25-25], In the present study, the commercial FE software Strand? (26) is used to construct three-dimensional geometric and material nonlinear FE models for the selected scaffolding frames. The material nonlinear analysis in Strand7 is a plastic-zone analysis using fiber-type beam flements, It is assumed that the steel materials elasic-perfecly plastic. Translation in the horizontal directions is restraint atthe top of the uprights, since in construction practic, the horizontal formwork is ‘ypically restrained by the completed permanent structure, Timber bearer beams generally impose constraints on the rotation of the U. heads. To model this effect, a rotational spring is applied atthe top of| cach upright with a rotational sifness of 29 KN m/ad, as suggested in 27]. In terms of the bottom boundary condition, a base plate eccen tricity of 15 mm is assumed, which is according to the tolerance spe cified in the Australian Formwork Standard [28]. Two types of initial geometric imperfections are considered, ie, the outof-straightness of the upright (member crookedness) and the out-of plumb of each storey Gintial frame sway). The effects of geometric imperfections are in corporated inthe FF models by explicitly moving the relevant FE nodes to their imperfect positions. The cuplock joints connecting uprights and horizontal ledgers are modelled as semi-rigid witha tr-linesr momer rotation relationship demonstrated in Fig. 5 [24]. The vertical loads from the formwork are modelled as point loads applied to each U-bead component, as illustrated in Fig. 2. Load eccentricity is explicitly modelled inthe FE analyses due to its adverse effect on the strength of| scaffolding systems. In the nonlinear analyses, the loads were increased incrementally until system collapse. Fig. | illustrates the FE model of the scaffolding frame shown in Fig, 2(€). More details on the nonlinear FFE modelling can be found in (25). In terms of computational cost, using the commercial FE software Stand7 on an Iatel Xeon CPU @ 3.6 Gtlz computer, the computational time of performing aa advanced nonlinear analyse (push down to frame collapse) of the 3% 3 bay frame (with 300 mm jack extension, 1.5-m lift height) is about one minute, ‘The FE models were verified against 15 full-scale scaffolding load, tests (25). Since the DDM is based on the ultimate frame stength, the accuracy of the advanced analysis for predicting the ultimate system strength i of most interest. The ultimate loac-carying capacities from the 15 load tests and predicted by the advanced analysis were com pared, and a certain degree of discrepancy was observed. The dis crepancy is due to the inevitable modelling simplifications and ideal sations such as the idealized trilinear momentotation response of| cuplock joints, and idealized boundary conditions in the FE analysis. ‘This modelling uncertainty of the DDM needs to be accounted for in developing the probabllity-based design criterion, The tue strength of a frame, denoted by Rc, can be expressed as Rag = Rene, © ln which Rpgy represents the predietion fom the advanced analysis, and € is the associated modelling uncertainty. The modelling un. certainty € i treated asa random variable, representing the rato of test to-numerical prediction. Fig. 4 plots the histogram of the modelling uncertainty © based on the 1S fullseale saffoding load tests. The 15 sample data of € have a mean of 1.02 and a coefficient of variation (COV) of 0.063. Ie is assumed that the modelling uncertainty fllows 2 normal distribution 3.2, Uncertain parameters of sel seafeling systems Steel scaffold systems are erected and dismantled many times during the lifetime. Consequently, the components of scaffold systems [Component properis of the safoing systems egerig Src: 216 2020) 10677 7 ‘andards ‘Nominal ede ameter of 43 am and thoes of 4 am old lomsed celeste abe (CS) grade 450 MPa mean = 1.02 eee Cov =0.063 Testto-FE prediction Fig 4. Histogram of the modelling uncertainty ofthe advanced analy. é ky 0 a ‘a Rotation Fig 5, Wilinear model of momentrotaton responce of cuplck joint. 'ypically pose imperfections and uncertainties, from poor handling and reuse of from the manufacturing of the component. The parameters ‘with variabilities are modelled a random variables in order to estimate the probabilistic characterstes of system strength. The random vari- ables considered in the present study include: «+ Initial geometric imperfections: out-ofstraightness of uprights, and frame out-of plumbness; # Standard-to-Ledger joins: joint stiffvess of the euplock connection; 1 Seandards: cross-sectional area, moment of inertia and yield stress ‘Jacks: cross-sectional area, moment of inertia and yield stress; * Load eccentricity ‘Model uncertainty of the advanced analysis; * Consteuetion loads (dead load and live load) ‘Among the six parameters (ky, a5) defining the tlinear moment-otation response of euplock joints, only ky fe and ky are modeled as random variables, while 2 and 25 have relatively small variabilities and are treated as deterministic, Moreover, the moment rotation response of the cuplock joints depends on the number of ed gers connected at the joint. Toble 2 summarizes the statistics of kk, ks for three euplock joint configuration, ie, 2-way (comer join), -way ‘Table Statistics of the joint stifness (yy A) Gin kNnad) and the rotation (G0 3) Gn re) for euplck joins (edge) and 4-way (interior). The nominal values of the stiffness para meters are equivalent to their mean values |24). The parameters 3.3: snd a) ate also given in Table 2. The statistical daca ofthe remaining random variables are preseated in able 3. The statistics ofthe moment of inertia and cross-sectional area of uprights and jacks were de- termined based on the measurement data of 80 used uprights and jacks 28]. The data of geometric imperfections and loading eccentricity was collected from a number of construction sites (20 For the type of steel scaffolding structures considered in this study, the typical flare mode is the buckling of the vertical members (upright or jack), while the failure of ledgers and brace members is generally preciuded (27). Thus the properties of uprights and jacks are modeled as random variables ‘The properties of ledgers and braces are treated as deterministic. Sen sitivity analysis has shown that the system steengths aze insensitive £0 the randomness in the properties of ledgers and braces (27). Note that the statistics of steel yield stress given in Table 3 is based on tests of| steels manufactured in the 1970's, Because of improved manufacturing processes, the yield stress of modern grades of stel has a somewhat smaller mesn-to-nominal resi and a smaller COV [31], However, it was found that using the updated statistical data for stel leads to only slightly higher relibilicy indices (31). The present paper stil adopts the coviinalstatistis of yield stress, which was used in the reliability ex libration of the first-generation probabilty-basd stel design codes (5) 3.3, Probabiltc models for construction loads ‘An accurate estimate of the load statistical information for scaffold systems is important in developing rliailty-based design criteria. The variability associated with the construction loads is typically large arising from different construction activities. However, limited data on the construction loads is accessible, implying thatthe understanding of load-esistance interaction is insufficient inthe Itrature and in prec: tice [54,35]. For scafold systems with lateral restraints at the op, a is fen the esse In practice, the dominant loads would be the dead load and live load. The dead load includes the vertical load exerted by formwork as well as the sef-weight of construction materials (eg concrete, reinforcement and others) The live load includes the weight of workers and equipments supported by the scaffold and the dynamic loading effects during concrete pouring, Fig. 6 illustrates the typical shore load history during the construction process containing three stages, Le, before, during and after concrete placement. Most support scaffolding failures occurred during the second stage (5), making the phase of concrete placement the most critical “The statistics of dead and live loads for scaffold systems have been lavestigated by fleld load surveys [25,34,35-40], which are ‘Staites ofthe random properties of seafold systems egerig Src: 216 2020) 10677 Oxtetargn f pe 0 ao o7s ‘eso wa cea of apo ‘300 tem oats ‘stoma fa lending ect a) i io Sos sepa bat Nome fet fc na) seas ee 0s rpm {Sem mean asf wp or sor oes lgsoal Baal Nomen of ort of echt) 676 a7 oes ‘ote baal ald efx epee OP) ‘so ‘98 on ona i Yk are rer) Ss aa on tenor 1 Tis te length of wp ° tember with pg on able 3 concrete Statistics of construction dead load and construction lve load Placement toad Moses cov a 2 ; sompleted ' Dead nd 1s @ eral conere | Ueto ass a Berene Te Pieement | ~ Based on nomial ive ond of 1.0 KPa sare | 1 time Before placement uring plement After placement Fig. 6, Shore loads in diferent construction stages ‘representative of the construction practices in different regions such as ‘Australia, Europe and the USA. The load survey conducted by (25) was based on a two-year shore load investigation undertaken in Sydney, ‘Australia, with 188 shore load measurements recorded in three con- struction sites. It was found that during concrete placement, the dead Toad effect isthe dominant component of the shore loads, and the live load effect is very small compared with the dead load effec. The dead load has a mean to nominal ratio close to unity, with a COV about 0.3. The (equivalent uniform) live load on shores has a mean of 0.81 KPa (yielding a mean-to-nominal value of 0.81 with a 1.0 kPa design live load) with a COV of about 0.46. The dead load and live load ean be modelled by normal and extreme type I distributions, respectively. In 36], it was found that the dead load has a mean-to-nominal value of 0.82 and COV of 0.35, while the (equivalent uniform) lve load has @ ‘mean-to-nominal of 0.74 (with a 1.0 kPa desiga lve load) and a COV of 0.71. The observations from (37) show thatthe mean-to-nominal and COV of dead load are 0.90 and 0.29 respectively, while the live load has ‘2 mean-to-nominal ratio of 0.99 and a COV of 0.31 based on a 1.0 kPa design live load. The results ofthese studies are not dissimilar; the dead Toad on saffolding systems las a meat-to-nominal ratio of about unity, and the variability ofthe dead load is quite significant, with a COV of approximately 0.30. The mean of the (equivalent uniform) live load on scaffolding systems varies between 0.74 and 0.99 KPa, with a COV between 0.46 and 0.7. Based on these load survey results, thi paper sstumes thatthe constuction dead load is & normal distribution with a ‘mean-te-nominal value of 1.05 and a COV of 0.3, and the constriction live load follows extteme Type I distribution with a mean-to-nominal value of 0.85 and a COV of 0.6, ax summarized in Iable In addition to the probabilistic modes of the dead and lve loads, the structural reliability of saffolding structures also depends on the live: todead load ratio. In the Australian Formwork Standard (28), the equivalent uniform design live Toad is 1.0 Xa forthe stage of concrete placement. Assuming that concrete slab thickness of typical residential and commercial buildings varies between 0.15 to 0.35 m, and the density of reinforced concrete is 2500 kg/m’, then the design live-o dead load ratio L/D, would vary ina range of 0.1 to 0.3. Ite noticed thatthe loading seenario of 8 scaffold system differs from that of an ordinary concrete structure, which typically has a live-todead load ratio of © varying from 0.5 to 1.5 (5 5.4, Staisties of system resistance Statistics ofthe strengths ofthe 36 scaffolding frames were obtained using Monte Carlo simulation with the Latin Hypercube sampling technique. For each frame, 350 Monte Carlo trials were conducted to construct 350 samples of the system resistance (Ri Bq, (4). It should ‘be noted that the purpose of Monte Carlo simulation i not to compute direcly the probability of failure, for which 350 simulations are not sulficient to capture the lower tail of structural resistance, Instead, the Monte Carlo simolation i used to estimate the statistics characteristics (mean, standard deviation, distribution type) of the structural re sistance over its complete range, which can be estimated reasonably well using relatively few simulations (typically 100 to 500) (15) In each Monte Carlo trial, the values ofthe random parameters of ables 2 and 3 are fist randomly sampled according to their distribu tions, and then incorporated in the FE model to compute the system resistance, which is subsequently multiplied by a random modelling uncertainty to obtain a sample of the system resistance. tis assumed thatthe initial frame sway at each storey is perfectly correlated, and thatthe loud eccentricity and out-of-straightness of uprights are ran domly positioned towards either side of each vpright with equal pos sibility. Wit the statistics of member imperfection in Table 3, arandom direction and magnitude for outotstraightness is assigned to each ‘upright independently in the x

You might also like