You are on page 1of 20

Structural safety

ELSEVIER
Structural Safety 17 (1996) 205-224

Crane load modelling 1


Hartmut Pasternak a, Bogdan Rozmarynowski I* , Yi-Kwei Wen
a Technical University of Braunschweig, D-3300 Braunschweig, Germany Technical Uniuersity of Gdahsk, Pl-80952 Gdarisk, Poland Unicersity of Illinois Urbana-Champaign, Urbana, IL 612801, USA

Abstract Based on the theory of stochastic processes, a model for the vertical load on columns of buildings due to one or two overhead travelling cranes was developed. In case of two cranes, it is assumed that they move either on the same crane track (then load modelling includes clustering effects) or in neighbouring bays. Characteristic values, partial safety factors and combination factors have been estimated on the basis of extensive numerical studies, which may be used for future code provisions on crane loads. Keywords: Load models; Load combination; Point-crossing method; Load coincidence method; Clustering effects

1. Introduction

One of the most important considerations in the design of buildings and structures is the determination of loads and load combinations for safety checking. For this purpose, the characteristic value load factor and combination factor are often used in the building code. In the new German DIN 18800/l [3] and EUROCODE 1 [4] they are defined as follows: . Characteristic values S, of action (loads) are based on the appropriate fractile value, i.e. 99%-fractile according to GRUSIBAU guideline [6] and 98%-fractile of the loads referred to one year as stated by EUROCODE 1. The latter deals with loads, but it does not include any specific prescription for crane loads yet. * A partial safety factor yF is used in such a way that the design value of an action load is S, = yFSk. This design value corresponds approximately to a safety index = 3.8. For simplicity, both codes assume that yF = 1.5 for variable actions. * Combination factor Q I 0.9 is used to account for the small chance of simultaneous occurrence of different variable actions (e.g. wind and live load).
* Corresponding author. Discussion is open until November

1996 (please submit your discussion

paper to the Editor,

Ross B. Corotis).

0167-4730/96/$15.00 0 1996 El sevier Science B.V. All rights reserved SSDI 0167-4730(95)00011-9

206

H. Pastemak

et al.

/Structural

Safety

17 (1996)

205-224

0.6.-

Fig. 1. Examples of combination craneway girder spans.

factors of crane loads I) for different

calculation

action zones as a function

of

For overhead travelling cranes, the existing deterministic nominal wheel loads have been introduced as characteristic values. Moreover, the partial safety factor has been assumed a value a priori (i.e. without any probabilistic reasoning). According to Kikin et al. [7] this coefficient is about 1.2 to 1.4. It depends mainly on the service intensity and the unevenness of the runway. Following the mentioned (and many other) codes, forces caused by two cranes are to be treated as one action, consequently, no combination factors have been provided. In contrast to this, the Russian code [ll] contains a combination factor for the case of two cranes. Depending on the load intensity, 4 is either 0.90 or 0.95. Belenja and Vasiljev [l] present the combination factor $ for different calculation zones, especially with the emphasis on the effect of the span of the craneway girder (Fig. 1). In the present paper, based on the theory of stochastic processes, a model for the vertical loads acting on buildings due to one or two overhead travelling cranes is developed. In case of two cranes, it is assumed that they move either on the same crane track (then load modelling includes clustering effects) or in neighbouring bays. Characteristic values, partial safety factors and combination factors are estimated and can be used for future code provisions on crane loads.

2. Model

for one crane

2.1. Problem

formulation

Considering the craneway as a system of simply supported continuous beams, the vertical reaction R, due to a moving crane can be described by a trapezoid pulse shape function as shown in Fig. 2. Such a model of the structural system supporting the craneway was motivated by the necessity of defining the pulse shape in a tolerably simple way. This is due to difficulties arised if the probabilistic analysis of the combined renewal pulse processes is to be performed [13]. The dashed line in Fig. 2 shows the actual influence line due to the superposition of two moving forces, whereas the solid line is the simplification of the actual one to facilitate further

H. Pasternak

et al. /Structural

SafetyI7

(1996)

205-224

207

Fig. 2. Influence

line (- - -) and approximated

pulse shape (-)

for vertical reaction

R,.

analysis. different

The vertical reactions of the crane wheels F,, F2 due to dead weight and loading bays (Fig. 3) may be obtained from

in

(1)

(2)
where Eqs. (1) and (2) are valid for bays L and R respectively; QKB = the weight of the crane bridge; Q = the lifted load; QK = the weight of the trolley; L = the span of the crane and y = the variable describing the trolley position. The extreme value Si of the ith load pulse, can be obtained from the following relationship (Fig. 2):

(3)
where eK7 is the wheel base and 1 denotes the girder span; F;. is given by Eqs. (1) and (2).

Fig. 3. Vertical

reactions of the crane load processes X, and X2 acting on columns.

208

H. Pasternak frequency 1

et al. /Structural

Safety

I7 (1996)

205-224

Fig. 4. Frequency

distributions

of the lifted loads 181.

In the analysis to follow the ith load process will be denoted by X, (Fig. 3). Two random variables X,* and X,* are used to describe the uncertainty of forces given by Eqs. (1) and (2): X,* = Q (lifted load) and X; = y (trolley position). The probability density function (PDF) of each random variable is assumed to be known. The first variable deals with the lifted load and is characterized in the form of empirical frequency diagrams (Fig. 4) [81. On the basis of these diagrams, the suitable PDFs are constructed. These four cases describe different regimens of the crane work, i.e. the diagram A in Fig. 4 shows that the greater part of lifted weights is of smaller quantity, the diagram D depicts the cranes work in the range of larger weights, wheras diagrams denoted by B and C represent intermediate cases. Note that in all cases the maximum value of the lifted load has been chosen 20% higher than the nominal load (overloading and small unevennesses of the runway are included). The second random variable is described by the PDF of the trolley position which is proposed to be of two types: uniform and triangular (Fig. 5). The triangular PDF represents bridge cranes working predominantly on one side of the bay. The triangular PDF of the trolley position may have the peak value on the left (as in Fig. S(bottom)) or on the right-hand side. This allows one to consider the effect of a non-uniform trolley position on the reaction. Assumming statistical independence between X,* and XT one gets

f x,*x;h4

=fx+Jfx;(4

(4)

H. Pastemak

et al. /Structural

Safety

17 (1996)

205-224

Fig. 5. PDFs of the trolley position.

in which f indicates probability density function (PDF). With the aid of Eq. (4) and using standard methods of the probabilistic analysis one obtains the PDFs of the peak value of the load pulse effect as follows: * Right bay f&w . Left bay = ;(&-x{ g$

1 Qm -s-B 4

(5)

(6)
where A = 2Ll; B = 21- e,,; the remaining symbols are shown in Fig. 2 and 3 and 5. The integration in Eqs. (5) and (6) can be performed by standard numerical procedures. For the further analysis of the load combinations one needs to consider two additional functions, i.e. the mean upcrossing rate function and the arbitrary-point-in-time (APIT) distribution function [9]. The upcrossing rate provides information of the individual load effect process from which the distribution of the maximum over a given time can be determined. For a pulse type process which returns to zero at the end of each occurrence, the mean upcrossing rate can be written as (7) iltr> = i.S,() where G,( *) is the complementary cumulative distribution function (CDF) of the peak value of the pulseand vi denotes the mean rate of arrival of ith load. Note that Eq. (7) also applies to triangle pulses [9].

210

H. Pasternak

et al. /Structural

Safety

17 (19961 205-224

When two processes are combined, one needs the APIT distribution function. For renewal pulse processes this APIT distribution is

where r denotes the level of the load effect; fX,,on(.) is the PDF of the value of ith load process given that it is on; 6(e) is the delta-Dirac function; p and q are the probabilities of the process being off and on respectively. For a renewal rectangular pulse process one gets

fx,,ok) =fs,(r)
where fS,( 0) is described by Eqs. (5) and (6). For other pulse shapes it is necessary to derive fx,+,, (0) taking into consideration the effect of the shape of the pulse and fS,( -1. The trapezoid pulse shape that is under consideration can be seen to be composed of two such parts that the pulse duration is d = 2d; + d;, in which d; and d; denote the pulse durations of the triangular and rectangular parts of the pulse, respectively. After such a decomposition of the pulse, one can write Eq. (8) as follows:

where q; = E(d; )/E(d) and q; = 2ECd; )/E(d); E(.l denotes the expected value; the third term in Eq. (10) describes the ~~,,,,(. > for the triangular pulse [13].
2.2. Numerical results for one crane

f& .>, i.e.

The numerical analysis of load effects for the case of one crane includes three functions: the PDF of the peak value of the pulse process i; f,< .l, i.e. the APIT function and Y;,(. 1,i.e. the mean upcrossing rate for the pulse process X, and appropriate fractiles. The main cranes data (i.e. the nominal values excerpted from the German code [3]) that were used in the numerical calculations are shown in Table 1. The cases to be considered are presented in Table 2. * Table 3 contains the computed 99% and 98%-fractiles ( = characteristic values) for different cases shown in Table 2.

Table 1 Significant Characteristic values

cranes data [3] L=15m 196.2 49.776 15.89 2.5 0.95 L=30m 196.2 156.96 19.23 4.56 0.95

Q 1kNl QKBMl QK Ml
eKT lrnl

X,in 14

H. Pasternak

et al. /Structural

Safety

I7 (1996)

205-224

211

Table 2 Calculated Frequency


A, B, C, D

cases distributions of the lifted load Crane span [m] 15 30 15 30


Cases

PDF of the trolley position Rectangular R-l R-2 R-3 R-4 Triangular T-l T-2 T-3 T-4

Girder 6 6 12 12

span [m]

Table 3 99% and 98% fractiles of the crane load effect PDF of the trolley position

R~nl.,x
[kN] 206.1

R, IL
[kN] 20.5

I
[m] 6

99% fractiles [kN] A 179.6 B


166.1

98% fractiles [kN] C


177.0

[m] 15

D
186.8 191.5 185.7

A
168.3

B 153.8 165.0 158.5


167.5 174.0

C 168.3
176.3 170.2

D
181.0

186.5 179.4 201.6 49.0 6 30 185.1 203.2 233.2 23.2 12 15 211.0 234.5 263.3 64.1 12 30 241.9

175.6 168.4
176.1 188.0

183.6 177.3 183.1 200.3 207.7 231.8 239.2

178.8 169.9 179.1 190.5 202.4 222.0 234.1

186.5 180.9 185.4 204.8 211.0 236.4 242.3

189.4 211.4
216.8

177.0 190.5
199.5

198.3 220.1 230.3

242.7 247.4

186.8 207.2 218.9

222.4 231.4

Fig. 6. PDFs of the pulse intensity. trolley position.

Type A of the frequency

distribution

of the lifted load, rectangular

PDF of the

212

H. Pasternak

et al. /Structural

Safety

17 11996) 205-224

s[kN]
Fig. 7. PDFs of the pulse intensity. trolley position. Type A of the frequency distribution of the lifted load, triangular PDF of the

. According to the lifted load and geometrical relations, the 99%-fractiles give 80% to 92% of the maximum vertical crane forces acting on the column (indicated by R,,,, in Table 3) for the uniform PDF of the trolley position and 85% to 94% for the triangular PDF while 98%-fractiles range between the adequate intervals as follows: 75% to 90% and 80% to 92% respectively. * The particular values, shown in Table 3 are the 99% and 98%-fractiles of the vertical crane forces, which are mainly dependent on the girder and crane spans, range between the following values of the nominal, i.e. deterministic reactions obtained as the summation of the

0,030

----7
\ ---. . . . .

-----L=6m L = 6 m I=12m
L=

7
L=15m L L=
L-

30m 15m
TJom

0.020

CA \\ I -.I I % \ ------- ------- 4 0.010 ; \\r... -., j , \ \. , -4 -I ,,, k, ( f,, , \k 0.000 I,, , , , , ,$
f--f\

,zrn

00 Fig. 8. Arbitrary-point-in-time distribution triangular PDF of the trolley position. functions

to.0 fx$s).

s[k&

80.0 distribution of the lifted load,

Type A of the frequency

H. Pastemak

et al. /Structural

Safety I7 (I 996) 205-224


T--------I=6m - ---... 1=6m I=IZm 1=121x1

213

__

L=ISm L=30m L=ISm L=30m : : '

0.0

I 0.0

II

11

40.0

s[kN 180.0 distribution of the lifted load, rectangular

1 1 I

Fig. 9. Mean upcrossing rate functions Y:(S). PDF of the trolley position. (NCOL = 10).

Type A of the frequency

products of each crane wheel load and influence ordinate under that load (intervals for 98%-fractiles are given in the parentheses): 1. for the uniform PDF of the trolley position 93% to 104% (87% to 102%), 2. for the triangular PDF of the trolley position 98% to 107% (94% to 104%). It should be reminded that in the foregoing analysis the maximum value of the lifted load is assumed to be 20% higher than its nominal value (see Fig. 4). * Fig. 6 and 7 show the PDFs for one chosen frequency distribution (case A in Fig. 4) using rectangular and triangular PDFs of the trolley position, respectively. The PDFs are presented for four cases of the girder and crane spans. It becomes noticeable that PDFs preserve the same shape for all cases. The peaks (also the modal values) of the PDFs have a slight tendency to be shifted to the right. The greatest peak has been always noticed for the second span relation (i.e. I= 6m, L = 30m, where 1 and L are girder and crane spans respectively). This is also true for other frequency distributions. * Fig. 8 shows the APIT functions, whereas Fig. 9 the mean upcrossing rate functions for the case A and analogically for different girder and crane spans. Fig. 8 concerns the triangular distribution of the trolley position whereas Fig. 9 the rectangular one.
3. Model for two cranes

3.1. Introduction When during its lifetime a building is subjected to two (or more) crane loads acting simultaneously, then the problem of their combined action should be considered. Most studies

214

H. Pasternak

et al. /Structural

Safety17

(1996)

205-224

Fig. 10. Pulse shapes formed by two cranes.

on load combination problems have been devoted to the approximate solutions of the probability of the maximum value of the combined load effects for independent loadings [9,10,13]. Recently, modelling and effect of load dependencies have received a great deal of attention [13]. The objective of the following analysis is to describe random vertical forces caused by two cranes that can move on the same crane track or in neighbouring bays. In Fig. 10 the load effect process due to two cranes is presented. The wheel base of the crane is described by eKT, (in the analysis hereafter cKr. is assumed to be identical for both cranes), fmin indicates the minimal distance between two cranes on the same track and c is a constant value relating eKT with time (i.e. c = LI-~, where c denotes a constant cranes velocity), pud stands for the mean pulse duration. For two cranes moving in neighbouring bays of a building: it is assumed that fmin = 0. In addition to the actual trapezoid pulse shape, the rectangular approximation is also indicated in Fig. 10. The problem of the load effect due to two (or more) cranes working separately or together on the same (or not) crane track is reduced hereafter to that of finding the mean upcrossing rate function P;(T) and CDF, i.e. FR (r, T) in which R, is the maximum value of combined load effects in time T whereas R is the combined load effect given by
R(t) = C,X,(t) + C2X2(f)

where Xi and X2 denote the vertical load processes of each working crane, C, and C, are constant load effect coefficients.

H. Pasternak

et al. /Structural

Safety

I7 (1996)

205-224

215

3.2. Point crossing method In the following part the simplified analysis of the load effect combination due to two cranes working independently is considered. The basic formula of the point crossing method describing the mean upcrossing rate function for the combination of two independent loads which return to zero is as follows [9]:
4w = /I -m &*)4l(~ 4 dz* + /= fx,(Z,)~XL(~ -21) dz, (12)

where V;,(Z) = VEGA,,,,, for z I 0, V;,(Z) lmv.F , X,,on(z) for z I 0, f;(. > is expressed by Eq. (81; vi is the mean arrival rate of pulses in the ith load process. It should be noted that Gx,,,J .> and FX,,,,( *> are the distributions of the value of the process given that it is on. For rectangular pulse processes these conditional distributions are equivalent to the distributions of the peak value, i.e. G,,,,( .) = GS,C .> and F,,,,( 0) = F,(. 1. Finally, if S, and S, are always positive and using relationship for the probability on of the ith process, i.e. qi = vdpd,, where pd, is the mean pulse duration, one can write Eq. (12) as follows:

+ u%,+&)

(13)

where vi2 = Y~Y~(P~,+ Pi,) [12] indicates the mean occurrence rate of coincidence for two independent loads; p, stands for the probability off of the ith process. In formula (13) S, + S, denotes the random variable that is the sum of the values of two on pulses. For two cranes such a situation is restricted by the physical reality that two cranes have to keep a minimal distance fmin which is determined by the type of the crane. The minimal shift between two pulses of two cranes is shown in Fig. 11. Depicted situation is the most inconvenient with respect to the superposition of two crane loads. Based on this specification of the problem the mean occurrence rate of coincidence for two cranes in Eq. (13)

---

= c*b
Fig. 11. The minimal

-0~roximote influence Of R

lines

-y)S,

;,,,=

fmin-eKT

shift between two pulses and the relation

between

random peak values.

216

H. Pasternak

et al. /Structural

Safety

I7 (1996)

205-224

can be rewritten as a particular case of the last formula derived in Appendix A, i.e. without a term describing dependencies among loads. It can readily be shown that
VI2 = 12(pd, pdz 2tmin) (14)

where fmin = Cfmin and c = u-r is the constant value aforementioned. It should be noted that for other pulse shapes one can directly utilize Eq. (12). The significance of Eq. (13) is as follows. For a load effect R(t) which is the sum of several load effect processes (see Eq. (11)) there is a random variable R, = max R(t) that is the maximum value of R(t) during the period of time T. The probability of events (R, is larger than r} can be closely approximated for high values of r which are of most engineering interest by
G,_(r, T) I v;(r)T+ GR&r)T = vR+(r)T (15)

in which YR+(Y)is the mean upcrossing rate of R(t) described by Eq. (13) with regard to Eq. (14); G,&r) is th e complementary CDF of the value of R(t) at c = 0. Therefore on the basis of vR+(rl one can also calculate appropriate fractiles of the exceeded r values.
3.3. Occurrence clustering among loads

Usually loads on columns of buildings due to two cranes that move on the same crane track have different arrival times and intensities. Sometimes they may have a much higher chance of coincidence (in extreme cases two cranes have to put up the same component part). In other words such loadings may be clustered around a common point in time. Processes with clustering effects that can be understood as a kind of correlated processes, are discussed by Cox and Lewis [2] and Wen [13]. The occurrence clustering among load processes consists in the following idea (Fig. 12). In Fig. 12 are denoted:
0
A X

- Generating (parent) point process with the occurrence rate p, - Delay point process (may occur with probability Pi>, - Noise process with the occurrence rate pi, - Delay time (random), - Intensity given occurrence, - Duration given occurrence.

T
i

di

Loadings may be clustered around a common point in time where there is a much higher chance of coincidence. In case of crane loads this point may be given by the manufacturing cycle in the building. The point is indicated by 0 and the process is called the parent process. This is a simple Poisson process with an occurrence rate p. The load may occur with a

H. Pasternak

et al. /Structural

Safety

17 (1996)

205-224

217

1 I

Clustering

Clustering

Fig. 12. Example

of two processes with clustering

effects.

probability Pi at a random delay time T,. This process is indicated by a and is called the delay point process. To make the process more general, an independent noise or subsidiary Poisson process with occurrence rate pi is generated and indicated by X . This process makes it possible to consider crane positions independent from each other (i.e. loadings which are not generated by the parent process). The relation between P,, p, pi and vi (an occurrence rate of the load Xi(t)) can be shown after Cox and Lewis [2] as vi =pi +P;p. (16) It appears that a superposition of A and X creates a simple stationary Poisson process with the occurrence rate expressed by Eq. (16). One of the most convenient ways for load combination analysis including the occurrence correlation of a bivariate point process is through the use of conditional occurrence rate (COR). It can be written [2,13] as follows: h:(t) = ~jm0(l/At)P{N(2)(t,t + At) 2 lJX,(t)isonat t = 0) (17) I -* where h\2( t) = the COR function of X,(t) given X,(t) is on at t = 0; N = number of occurrences of X,( t 1. The hy is similarly defined by switching indices 1 and 2. One can show that the following relation is fulfilled:

218

H. Pasternak

et al. /Structural

Safety

I7 (1996)

205-224

P---c

Pulse

from

CR2

fl

Pulse

:y

CR1

l-s c f,,=tm,
Fig. 13. Two situations of moving cranes.

where frLe71(t) defines PDF of the difference of the delay time T2 - T,; V, stands for mean occurrence rate of a load X,; Pi and p are defined in Eq. (16). Exponential PDFs for the independent random variables T, and T2 are considered. Two possible occurrences of moving cranes with the minimum distance between them are shown in Fig. 13. Assuming the Poisson type of upcrossings of the maximum value of load effects over a time period T one can get an approximation of the probability distribution P[ R,, < r in (0, T)] = FR,,(r, T)r = ew[ -+Gx,,,,(r) where v,~ can be specified
V12g
VlV2(Pd, + Pdz -

- v~TG~~,,,(~) v12TG x,+x2,(r)] A) (19)

by (see the derivtions


2tmin> +

in Appendix

PlP2
-PVlV2
a1

1
+a2 I

(20)

where a, denotes the mean value of the delay time of the process i, i.e. ai = E(q). It should be noted that after the rearranging of Eq. (20) one can apply Eq. (19) also to the situation when two cranes move in neighbouring bays. In this case, the term connected with the clustering effect in Eq. (20) as well as the shift between cranes should be omitted i.e. p = 0 and rmin = 0. In this way one obtains the probability distribution of the maximum load effect in which the simultaneous occurrence among independent loads is assumed [12,13]. 3.4. Numerical results for two cranes

Assume that Q,, and Q,, denote the appropriate fractiles (98% or 99%) for the particular crane load effects, QR m stands for the fractile calculated by means of Eq. (19). Note that Q,, is

H. Pasternak

et al. / Structurul

Safety

I7 (I 996) 205-224

219

500.0

____

----c--------,----~--~-,

400.0

ZI ------. -----.-..~~

Prob.'off Prob. Prob. Prob. Prob. Prob. Prob.

off off off off off off

= = = = = = =

0.00 0.05 0.10 0.20 0.40 0.60 0.90

I I I ' ' -; ,

0.0) 0.0

,,,,,,I,,; 50.0

,,,,,,,,, 100.0

iI,!

I,,,,

rkN1

15

? 0.0

Fig. 14. Mean upcrossing rate functions ui fupcossings during 8 hours). Type A of the frequency lifted load, rectangular PDF of the trolley position. Rectangular pulses, P, = 1.0.

distribution

of the

calculated on the basis of the PDF with a deterministic factor ( < 1) which refers to the position of the second crane according to Fig. 11. A load combination factor may be introduced as

Qh
=

Q,,

Qx2

On the basis of the calculated examples one can notice that: - The effect of different values of probability that loading is off on the mean upcrossing rate function vR+(Y) for rectangular pulse shapes is quite important as shown in Fig. 14. - Generally, the clustering effect between loads causes the G,,,Jr) to shift to the right and thus increases the fractiles significantly (compare Table 4).

Table 4 Combination Pulse shape 17 r\

factors 9. Case A-R-l, P, 1.0 0.0 1.0 0.0 1 - 4, 0.0 0.93 0.77 0.98 0.9

a, = 5 s, T = 10 days 0.1 0.93 0.98 0.2 0.92 0.98 0.4 0.92 0.97 0.6 0.91 0.97 0.8 0.86 0.53 0.95 0.57 0.99 0.72 0.87 0.999 0.57 0.74 -

220

H. Pasternak

et al. /Structural

Safety

17 (1996)

205-224

Fig. 15. Influence of the noise process on G,_(r). Type D of the frequency PDF of the trolley position. Trapezoid pulses: pi = 0.0, P, = 1.0.

distribution

of the lifted load, triangular

Increasing ratios of the noise occurrence rate to the parent process occurrence rate lead to a slight decrease of clustering effect (Fig. 15). Increase in the expected value of the delay time decreases the clustering effect. This influence is more conspicuous for smaller Y values (see Fig. 16). The maximum of the crane load combination effect increases with the period of time. This can be seen from Fig. 17. However, for sufficiently long periods the maximum values are no

Fig. 16. Effect of the delay time process. Type D of the frequency the trolley position. Trapezoid pulses, p, = 0.9, P, = 1.0.

distribution

of the lifted load, triangular

PDF of

H. Pasternak

et al. /Structural

Safety

I7 (19961 205-224

221

day
days days
days

Fig. 17. Influence of the period of time T. Type D of the frequency the trolley position. Trapezoid pulses, p, = 0.9, P, = 1.0.

distribution

of the lifted load,

triangular

PDF of

longer statistically independent. Because of this one should not extrapolate to periods beyond, e.g., 5 years. From tests it is known that the distribution of the load effects stabilizes after 1 month [6] or 1000-3000 crane passages [7].

4. Conclusions

The problem of the combination of load effects considered as column reactions due to overhead travelling cranes was investigated in this study. Cranes can work on the same crane track or in neighbouring bays. In the first case, processes with clustering effects were used to include the occurrence dependence between crane loads. The complementary CDF of the maximum value of combined load effects over a time period was discussed. Two types of pulse shapes (the trapezoid and its rectangular approximation) were considered with the purpose of comparing numerical results. For rectangular pulses the point-crossing method was also presented. The following conclusions can be drawn from the analysis: 1. In the paper the 99% (GRUSIBAU [5]) and 98% (EUROCODE 1 [4]) fractiles were considered as characteristic values of crane loads. According to the lifted load and geometrical relations, the 99%-fractiles give 80% to 92% of the maximum vertical forces (denoted by Rl3 in Table 3) for the uniform PDF of the trolley position and 85% to 94% for the triiigular PDF while 98%-fractiles range between the adequate intervals as follows: 75% to 90% and 80% to 92% respectively. The values from 1.09 to 1.33 (for the uniform PDF) and from 1.06 to 1.25 (for the triangular PDF) calculated on the basis of two different fractiles

222

H. Pasternak

et al. /Structural

Safety

I7 11996) 205-224

Table 5 Combination Pulse shape

factors $ for a, = 5 s and T = 10 days Case l- 4, 0.0 0.91 0.17 0.90 0.76 0.9 0.88 0.74 0.87 0.73

r\

D-T-l D-T-2 D-T-3 D-T-4

2. 3.

4.

5. 6.

give partial coefficients smaller than that suggested by DIN 18800/l [3] and EUROCODE 1 [4], i.e. yf= 1.5. Fractile values of the vertical crane forces calculated in the paper are more dependent on the distribution of the lifted load than the girder and crane spans. For many practical cases, especially for: * the independent operation of cranes (without clustering) * the occurrences (= lo%-50% of the time) of the load process with clustering effects (see Table 1 and 2 and 4) * larger crane and girder spans (compare Tables 2, 3 and 5) the combination factor + is smaller than 0.9 (given as a combination factor for two variable loadings according to DIN 18800/l and EUROCODE 1. The combination factors are closer to the values recommended by Russian Code [ll]. Time dependencies between the operations of two cranes (i.e. clustering effects) on one crane track may increase the load combination factor (compare Table 1 and 2 and 4). These dependencies should not be neglected. The dependency of the factor Q on the girder span is weaker than suggested by Bjelenja (compare Table 6 and Fig. 1). A cubic polynomial approximation is needed for describing the pulse shape if one takes into account the continuous beam model of the craneway. The maximum value of the pulse shape for such a model appears to be bigger than in the case of simply supported beams. The error introduced by the simplified model depends on the quantity e,,/l. For different cases analysed in the paper ( Table 1 and 2) the maximum error was not greater than 13%. However, as it was previously mentioned, the maximum value of the lifted load was always chosen 20% higher than the nominal load. This makes the results obtained herein certain to be on the safe side.
factors 4. Case C-R-1
T -9,

Table 6 Combination

Girder 3 0.94 0.89 0.94 0.88

span [m] 6 0.93 0.87 0.93 0.86 12 0.92 0.85 0.91 0.84 24 0.90 0.82 0.89 0.81

[days]
5 5 20 20 10 10 30 30

1-l
0.9 0.0 0.9 0.0

H. Pasternak

et al. /Structural

Safety

I7 (1996)

205-224

223

Acknowledgements The authors wish to thank the German the analysis. Research Society (DFG) for the financial support of

Appendix

A. Derivation

of the mean occurrence

rate of coincidence respectively

The PDFs f,,pT,(r) - Case A (7 2 d)

for two cases from Fig. 13 can be expressed

- Case B (7 2 0)

where a, = E(T,) is the mean value of the delay time of the process i. According to Fig. 13 the coincidence of processes X,(t) and X&t) mutually exclusive ways: 1. X,(t) is on at t = 7 and X,(t) is on in (7, 7 + d,), 2. X,(t) is on at t = T and X,(t) is on in (T, T + d,), where d, and d, stand for durations of pulses. Therefore one can write P (coincidence Id, , d 2) E
(~,~+ht)]

can happen

in two

=P[N(2(~,r+d,)~11X,(t)ison .P[X,(t)ison E (r,r+At)]

+P[Nc1)(r,~+d2)211X2(t)ison ~P[X,(t)ison = g2(+& + g,(d&& E (r,r+At)]

E (~,r+At)]

in which g, and g, indicate the conditional probabilities. After taking expectation with respect to d, and d, and dividing by At -+ 0 (in the limit sense), one obtains a first-order approximation of the mean rate of coincidence VI2 =g2b-%flh in which Pi, and Pi, +
&(PdJ~2

are the mean load effect durations.

224

H. Pasternak

et al. /Structural

Safety I7 (1996) 205-224

It can be shown that g*(x) = 1 - exp by case A one gets (for x = pd, - tmin)
2(Pd, min) -

and for the PDF described

If bd, k?,(kf,)

g2(Pd,)

l -

exP

i for g, indicated
min))

tmin)/a2 -K 1, then an approximation


= v2 + PIP2 Vl P ~ a1 ( Pd, +a2 -

by k2 is

i by analogy
&bd,)

one can write


= Vl +

and finally,
v12 g

the mean occurrence


Iv2(Pd, + Pd, 2trnin)

rate of coincidence
+ PIP2 -PVlV2 1 a1 +a2

is

References
[l] E.E. Belenja and A. Vasiljev, Osobiennosti dejstvitielnoj raboty podkranovych konstrukcyj (Singularity of structural behaviour under crane loads), Stavebnicky Cusopis, 24 (1976) 857-886 (in Russian). [2] D.R. Cox and P.A.W. Lewis, Multivariate point processes, Proc. 6th Berkeley Symp. on Mathematical Statistics, 1972, Vol. 3, pp. 401-448. 131 DIN 18800/l, Stahlbauten. Bemessung und Konstruktion (Steel structures. Design and calculation), Beuth-Verlag, Berlin (1990) (in German). [4] ENV 1991-1, Eurocode 1, Basis of design and actions on structures. Part 1, Basis of design, CEN/TC 250, 1992. [5] Grundlagen zur Festlegung uon Sicherheitsanforderungen fiir bauliche Anlagen GRUSIBAU (Basis for the assumption of safety requirements for buildings), Beuth-Verlag, Berlin (1981). [6] I.V. Izosimov et al., Issliedovanije silovych vozdiejstvij ot mostovych kranov (Investigations of actions on overhead travelling cranes), Metalliceskije Konstrukcji, Stroizdat (1966) 164-178. 171 A.I. Kikin et al., Povysenije dolgovecnosti metalliceskich konstrukcyj promyslennych zdanij (Extension of the time life of the steel structure of industrial buildings), Izd. po Stroit., Moscow, 1969. [8] U. Kiippe, Nutzlastkollektive von Kranen (Distribution of the lifted load for cranes), Hebezeuge und Fiirdermittel 21 (1981) 36-39. [9] R.D. Larrabee and CA. Cornell, Combination of various load processes, J. Strut. Die, ASCE, 107 (1) (1981) 223-239. [lo] H.O. Madsen, Load models and load combinatins, Ph.D. Thesis, Technical University of Denmark, 1979. [ill SNiP 11-6-74 (Russian Code), Stroitielnyje normy i pravila, Nagruzki i wozdiejstvija (Standards and regulations, Loads and actions), Stroizdat, Moscow, 1976. 1121 Y.K. Wen, Statistical combination of extreme loads, J. Strut. Die, ASCE, 103 (5) (1977) 1079-1093. [13] Y.K. Wen, Structural Load Modelling and Combination for Performance and Safety Evaluation, Elsevier, Amsterdam (1990).

You might also like