Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 41, 4, pp. 873-885, Warsaw 2003 LOAD POSITIONING AND MINIMIZATION OF LOAD OSCILLATIONS IN ROTARY CRANES Andrzej Maczyński Department of Mechanics and Computer Methods, University of Bielsko-Biała e-mail: amaczynski@ath.bielsko.pl In the paper a method of load positioning and minimization of load final oscillations in rotary cranes has been presented. Drive functions of the slewing of an upper structure have been determined by means of dynamic optimisation. In the optimisation task a simplified model has been used. A completely stiff supporting structure of the crane has been assumed. A feedback control system has been proposed to compensa- te for influences of flexibilities that have not been taken into account during the optimisation and for inaccurate knowledge of parameters of the model. The effectiveness of the control system for two different con- trolled variables has been analysed.A special coefficient for quantitative analysis has been proposed. Results of numerical simulations have been presented. Key words: crane, positioning, control, numerical simulation 1. Introduction The limitation of load oscillations during the slewing motion of a crane upper structure and, especially, during the final load positioning is a vital problem for the effectiveness and occupational safety of the reloading and assemblywork. This explains why the problemofminimization of load oscilla- tions in rotary cranes is taken up inmanypapers (Abdel-Rahman andNayfeh, 2002;Balachandran et al., 1999;Bednarski et al., 1997;Kłosiński, 2000; Parker et al., 1995; Sakawa et al., 1981). Mobile cranes belong to the kind of cranes for which accident hazard is extreme (Neitzel et al., 2001; Yow et al., 2000). That is why the author has been carrying out research connected with load positioning at the end of the slewing motion of a mobile telescopic crane for 874 A.Maczyński several years. One of the most important assumptions in author’s investiga- tions is that it is possible to determine the drive functionswhichminimize the final load oscillations by using optimisation methods and a simplified model. In this simplified model, a completely stiff supporting structure of the crane may be assumed. The influence of flexibilities that are not taken into account during the optimisation, and inaccurate knowledge of parameters of themodel can be compensated by a feedback control systemwith a PID controller. The appropriate choice of a controlled variable is fundamental for the operation efficiency of the control system. This variable should also be easily measured. Comparison of the control system effectiveness for two different controlled variables is presented in the paper. 2. Determination of functions of the upper structure slewing. The dynamic model of a crane Drive functions of the upper structure slewing have been determined from the point of view ofminimization of final load oscillations and, simultaneously, load positioning (Maczyński andWojciech, 2000). In this task a simplifiedmo- del, taking into account the complete stiffness of the crane supporting structu- re, has been used. TheNelder-Meadsmethod has been applied. The objective function is defined as follows F =C1 1 2 mLv 2 LT +C2‖rLT −rLF‖ 2 (2.1) where rLT = rL ∣ ∣ ∣ t=T vLT = vL ∣ ∣ ∣ t=T and mL – mass of the load rLF – vector of expected load coordinates at the moment t=T rL – vector of load coordinates vL – velocity of the load T – time of the upper structure’s slewing C1,C2 – coefficients (weights). This objective function means that one can expect that at the end of the slewing motion the load is at a particular point in space and, furthermore, its kinetic energy is minimal. Values of the coefficients C1 and C2 have been Load positioning and minimization... 875 determined during numerical simulations. The main criterion of this deter- mination was the best quality of load positioning at the end of the slewing motion. Fig. 1. Flexible model of a mobile crane Figure 1 shows themodel of a crane that has been used to performnumeri- cal simulations presented below. It is a 3Dmodel which takes into account the flexibility of the crane supporting structure and damping in selected subsys- tems. It allows dynamic analysis of a telescopic mobile crane in every respect, and is described in detail inMaczyński and Szczotka (2002). 3. The feedback control system – determination of controlled variables In this section the feedback control system, which compensates for the influenceof theflexibilities ignoredduring the optimisation of loadpositioning, is briefly discussed. The control system can also eliminate the influence of inaccurate knowledge of parameters of the model. Figure 2presents ablockdiagramof the control system. It is aprogrammed control system. Functions of the controlled variable Φt have been calculated for the simplifiedmodel, the optimal drive function and for certain operating parameters. The analysis has been carried out for two different controlled variables Φt: • angle ϕL of deflection of the load • angle τ of tangential deflection of the load. 876 A.Maczyński Fig. 2. Block diagram of the control system; ϕO – slewing angle of the crane determined according to Maczyński andWojciech (2000) (optimal for the simplified model), Φt – controlled variable determined for the simplified model, Φb – actual value of the controlled variable 3.1. Angle ϕL of deflection of the load as the controlled variable The definition of the angle ϕL of deflection of the load is presented in Fig.3. The plane marked out by the coordinate system 0′X′Y ′ is parallel to the ground. The origin of the coordinate system lies on the axis of rotation of the crane upper structure, the axis 0′X′ is parallel to the axis X in Fig.1. Points G′ and L′ define the projections of the end of the jib and the load on the plane X′Y ′, respectively. Fig. 3. Angle ϕL of deflection of the load Author’s earlier research demonstrates that assuming the angle ϕL as the controlled variable can give satisfactory results of numerical analysis. The ti- me course of the angle ϕL corresponds well with load oscillations, especially tangential ones, and enables them to be effectively eliminated. However, me- asurement of the angle ϕL in a real crane is inconvenient and requires the use of expensive measuring devices, for instance laser devices. This solution Load positioning and minimization... 877 would mean a significant increase in the price of the crane, especially in the case of small simple versions. Therefore, another parameter that could be used as the controlled variable has been sought; it is assumed that the angle τ of tangential deflection of the load can be treated as the controlled variable. 3.2. Angle τ of tangential deflection of the load as the controlled variable The diagram in Fig.4 shows the end of the jib (point G) and the load L that are linked by the rope (GL). The angles τ of tangential and β of radial deflection of the rope are marked in the figure. Fig. 4. Angle τ of tangential deflection of the load 4. Numerical calculations Masses and geometrical parameters of themodel have been chosen for nu- merical calculations based upon the technical documentation of theDUT0203 crane.Themass of the loadwas equal to 3000kg, the crane radius about8.7m. The control system efficiency in compensating the influence of the flexibilities of the crane supporting structure has been analysed for both controlled varia- bles for the following cases of motion: • slewing by 60◦ over a period of 12s • slewing by 75◦ over a period of 13.5s • slewing by 90◦ over a period of 15s. 878 A.Maczyński In graphs presented in this paper, PID I denotes courses obtained for the control system with the angle ϕL as the controlled variable, PID II – with the angle τ. The selection of controller settings has been carried out using the criterion ofminimizing load oscillations based on the positioning efficiency coefficient PEG = √ x20+y 2 0 (4.1) where x0 = |xL−x ′ LF |max y0 = |yL−y ′ LF |max and xL,yL – coordinates of the load x′ LF ,y′ LF – expected load coordinates for t=T |xL−x ′ LF |max, |yL−y ′ LF |max – maximal absolute value of the diffe- rence between the coordinates after the end of the slewing motion. The load coordinates x′ LF ,y′ LF expected at the time t = T have been determined for the static load of the crane for a given final angle of the slewing andmass of the load.During calculation of the coefficient value, it is important that the time of load observation after the end of the crane slewing must be longer than the time of load oscillations. Numerical simulations showed that good quality of load positioning, for all three analysed motions, can be obtained using the controller type P. Ap- plication of integrating and/or differential elements of the controller did not cause significant improvement of the positioning efficiency. In the analyses it has been assumed that the proportional gain Kc was equal to 20 for PID I and 0.95 for PID II. Figure 5 shows the final parts of the load trajectories for motions 1-3, respectively. In graphs, besides curves denoted byPID I andPID II, the curve obtained for the drive without the feedback control system is also presented. Additionally, point � is the theoretical point at which the load should stop at the end of the slewing motion. This theoretical point has been determined for the static load and for themodel taking into consideration the flexibilities of the crane supporting structure. In Table 1 values of the PEG coefficient for the analysed cases are com- pared. Load positioning and minimization... 879 Fig. 5. Slewing by 60◦ (a), 75◦ (b) and 90◦ (c) Table 1.Comparison of PEG coefficient values Angle PEG coefficient value [cm] without control system PID I PID II 60◦ 4.37 2.57 1.56 75◦ 4.59 2.71 1.81 90◦ 4.42 2.70 2.21 The data in Table 1 show considerable differences between the values ob- tained for the control system PID I and PID II. According to the case of the slewing analysed, the differences range from 22% to 65%. The assessment of the final load positioning carried out bymeans of observation of the load tra- jectories does not confirm these differences (Fig.5). The quality assessment in fact shows that the final load oscillations are more or less equal for both feedback control systems. However, a certain displacement between the load 880 A.Maczyński trajectories after the end of the crane slewing can be seen. This displacement causes differences in values of the coefficient PEG, as shown in Table 1. The PEG coefficient was defined for the drivewithout the feedback control system. In this type of drive, the expected load position after the end of the motion is identical to the load position in static conditions. However, in the case of the drive with the feedback control system, the expected load position depends on the controlled variables. The angle ϕL contains information about the load position in static conditions. When determining the desired value of the angle ϕL, the simplified model omitting the flexibilities of the crane structure is used. Thus, the feedback control system tries to change the final load position and to place the load as in the simplified model. However, the angle τ contains information only about the relative position of the load and the end of the jib. In this case, the expected load position after the end of the motion is in agreement with the state of equilibrium for themodel taking into account the flexibilities. The graph inFig.6 confirms this hypothesis. Figure 6 shows the upper structure slewing by 90◦. In addition to curves from Fig.5c, the projection of the load trajectory for the drive without the control system obtained for the simplified model is presented. It is denoted as ”simp”. Fig. 6. Comparison of load trajectories for slewing by 90◦ The differences occurring between the values of the positioning efficiency coefficient PEG led the author to define the coefficient once again. This new coefficient is denoted as PE PE = √ (xmax−xmin 2 )2 + (ymax−ymin 2 )2 (4.2) Load positioning and minimization... 881 where xmax, xmin, ymax, ymin –maximal andminimal values of the load coor- dinates registered after the end of the slewing of the upper structure, Fig.7. Fig. 7. Geometrical interpretation of the PE coefficient Ageometrical interpretation of the PE coefficient is presented in Fig.7. It is the radius of the circle circumscribed onto the rectangle in which the load trajectory (after the end of slewing) is inscribed. Values of the PE coefficient calculated for described motions are compared in Table 2. Table 2.Comparison of PE coefficient values Angle PE coefficient value [cm] PID I PID II 60◦ 1.64 1.51 75◦ 1.86 1.76 90◦ 2.25 2.19 Results presented in Table 2 show that the PE coefficient lacks the PEG coefficient error. It is in agreement with the quality assessment performed by means of observing the load trajectories. Figure 8 presents dependencies of the PE coefficient values on the pro- portional gain Kc. The curves have been obtained for cases of motion 1-3. Graphs in Fig.8a and Fig.8b relate to the PID I control system (in Fig.8b the proportional gain Kc is held within the range 0÷5), Fig.8c relates to the PID II control system. 882 A.Maczyński Fig. 8. PE coefficient versus proportional gain Kc; (a) – PID I, (b) – PID I and Kc =0÷5, (c) – PID II Graphs in Fig.8 show that the control system with the angle ϕL (PID I) as the controlled variable is less sensitive to the selection of the proportional gain Kc. For the proportional gain Kc in the range 5÷50 and for all three analysedmotions, the PE coefficientmaintains an almost constant value.This value is only slightly greater than theminimal value of the PE coefficient. The minimal values of the PE coefficient are obtained for lower values of the pro- portional gain Kc. Theproportional gain Kc that ensures theminimumof the PE coefficient is different for every slewing. Thus, it is impossible to obtain theminimal value of the PE coefficient for one value of the proportional gain Kc and for all cases of motion. When the value of the proportional gain Kc exceeds 60, then sudden deterioration in the positioning quality occurs. In the case of the control system with the τ angle (PID II) as the controlled varia- Load positioning and minimization... 883 ble, the most advantageous values of the PE coefficient are obtained for the following ranges of the proportional gain Kc: motion 1 – 〈0.8,1.05〉, motion 2 – 〈0.7,1〉 andmotion 3 – 〈0.6,1〉. For the PID II, sudden deterioration in the positioning quality occurs when the proportional gain Kc exceeds 1.05. 5. Summary Numerical simulations prove that the feedback control system presented in the paper effectively compensates for the influence of the flexibility of the crane supporting structure on the final load positioning. Its effectiveness does not depend on the choice of the controlled variable. Results obtained for both angles: ϕL of deflection of the load, and τ of tangential deflection, are com- parable. The feedback control system minimizes the load oscillations in the tangential direction very effectively but it leaves some in the radial direction. It is impossible to completely eliminate the load oscillations in two directions using for this purpose only one drivemotion – the slewingmotion of the upper structure. In his future research, the author plans to elaborate a conception of a simple supplementary system that should eliminate the remaining radial oscillations. From the practical point of view, it is important that taking the τ angle as the controlled variable has effects similar to those obtained for the angle ϕL. The only real difference between the results obtained is the small displacement discussed in Section 4. The value of the displacement does not exceed 2cm, so it is a negligible variable for an operator. The operator is mainly interested in elimination of the final load oscillations. Fig.8 shows that the PID I system is more convenient in terms of the controller setting selection; it enables the posi- tioning quality to be improved over a wide range of values of the proportional gain Kc. For the PID II, this range is much narrower and, even if exceeded only slightly, can cause clear deterioration of the load positioning quality. The analysis of the results shows that the positioning efficiency coeffi- cient PEG, used so far by the author, can, in the case of the drive with the feedback control system, be misleading. That is why the new universal coef- ficient PE has been proposed. This new coefficient can be useful during the assessmentof the loadpositioning fordriveswith orwithout the control system and any controlled variable. On the basis of the graphs showing dependence of the PE coefficient on the Kc proportional gain (Fig.8), it is easy to determine the most advantageous settings of the controller. Simulations indicate that a 884 A.Maczyński considerable improvement in the quality of the load positioning is possible for a wide range of drives, using only one value of the controller settings. Theproposedmethod of the load positioning and oscillationminimization, which has been discussed using the example of a telescopic mobile crane, can be also utilised in the case of other rotary cranes. References 1. Abdel-Rahman E.M., Nayfeh A.H., 2002, Pendulation reduction in boom cranes using cable length manipulation,Nonlinear Dynamics, 27, 3, 255-269 2. BalachandranB., Li Y.-Y., FangC.-C., 1999,Amechanical filter concept for controlofnon-linear crane-loadoscillations,Journal of Sound andVibration, 228, 3, 651-682 3. Bednarski S.,Cink J.,TomczykJ., 1997,Pozycjonowanie ładunkuwruchu roboczym żurawia portowego,Proc. X Konferencji Problemy Rozwoju Maszyn Roboczych, part I, 25-31 4. Kłosiński J., 2000, Slewing motion control in mobile crane ensuring stable positioning of carried load, The Archive of Mechanical Engineering, XLVII, 119-138 5. Maczyński A., SzczotkaM., 2002, Comparison ofmodels for dynamic ana- lysis of amobile telescopic crane,Journal of Theoretical andAppliedMechanics, 40, 4, 1051- 1074 6. Maczyński A., Wojciech S., 2000, Selection of function describing crane slewingmotion in order tominimize load oscillations,Proc. XII Conference on ”Drives, Control and Automation ofWorkingMachines andVehicles”, 317-326 (in Polish) 7. Neitzel R.L., Seixas N.S., Ren K.K., 2001,A review of crane safety in the construction industry, Applied Occupational and Environmental Hygiene, 16, 12, 1106-1117 8. ParkerG.G.,PettersonB.,DohrmannC., RobinettR.D., 1995,Com- mand shaping for residualvibration free cranemaneuvers,Proc.AmericanCon- trol Conference, Seattle,Washington, 934-938 9. Sakawa Y., Shindo Y., Hashimoto Y., 1981, Optimal control of a rotary crane, Journal of Optimzation Theory and Applications, 35, 535-557 10. Yow P., Rooth R., Fry K., 2000, Crane accidents 1997-1999: a report of the crane unit of the division of occupational safety and health, Division of Occupational Safety andHealth, CaliforniaDepartment of Industrial Relations Load positioning and minimization... 885 Pozycjonowanie i minimalizacja wahań ładunku w żurawiach obrotowych Streszczenie W pracy zaprezentowanometodę pozycjonowania i minimalizacji końcowychwa- hań ładunku w żurawiach obrotowych. Funkcja napędowa obrotu nadwozia została dobrana na drodze optymalizacji dynamicznej dla uproszczonegomodelu zakładają- cego całkowitą sztywność układu nośnego żurawia. Zaproponowano układ regulacji kompensujący wpływ nieuwzględnionych podczas optymalizacji podatności i niedo- kładnej znajomości parametrówmodelu. Analizowano działanie układu regulacji dla dwóch różnych wielkości zadanych. Do ilościowej oceny jakości pozycjonowania za- proponowano specjalny wskaźnik. Zaprezentowano wyniki symulacji numerycznych. Manuscript received May 5, 2003; accepted for print June 16, 2003