液压挖掘机的半自动控制系统@中英文翻译@外文翻译

.Automation in Construction 10 2001 477486rlocaterautconSemi-automatic control system for hydraulic shovelHirokazu Araya), Masayuki KagoshimaMechanical Engineering Research Laboratory, Kobe Steel, Ltd., Nishi-ku, Kobe Hyogo 651 2271, JapanAccepted 27 June 2000AbstractA semi-automatic control system for a hydraulic shovel has been developed. Using this system, unskilled operators canoperate a hydraulic shovel easily and accurately. A mathematical control model of a hydraulic shovel with a controller wasconstructed and a control algorithm was developed by simulation. This algorithm was applied to a hydraulic shovel and itseffectiveness was evaluated. High control accuracy and high-stability performance were achieved by feedback plusfeedforward control, nonlinear compensation, state feedback and gain scheduling according to the attitude. q2001 ElsevierScience B.V. All rights reserved.Keywords: Construction machinery; Hydraulic shovel; Feedforward; State feedback; Operation1. IntroductionA hydraulic shovel is a construction machinerythat can be regarded as a large articulated robot.Digging and loading operations using this machinerequire a high level of skill, and cause considerablefatigue even in skilled operators. On the other hand,operators grow older, and the number of skilledoperators has thus decreased. The situation calls forhydraulic shovels, which can be operated easily bywxany person 15 .The reasons why hydraulic shovel requires a highlevel of skill are as follows.1. More than two levers must be operated simulta-neously and adjusted well in such operations.)Corresponding author.E-mail address: arayahrknedo.go.jp H. Araya .2. The direction of lever operations is differentfrom that of a shovels attachment movement.For example, in level crowding by a hydraulicshovel, we must operate three levers arm, boom,.bucket simultaneously to move the top of a bucket.along a level surface Fig. 1 . In this case, the leveroperation indicates the direction of the actuator, butthis direction differs from the working direction.If an operator use only one lever and other free-doms are operated automatically, the operation be-comes very easily. We call this system a semi-auto-matic control system.When we develop this semi-automatic controlsystem, these two technical problems must be solved.1. We must use ordinary control valves for auto-matic control.2. We must compensate dynamic characteristicsof a hydraulic shovel to improve the precisionof control.0926-5805r01r$ - see front matter q2001 Elsevier Science B.V. All rights reserved.PII: S0926-5805 00 00083-2()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486478Fig. 1. Level crowding of an excavator and frame model of anexcavator.We have developed a control algorithm to solvethese technical problems and confirm the effect ofthis control algorithm by experiments with actualhydraulic shovels. Using this control algorithm, wehave completed a semi-automatic control system forhydraulic shovels. We then report these items.2. Hydraulic shovel modelTo study control algorithms, we have to analyzenumerical models of a hydraulic shovel. The hy-draulic shovel, whose boom, arm, and bucket jointsare hydraulically driven, is modeled as shown in Fig.2. The details of the model are described in thefollowing.2.1. Dynamic model 6Supposing that each attachment is a solid body,from Lagranges equations of motion, the followingexpressions are obtained:22J u qJ cos u yuuqJ cos u yuuqJ sin u yuuqJ sin u yuuyK sinu st. . . .11 l 12 1 2 2 13 1 3 3 12 1 2 2 13 1 3 3 1 1 122 J cos u yuuqJ u qJ cos u yuuyJ sin u yuuqJ sin u yuuyK sinu st. . . .12 1212223 23312 12123 233 2 22J cos u yuuqJ cos u yuuqJ u yJ sin u yuuqJ sin u yuuyK sinu st. . . .13 1 3 1 23 2 3 2 33 3 13 1 3 1 23 2 3 3 3 3 31.2.2where, J s m 1 q m q m 1 q I ; J s11 1 g123 112m 11 qm 11 ; J sm 11 ; J sm 12q21g 231g 313 31g 322 2g 2m 12qI ; J sm 11 ; J sm 12qI ; K s32 2 23 32g 333 3g 33 1.m 1 qm 1 qm 1 g; K s m 1 qm 1 g;1 g1213 2 2g 233K sm 1 g; and gsgravitational acceleration.33g 3u is the joint angle, t is the supply torque, 1 isiithe attachment length, 1 is the distance betweengithe fulcrum and the center of gravity, m is the massiof the attachment, I is the moment of inertia aroundithe center of gravity subscripts is13, mean boom,.arm, and bucket, respectively .2.2. Hydraulic modelEach joint is driven by a hydraulic cylinder whoseflow is controlled by a spool valve, as shown in Fig.3. We can assume the following:1. The open area of a valve is proportional to thespool displacement.2. There is no oil leak.3. No pressure drop occurs when oil flows ()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486 479Fig. 2. Model of hydraulic shovel.4. The effective sectional area of the cylinder isthe same on both the head and the rod sides.In this problem, for each joint, we have thefollowing equation from the pressure flow character-istics of the cylinder:ViAhsKXPysgn X P y P 2( . .ii 0 ii si i 1i 1iKwhen,K scB 2rg P sP yP0 ii 1i 1i 2 iwhere, A seffective cross-sectional area of cylin-ider; h scylinder length; X sspool displacement;P ssupply pressure; P scylinder head-side pres-si 1isure; P scylinder rod-side pressure; V soil vol-2 iiume in the cylinder and piping; B sspool width;igsoil density; Ksbulk modulus of oil; and csflow coefficient.2.3. Link relationsIn the model shown in Fig. 1, the relation be-tween the cylinder length change rate and the attach-ment rotational angular velocity is given as.follows: 1 boomf u.11h1su1OA OC sin u qb.11 1 1sy ,22(OA qOC q2OA OC cos u qb.11 1111.2 arm.f u ,u212h2su yu21.OAOCsin u yu qb qa2222 2 1 2 2sy ,22( .OAqOCq2OAOCcos u yu qb qa22 22 2222 2 1 2 2.3 bucketwhen ODsOBsBCsCD33 33 33 33YhABCsin u yu qg ya qu.3333232f u ,u ssy . 3.32322 Yu yu (AB qBC q2ABBCcos u yu qg ya qu.3233 33 3333 3 2 3 3 2()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486480Fig. 3. Model of hydraulic cylinder and valve.2.4. Torque relationsFrom the link relations of Section 2.3, the supplytorque t is given as follows, taking cylinder frictioniinto consideration:t syf u P1 A qf u ,u P1 A. .111 121222qf u ,u P1 A y Cfuu. .323 33 c11 1 1qsgn u Ffu 4. ./1111t syf u ,u P1 A y Cfu ,uuyu. . /221222c22 1 2 2 1qsgn u yu Ffu ,u.5/212212t syf u ,u P1 A y Cfu ,uuyu. . /332333c33 2 3 3 2qsgn u yu Ffu ,u .5/323323Where, C is the viscous friction coefficient andciF is kinetic frictional force of a cylinder.i2.5. Response characteristics of the spoolSpool action has a great effect on control charac-teristics. Thus, we are assuming that the spool hasthe following first-order lag against the referenceinput.1XX s X .5.iiTSq1spiWhere, XXis the reference input of spool dis-iplacement and T is a time constant.spi3. Angle control systemAs shown in Fig. 4, the angle u is basicallycontrolled to follow the reference angle u by posi-gtion feedback. In order to obtain more accuratecontrol, nonlinear compensation and state feedbackare added to the position feedback. We will discussdetails of these algorithms as follows.3.1. Nonlinear compensationIn the ordinary automatic control systems, newcontrol devices such as servo valves are used. In oursemi-automatic system, in order to realize the coexis-tence of manual and automatic operations, we mustuse the main control valves, which are used inmanual operation. In these valves, the relation be-tween spool displacement and open area is nonlinear.Then, in automatic operation, using this relation, thespool displacement is inversely calculated from therequired open area, and the nonlinearity is compen-.sated Fig. 5 .Fig. 4. Block diagram of control system u .()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486 481Fig. 5. Nonlinear compensation.3.2. State feedbackBased on the model discussed in Section 2, if thedynamic characteristics for boom angle control arelinearized in the vicinity of a certain standard condi-tion spool displacement X , cylinder differential10.pressure P , and boom angle u , the closed-loop110 10transfer function can be expressed byKpu s ug 6.1132asqasqasqK210pwhere, K is position feedback gain; andpf u ACfu. .110 1 c11 10a sq02 APyP.KPyP(1 s111001 s1110XJqJ cosuXqJ cos uXquXDC44.10 11 12 2 13 2 3a s12 Af u P yP. .11 10 s1 110Cfu V.c11 10 1qAKK P yP(101s1110VJqJ cosuXqJ cos uXquXDC44.111 12 2 13 2 3a s .2Af u KK P yP. (1 1 10 01 s1 110This system has a comparatively small coefficienta , so the response is oscillatory. For instance, if in1our large SK-16 hydraulic shovel, X is 0, the10coefficients are given as a s2.7=10y2, a s6.001=10y6, a s1.2=10y3. Addingthe acceleration2feedback of gain K , to this the upper loop in Fig.a.4 , the closed loop transfer function is given asKpu s u .7.1 r132asq a qKsqasqK.21a 0 pAdding this factor, the coefficient of s2becomeslarger, thus, the system becomes stable. In this way,acceleration feedback is effective in improving theresponse characteristics.However, it is generally difficult to detect acceler-ation accurately. To overcome this difficulty, cylin-der force feedback was applied instead of accelera-.tion feedback the lower loop in Fig. 4 . In this case,cylinder force is calculated from detected cylinderpressure and filtered in its lower-frequency portionwx7,8 . This is called pressure feedback.4. Servo control systemWhen one joint is manually operated and anotherjoint is controlled automatically to follow the manualoperation, a servo control system must be required.For example, as shown in Fig. 6, in the level crowd-ing control, the boom is controlled to keep the arm.end height Z calculated from u and u to refer-12ence Zr. In order to obtain more accurate control, thefollowing control actions are ()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486482.Fig. 6. Block diagram of control system Z .4.1. Feedforward controlCalculating Z from Fig. 1, we obtainZs1 cosu q1 cosu .8.1122.Differentiating both sides of Eq. 8 with respectto time, we have the following relation,Z 1 sinu22u sy y u .9.121 sinu 1 sinu11 11The first term of the right-hand side can be taken.as the expression feedback portion to convert Z tou , and the second term of the right-hand side is the1expression feedforward portion to calculate howmuch u should be changed when u is changed12manually.Actually, u is determined using the difference2value of Du . To optimize the feedforward rate,2feedforward gain K is tunned.ffThere may be a method to detect and use the arm.operating-lever condition i.e. angle instead of armangular velocity, since the arm is driven at an angu-lar velocity nearly proportional to this lever condi-tion.4.2. Adaptie gain scheduling according to the atti-tudeIn articulated machines like hydraulic shovels,dynamic characteristics are greatly susceptible to theattitude. Therefore, it is difficult to control the ma-chine stably at all attitudes with constant gain. Tosolve this problem, the adaptive gain schedulingaccording to the attitude is multiplied in the feedback.loop Fig. 6 . As shown in Fig. 7, the adaptive gain.KZ or Ku is characterized as a function of twovariables, uXand Z. uXmeans how the arm is22extended, and Z means the height of the bucket.5. Simulation resultsThe level crowding control was simulated byapplying the control algorithm described in Section 4to the hydraulic shovel model discussed in Section 2.In the simulation, our large SK-16 hydraulic shovel.was employed. Fig. 8 shows one of the results. Fiveseconds after the control started, load ()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486 483Fig. 7. Gain scheduling according to the attitude.was applied stepwise. Fig. 9 shows the use of feed-forward control can reduce control error.6. Semi-automatic control systemBased on the simulation, a semi-automatic controlsystem was manufactured for trial, and applied to theSK-16 shovel. Performance was then ascertained byfield tests. This section will discuss the configurationand functions of the control system.6.1. ConfigurationAs illustrated in Fig. 10, the control system con-sists of a controller, sensors, manmachine interface,and hydraulic control system.The controller is based on a 16-bit microcomputerwhich receives angle input signals of the boom, arm,and bucket from the sensor; determines the conditionof each control lever; selects control modes andcalculates actuating variables; and outputs the resultsfrom the amplifier as electrical signals. The hy-Fig. 8. Simulation result of level crowding.draulic control system generates hydraulic pressureproportional to the electrical signals from the electro-magnetic proportional-reducing valve, positions thespool of the main control valve, and controls theflow rate to the hydraulic cylinder.In order to realize high-speed, high-accuracy con-trol, a numeric data processor is employed for theFig. 9. Effect of feedforward control on control error of Z.()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486484Fig. 10. Schema of control system.controller, and a high-resolution magnetic encoder isused for the sensor. In addition to these, a pressuretransducer is installed in each cylinder to achievepressure feedback.The measured data are stored up to the memory,and can be taken out from the communication port.6.2. Control functionsThis control system has three control modes,which are automatically switched in accordance withlever operation and selector switches. These func-tions are the following.1 Level crowding mode: during the manual armpushing operation with the level crowding switch,the system automatically controls the boom and holdsthe arm end movement level. In this case, the refer-ence position is the height of the arm end from theground when the arm lever began to be operated.Operation of the boom lever can interrupt automaticcontrol temporarily, because priority is given to man-ual operation.2 Horizontal bucket lifting mode: during themanual boom raising operation with the horizontalbucket lifting switch, the system automatically con-trols the bucket. Keeping the bucket angle equal tothat at the beginning of operation prevents materialspillage from the bucket.3 Manual operation mode: when neither thelevel crowding switch nor the horizontal bucket lift-ing switch are selected, the boom, arm, and bucketare controlled by manual operation only.The program realizing these functions is primarilywritten in C language, and has well-structured mod-ule to improve maintainability.7. Results and analysis of field testWe put the field test with the system. We con-firmed that the system worked correctly and theeffects of the control algorithm described in Chaps. 3and 4 were ascertained as follows.7.1. Automatic control tests of indiidual attach-mentsFor each attachment of the boom, arm, and bucket,the reference angle was changed 58 stepwise fromthe initial value, and the responses were measured;thus, the effects of the control algorithm described inChap. 3 were ()H. Araya, M. KagoshimarAutomation in Construction 10 2001 477486 485Fig. 11. Effect of nonlinear compensation on boom angle.7.1.1. Effect of nonlinear compensationFig. 11 shows the test results of boom lowering.Because dead zones exist in the electro-hydraulicsystems, steady-state error remains when simple po-sition feedback without compensation is applied OFF.in the figure . Addition of nonlinear compensation.ON in the figure can reduce this error.7.1.2. Effect of state feedback controlFor the arm and bucket, stable response can beobtained by position feedback only, but adding ac-celeration or pressure feedback can improve fast-re-sponse capability. As regards the boom, with onlythe position feedback, the response becomes oscilla-tory. Adding acceleration or pressure feedback madethe response stable without impairing fast-responsecapability. As an example, Fig. 12 shows the testresults when pressure feedback compensation wasapplied during boom lowering.7.2. Leel crowding control testControl tests were conducted under various con-trol and operating conditions to observe the controlFig. 12. Effect of pressure feedback control on boom angle.Fig. 13. Effect of feedforward control on control error of Z.characteristics, and at the same time to determine theoptimal control parameters such as the control gains.shown in Fig. 6 .7.2.1. Effects of feedforward controlIn the case of position feedback only, increasinggain K to decrease error DZ causes oscillation duepto the time delay in the system, as shown by AOFFBin Fig. 13. That is, K cannot be increased. Apply-ping the feedforward of the arm lever value describedin Section 4.1 can decrease error without increasingK as shown by AONB in the figure.p7.2.2. Effects of compensation in attitudeLevel crowding is apt to become oscillatory at theraised position or when crowding is almost com-pleted. This oscillation can be prevented by changinggain K according to the attitude, as has beenpdiscussed in Section 4.2. The effect is shown in Fig.14. This shows the result when the level crowdingwas done at around 2 m above ground. Compared tothe case without the compensation, denoted by OFFin the figure, the ON case with the compensationprovides stable response.Fig. 14. Effect of adaptive gain control on control error of Z.()H. Araya, M. KagoshimarAutomation in Construction 10 2001 4774864867.2.3. Effects of control interalThe effects of control interval on control perfor-mance were investigated. The results are:1. when the control interval is set to more than100 ms, oscillation becomes greater at attitudeswith large moments of inerti