【作者】 刘晓峰；

【导师】 王龙山； 刘昕晖；

【作者基本信息】 吉林大学 ， 机械制造及其自动化， 2012， 博士

【摘要】 液压起升系统是履带起重机最重要的组成部分，它的安全性及稳定性直接关系到整机的工作性能，也是评价一台起重机性能优劣的重要指标。为了保证工作的可靠性，大型履带起重机常常采用单钩双卷扬的起升结构，即由两个结构、参数相同的液压马达共同提升同一个吊钩，完成对重物的起吊工作。但由于液压波动、系统泄漏，外部干扰等因素的影响，常常出现同步误差，如何保证履带起重机双卷扬系统的同步控制精度是摆在研究人员面前的首要问题。就目前液压同步控制方法而言，大多数现有的控制方法往往过于复杂，或者是附加条件过多，并带有一定针对性，在应用上受到很大的局限。因此，本文从简便实用的观点出发，结合校企合作项目对450t履带起重机做了大量的分析与研究工作。研究的目的就是为了寻求一种合适的控制规律，使双马达同步控制系统得到较高的同步控制精度。本文主要研究工作如下：1、在建立系统数学模型的基础上，对液压起升系统及其关键元件的结构和原理进行深入研究，并通过传递函数法对系统的动态特性进行分析，找出影响双马达同步控制精度的相关因素。针对平衡阀阀口通流面积梯度对平衡阀动态特性的影响，提出一种新的平衡阀阀口结构形式，以改善阀口通流面积梯度的突变对液压波动的影响；同时采取优化平衡阀和马达压力切断阀的结构参数等措施，降低系统压力波动的幅度，以减少压力波动对同步控制精度的影响。仿真和实验结果证明，优化设计方案有效，可行。2、寻求一种新型的同步控制方法是本文的核心内容。神经网络具有自学习功能，不需要对被控对象进行精确的辨识和建模，就可以用简单的方法实现对复杂系统的有效控制。本文创新性地将神经网络智能控制策略应用于履带起重机的双卷扬控制系统中，并把单神经元控制与传统PID控制相结合，提出一种单神经元PID控制策略。把神经元的连接权重分别与PID三个控制参数相对应，可以实时对其进行调节，克服了传统PID控制参数不能在线自动整定的不足，从而适应实际工作过程中环境的不断变化。3、依靠神经网络自学习、自适应功能，采用有监督Delta学习规则，按差值最小准则连续地对连接强度进行修正，使期望输出和实际输出的差值与两个神经元之间连接权值的变化量成正比，有效地加快收敛速度，具有控制简单，容易实现、鲁棒性强，同步控制精度高等优点。4、单神经元控制是通过对连接权重进行调整的一个非线性优化调节过程，即权值是以系统的误差函数相应于其负梯度方向来进行自动调节的。所以，本文采用广义Lyapounoy非线性稳定性理论对单神经元PID控制系统进行稳定性分析，总结出提高系统稳定性的方法，即尽量使学习速率ηi取小值，以提高系统的稳定性。采用这种方法不必求解系统的微分方程，就可以进行稳定性判别，简单、可靠。5、提出一种交叉耦合同步控制方式用于履带起重机双卷扬系统中，在两个子系统的输出量都作为反馈信号的同时，把输出量的差值也作为一个附加的反馈信号进行跟踪比较，最终实现同步控制的目的。这种控制方式较同等方式和主从方式相比，具有更快的收敛速度，并能很好的适应负载变化，具有较高的同步控制精度。6、为了验证理论分析的正确性，以及采用控制策略的合理性，应用AMESim软件与Matlab/Simulink软件进行联合仿真。并把传统PID控制和单神经元PID控制效果进行对比，得出最终结论，即：单神经元PID的控制性能优于传统PID控制，更具智能性。本文所提出的控制策略基本符合最初设计目的，为液压同步控制提供了新的思路。更多还原

【Abstract】 Hydraulic lifting system is a most important part of crawler crane, itssecurity, stability and handling performance is directly related to the machine’sperformance, it is also an important indicator of the performance merits of theevaluation of a crane. In order to ensure the reliability of the work, a largecrawler crane is often used single hook dual winch lifting structure consists oftwo structures, the same parameters of the hydraulic motor upgrade with ahook to complete the lifting of heavy objects. However, due to hydraulicfluctuations, the system leakage and external interference factors there areoften the synchronization error, how to ensure the synchronization controlaccuracy of the double-winch system to become the most important issuebefore the designerHowever, hydraulic synchronization control methods, most of the existingcontrol methods are often too complex or too many conditions attached, withsome targeted and significant limitations on the application. Therefore, startingfrom the simple and practical point of view, the combination of analysis andresearch work done a lot of school-enterprise cooperation project of 450tcrawler crane, the purpose of the study is to find a suitable control law,dual-motor synchronous control to obtain a higher synchronization accuracy.The main work is as follows:1. The system mathematical model based on the structure and principlesof the hydraulic lifting system and its key components in-depth study andanalysis on the dynamic characteristics of the system transfer function, find outthe impact of dual motor synchronous control The accuracy of the relevantfactors.Gradient balancing valve dynamic characteristics of the flow area for thebalance valve valve port, put forward a new kind of balance valve valve port structure to improve the valve port through flow area gradient mutations affectthe hydraulic fluctuations; taken to optimize the balance valve and motorpressure shut-off valve structure parameters and other measures to reduce thevolatility of the system pressure in order to reduce pressure fluctuations in thesynchronization control accuracy. Proved through simulation analysis andexperimental studies to optimize the design of the program is effective andfeasible.2. To seek a new type of synchronization control method is the corecontent of this article. The neural network has the accurate identification andmodeling of self-learning function, does not require the controlled object, youcan use a simple method to achieve effective control of complex systems. Thisinnovation to single-neuron intelligent control strategy applied to crawlercranes, dual hoist control system and neural network control with conventionalPID control the combination, for a single neuron PID control strategy. Neuronconnection weights to three and PID control parameters corresponding to thereal-time be adjusted to overcome the traditional PID control parameters cannot be on-line automatic tuning deficiencies, and thus adapt to the changingenvironment in the actual work process.3. Relying on self-learning neural network, adaptive function, the use ofsupervision Delta learning rule, the difference between the minimum criteria forcontinuous correction of the connection strength, so that the differencebetween the desired output and actual output and the connection weightsbetween two neurons the change is proportional to the amount of effectivelyspeed up the convergence, with a simple control and easy to implement,robust and strong, synchronized control of high precision.4. The single neuron control is to optimize the adjustment process througha non-linear adjustment of connection weights, the weights are the errorfunction corresponding to its negative gradient direction to automatically adjust.Therefore, the the generalized Lyapounoy nonlinear stability theory for stabilityanalysis of single neuron PID control system, summed up the method to improve system stability, that is, as far as possible, the learning rate to take asmall value in order to improve system stability. Using this method do not haveto solve the system of differential equations, stability of discrimination, simple,reliable.5. Cross-coupling synchronization control for crawler crane winch systemin two subsystems of the output as the feedback signal at the same time, thedifference between the two output as an additional feedback signal trackingeventually achieve the synchronization control purposes. This control methodthan the same manner as compared to the master-slave mode, with a fasterconvergence speed, and are well suited to the load changes, has a highsynchronization control accuracy.6. In order to verify the theoretical analysis of the correctness of rationality,and control strategy of the AMESim software and MATLAB / Simulink softwareco-simulation, and the traditional PID control and the single neuron PID controltwo control strategies compared and concluded a final conclusion, that thesingle neuron PID control on the control performance is much better thantraditional PID control, more intelligent. That the proposed control strategy inline with the initial design purpose, it provides a new thinking of hydraulicsynchronization control.更多还原