不确定条件下混装和作业车间调度问题研究

发布时间：2019-05-07 02:30

【摘要】：在现代制造模式中多品种、小批量生产愈来愈多，对产品成本和质量的要求越来越高，因此对车间运作管理也提出了标准化、精细化的要求，致使管理者愈发关注生产中存在的不确定性及其对生产的影响。在实际工作中信息的获得具有不及时和不完整的特点。生产调度需及时了解、充分考虑这些影响因素，在调度方案制定前需防范此因素对生产造成的不平衡隐患，在执行过程中调度方案需随时动态调整以适应这些变化。在总结以往工作的基础上，本文研究混装和作业车间调度时处理不确定的框架、机制和措施，提出在不确定条件下的鲁棒调度方法和动态自适应反应式策略，并对生产过程中的不确定信息处理和参数校正方法进行探讨。本文的主要工作如下：以系统性消除不确定因素的影响为目标，构造了结合预防式调度、反应式调度与不确定推理于一体的整体调度框架。以具有不确定吸收能力的鲁棒调度方案作为生产开始前的预调度方案，基于调度结果通过贝叶斯推理对预估的不确定参数分布进行再处理和修正；利用具有不确定反馈能力的反应式调度，应对生产中各种突发事件，评估并修正应对策略，，为下一阶段的预防式和反应式调度提供更可靠的决策依据。基于预防式调度思想，研究具有不确定吸收能力的鲁棒调度方法。针对具有不确定操作时间的混合装配线平衡问题，基于混合整型线性规划建立相应的鲁棒对等模型；针对具有不确定操作时间的作业车间调度问题，建立基于调度目标期望值的目标规划模型并开发相应的智能算法对模型进行求解。针对生产过程中出现的设备故障、订单改变等突发事件，通过调整系统参数中的设备和工件等，提出具有自适应能力的反应式调度方法。并针对柔性作业车间调度问题，开发具有双层编码的遗传算法。研究不确定信息的处理及不确定参数的校正方法。以随机变量的上界、下界、均值和方差为不确定参数描述手段，建立基于随机变量的鲁棒解与基于均值的确定解之间的对应关系。以贝叶斯网络为工具，结合后验信息与先验统计进行分布参数的校正处理，以获得更符合实际情况的分布参数。为降低调度问题的计算复杂性，研究两种快速算法——针对装配线平衡问题的摹加代数方法和针对作业车间调度的Hopfield-神经网络算法。对于前者，通过数学命题证明在摹加代数意义上，简单装配线平衡问题可等价于旅行商问题；对于后者，基于Lyapunov稳定性理论证明方法的收敛性。并通过实际算例验证两种方法的有效性。
[Abstract]:In the modern manufacturing mode, there are more and more varieties, more and more small batch production, higher and higher cost and quality of the product, so the management of the workshop operation is also put forward the standardization, the refined request. Managers pay more attention to the uncertainty in production and its influence on production. In practical work, the acquisition of information is not timely and incomplete. Production scheduling needs to be understood in time, fully considering these factors, and the hidden danger of imbalance caused by this factor should be prevented before the scheduling plan is formulated. The scheduling scheme should be dynamically adjusted to adapt to these changes in the process of execution. On the basis of summarizing previous work, this paper studies the framework, mechanism and measures for dealing with uncertainty in mixed-loading and job-shop scheduling, and proposes robust scheduling method and dynamic adaptive reactive strategy under uncertain conditions. The methods of uncertain information processing and parameter correction in production process are also discussed. The main work of this paper is as follows: aiming at systematically eliminating the influence of uncertain factors, a whole scheduling framework is constructed, which combines preventive scheduling, reactive scheduling and uncertain reasoning. The robust scheduling scheme with uncertain absorption ability is used as the pre-scheduling scheme before the production start. Based on the scheduling results, the estimated uncertain parameter distribution is re-processed and modified by Bayesian reasoning. The reactive scheduling with uncertain feedback ability is used to deal with all kinds of emergency events in production and to evaluate and revise the response strategies so as to provide more reliable decision-making basis for preventive and reactive scheduling in the next stage. Based on the idea of preventive scheduling, a robust scheduling method with uncertain absorbing ability is studied. To solve the problem of hybrid assembly line balance with uncertain operating time, a robust equivalent model based on hybrid integral linear programming is established. In order to solve the job shop scheduling problem with uncertain operation time, the objective programming model based on scheduling target expectation is established and the corresponding intelligent algorithm is developed to solve the model. Aiming at the emergency events such as equipment failure and order change in the production process, a reactive scheduling method with adaptive ability is proposed by adjusting the equipment and workpieces in the system parameters. In order to solve the flexible job shop scheduling problem, a two-level coded genetic algorithm is developed. The processing of uncertain information and the correction method of uncertain parameters are studied. In this paper, the upper bound, lower bound, mean value and variance of random variables are used to describe the uncertain parameters, and the corresponding relation between the robust solutions based on random variables and the definite solutions based on mean is established. The Bayesian network is used as a tool, and the posterior information and prior statistics are used to correct the distribution parameters in order to obtain the distribution parameters which are more in line with the actual situation. In order to reduce the computational complexity of scheduling problem, two fast algorithms are studied-the modeling algebra method for assembly line balance problem and the Hopfield- neural network algorithm for job shop scheduling. For the former, it is proved by mathematical propositions that the simple assembly line equilibrium problem can be equivalent to the traveling salesman problem in the sense of mimetic algebra, and for the latter, the convergence of the method is proved based on the Lyapunov stability theory. The effectiveness of the two methods is verified by a practical example.
【学位授予单位】：武汉科技大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TH186

【参考文献】