Research On Preemptive Scheduling Of Design Tasks With Random Rework

Customized equipment enterprises are a kind of order-oriented design and manufacturing enterprises,which are based on customer demand-oriented product design.The product design characteristics of these enterprises are low repeatability,complex design process and long design cycle.Due to the unclear product requirements and high quality requirements of product design,designers need to submit to the customer for confirmation or to the team leader/supervisor for review after each design task is completed.The results of customer confirmation and team leader/supervisor review may be passed or multiple rework of the design may be required.In the task review phase,designers often start another design task,and when rework occurs,they may stop the current task to prioritize the rework task.It can be seen that the design task process is complex and the completion time is highly uncertain,which easily causes the design project to be delayed.Product design completion time directly affects subsequent material procurement and spare parts production,which is very important for customized equipment production management.For this reason,considering that the design task has stochastic rework and preemption,this paper studies the design task dynamic scheduling strategy.The specific research contents are as follows:(1)The characteristics,scheduling process and shortcomings of design task scheduling method in engineering practice of customized equipment enterprises are analyzed.(2)To solve the problem of design task scheduling with stochastic rework and preemption,a scheduling model and method based on Markov decision process is proposed.Firstly,a scheduling decision model is constructed based on Markov decision process theory,and the optimal strategy is solved by value iteration algorithm.Then,aiming at the dimension disaster problem,a sub-optimal strategy based on single priority rule and multi-rule combination is proposed.(3)To solve the problem of design task scheduling with stochastic rework and preemption,a scheduling method based on supervised machine learning is proposed.Firstly,the potential factors affecting the decision-making of task scheduling are analyzed,and the macro-characteristics of different scheduling environments are constructed.Then,the training samples are obtained by using the above multi-rule combination algorithm to solve a smallscale example.Secondly,four different machine learning models are used to train and learn,and the performance of different models is compared.Finally,the approximation strategy obtained by machine learning is compared with the single priority rule through simulation experiments,which verifies the reliability of the machine learning model.(4)On the basis of the research on design task scheduling,and combined with the actual investigation in enterprises,a design task scheduling management system for customized equipment enterprises is developed.The machine learning strategy obtained in this paper is applied to the system to achieve efficient scheduling of design tasks.Finally,the design of the system is introduced in detail,and some functional interfaces are shown.