APS dynamic production plan scheduling algorithm

Abstract

The invention relates to an APS dynamic production plan scheduling algorithm, which comprises the following steps of: generating a static scheduling plan by using a GA (Genetic Algorithm) on the premise of not considering dynamic characteristics; when the dynamic characteristics appear, optimizing the dynamic characteristics in actual production by using a reinforcement learning method; carrying uncertainty probability distribution through a neural network, and guiding and adjusting the scheduling result of the next GA by taking the probability distribution as feedforward prediction; therefore, positive feedback closed loop from individual intelligence emerging to group intelligence and from group intelligence to evolutionary individual intelligence is realized until intelligence convergence. According to the method, the probability distribution of various uncertainties is learned through the neural network, and various uncertainties are dynamically dealt with through an online learning mechanism provided by reinforcement learning, so that the FJSSP problem is solved through the combination of GA, the neural network and reinforcement learning, namely the crowd-sourcing evolutionary algorithm, and a complete solution is provided for intelligent manufacturing upgrading of a factory.

Description

technical field [0001] The invention relates to the technical field of production planning and scheduling, in particular to an APS dynamic production planning and scheduling algorithm. Background technique [0002] Production planning and scheduling APS (Advanced Planning and Scheduling) problems can be generally divided into three types: discrete (JSSP), process-based (FSSP) and open (OSSP). Among them, process-based scheduling problems can be regarded as discrete-type scheduling problems. A special case of the process, which abstracts an entire automated production line in the process-based scheduling problem into a single process. In the actual production process, most of the problems are the mixed type of JSSP and FSSP. In addition, in actual production activities, a process can usually be executed on a set of replaceable machines, rather than intelligently executed on a special machine. Therefore, based on the serial structure of the original JSSP (or FSSP, OSSP) On th...

AI Technical Summary

Problems solved by technology

[0007] However, there are many uncertainties in actual production scenarios, such as fluctuations in manufacturing time, machine failures, preventive maintenance, dynamic order insertion, etc. Therefore, it can be said that the actual production scheduling problem is a dynamic FJSSP problem (Dynamic / Fuzzy FJSSP )

However, the existing GA algorithm has been unable to solve the problem of such uncertainty.

Images