Aerospace detonator production scheduling method based on deep reinforcement learning

Abstract

The invention discloses an aerospace initiator production scheduling method based on deep reinforcement learning, and mainly relates to the field of machine learning and intelligent manufacturing. The method comprises the following steps: S1, acquiring real-time information of production and processing from an exploder production workshop; S2, according to real-time information collected by an exploder production workshop, determining an exploder flexible production workshop scheduling problem description and related hypotheses; S3, determining an objective function and a constraint condition for scheduling optimization of the detonator production workshop; S4, constructing an exploder production scheduling problem into a Markov decision model, and converting real-time information into a real-time state; S5, storing the real-time state information in a memory bank and serving as input of deep reinforcement learning DQN algorithm training; S6, training a deep reinforcement learning DQN algorithm; and S7, real-time scheduling of production of the exploder. According to the method, the adaptivity and the real-time performance of production scheduling can be improved, and the scheduling scheme can better adapt to the complex and dynamic actual production process.

Description

technical field [0001] The invention relates to the field of machine learning and intelligent manufacturing, in particular to a production scheduling method for aerospace detonators based on deep reinforcement learning. Background technique [0002] With the rapid development of the aerospace industry, the demand for aerospace explosives continues to increase, and the traditional production mode based on manual production is far from meeting the development needs. At present, the discrete manufacturing enterprises producing aerospace explosives are in the transformation period of automated production, and are transforming from the traditional mass production mode to the small batch and multi-batch flexible production mode. Flexible production and processing scheduling is more flexible than traditional production scheduling and can better adapt to dynamic changes in the external environment. In complex and dynamic processing, scheduling schemes and parameters are usually org...

AI Technical Summary

Problems solved by technology

[0004] 1) Lost the advantage of simplicity and ease of scheduling rules, the scheduling rules generated by the algorithm are more complicated in form;

[0005] 2) Ignored the value of experience records, and did not pay attention to the historical data of the company's past production and processing scheduling;

[0006] 3) Lack of consideration of the learning ability of production control methods, which cannot meet the needs of the current intelligent manufacturing system construction

[0007] 4) Lack of consideration of the influence of the trial installation of the detonator and the impact of the time of curing and filling the pressure on the quality

[0008] 5) In the process of preparing materials, considering the influence of environmental factors such as medicaments and glue solutions, secondary scheduling is required

Images