Reworking automobile coating reordering method based on deep reinforcement learning

Abstract

The invention belongs to the field of resource, workflow, personnel or project management, and relates to a reworking automobile coating reordering method based on deep reinforcement learning, which can respond to the reworking condition in a coating workshop in real time and dynamically adjust a subsequent spraying plan. A reordering scheduling algorithm comprises a coating interaction environment construction module, an Actor-Critic algorithm implementation module, an offline training module and an online ordering module. The coating interaction environment construction module comprises order data import, parameter setting and coating interaction environment initialization. The algorithm implementation module comprises mathematical model construction, state space definition, action space definition, reward function definition and algorithm structure design, and the module is a core module of the invention. The offline training module is used for training order data according to the realized algorithm to obtain a reordering scheduling model. The online ordering module can perform real-time online ordering according to a model obtained after offline training of the order set.

Description

technical field [0001] The invention relates to a method for reordering automobile painting with rework based on deep reinforcement learning, which belongs to the field of resource, workflow, personnel or project management. Background technique [0002] In the modern automobile manufacturing system, after the car is processed from the body shop, it enters the WBS (White Body Storage, white body buffer zone) buffer zone in the order of the assembly demand sequence and waits to be processed in the paint shop. In order to reduce the operating cost of color switching, the paint shop needs to Reorder the current order sequence. At the same time, in order to reduce the production delay of the downstream assembly workshop, ensure that the order can comply with the planned assembly sequence and deliver on time, it is also necessary to reduce the deviation from the assembly demand sequence when reordering. However, there are still rework and repairs caused by serious quality problem...

AI Technical Summary

Problems solved by technology

[0005] The published reordering methods for automobile painting workshops have proposed some solutions to the production needs of the painting workshops, but there are still the following deficiencies: First, only the reduction of the number of painting workshops is considered when rescheduling the automobile painting. The number of color switching times or only solve the sequence recovery problem of the downstream buffer zone, without considering the demand linkage between the upstream painting shop and the downstream assembly shop, when making the spraying plan for the order collection in the WBS buffer zone, the reduction of painting is not considered at the same time The number of color switches in the sequence and the deviation from the final assembly demand sequence; second, the current reordering scheduling methods are mostly based on specified rules and traditional heuristic algorithms, which fail to abstract the dynamic changes of the painting environment and sequence well Third, the currently published results fail to propose a solution to the rework caused by serious quality defects in the paint shop and the secondary spraying of the body after rework. The effect will have an impact, which will directly increase the deviation of the final assembly demand sequence, resulting in production delays. It is necessary to dynamically adjust the painting plan according to the rework situation and the need for secondary spraying.

[0006] Based on the actual painting requirements, the present invention proposes a more complete and effective reordering scheduling method considering the dynamic rework situation. While frequent color switching in the painting workshop, reduce the inconsistency with the downstream assembly demand sequence, reduce the waste of painting paint, reduce production efficiency, and delay delivery

Images