Fault tracing method in lithium ion battery production process

Abstract

The invention discloses a fault tracing method in a lithium ion battery production process. The method comprises the following steps of: firstly, acquiring a normal state sample set and a fault statesample set in the lithium ion battery production process,and performing data binarization operation; constructing a binary feature combination optimization model for distinguishing normal state data from fault state data; calculating a binary feature group set enabling the model objective function to reach an extreme value, and reversely converting the binary feature group set into a physical-quantity-based production rule set; and performing fault tracing analysis according to the production rule set. The fault tracing method does not depend on a system mechanism and priori knowledge in the lithium ion battery production process, innovatively constructs the binary feature combination optimization model from the perspective of data, and realizes parallel discovery of multiple fault judgment rules to realize fault tracing analysis.

Description

technical field [0001] The invention relates to the field of fault diagnosis in industrial processes, in particular to a fault tracing method in the production process of lithium-ion batteries. Background technique [0002] In the production process of lithium-ion batteries, the level of automation is high, the degree of integration and complexity is increasing day by day, and the correlation and coupling between different process variables, which makes any small problems such as human misoperation and abnormal equipment parts can trigger a chain reaction, resulting in The operation failure of the whole system will cause huge economic losses. Fault detection and diagnosis technology is one of the key technologies to ensure the safety of industrial production and reduce maintenance costs. It is mainly divided into qualitative and quantitative methods. Qualitative methods analyze and diagnose faults based on the experience and knowledge accumulated by domain experts or profes...

AI Technical Summary

Problems solved by technology

[0003] However, a common problem with commonly used diagnostic models based on data is the interpretability of complex models.

Taking the neural network as an example, the feature transformation performed internally is an incomprehensible black box for users, from which the cause and effect of faults and influencing variables cannot be obtained, and the stability of its diagnostic performance is also difficult to guarantee

Images