Method, device, computer storage medium and terminal for fault diagnosis

Abstract

A fault diagnosis method, device, computer storage medium, and terminal, comprising: extracting equipment signal features from equipment operation-related data; marking a preset number of fault samples to obtain the category to which the fault samples belong; Add the category to the preset machine learning algorithm for training to obtain a fault diagnosis model; diagnose the fault according to the fault diagnosis model obtained through training to obtain diagnostic data; wherein, the data related to the operation of the equipment includes: fault-related Multi-dimensional time-series signal data; the diagnosis data includes: diagnosis results and diagnosis basis corresponding to the diagnosis results. The embodiments of the present invention improve the efficiency of fault diagnosis.

Description

technical field [0001] This article involves but is not limited to automation technology, especially a method, device, computer storage medium and terminal for fault diagnosis. Background technique [0002] During the operation of the equipment, all parts of the equipment need to be kept in a safe and stable working state for a long time. In order to ensure the stable operation of the equipment, one of the key links is to carry out fault diagnosis, that is, to quickly find the crux of the fault after the fault occurs, so as to quickly eliminate the fault; another important link is to record the equipment maintenance process, which is concise and complete. The records not only help in the later maintenance of the equipment, but also serve as a reference for experts. Take the train's traction control unit (TCU) as an example, which is a modular microprocessor control unit for train traction machines. As an important part of train control, TCU is used to control electric driv...

AI Technical Summary

Problems solved by technology

[0004] The expert model has the following problems: 1. It relies heavily on the experience data of experts; as the age of experts grows and the turnover rate of employees rises, the data required by the expert model becomes more and more unstable; 2. The data is scattered; the experience data mastered by different experts is generally only A certain type or several types of empirical data, scattered data, may lead to low efficiency of the expert model for fault detection; take TCU as an example, because the TCU module design is very sophisticated, each expert can only master the experience of certain types of faults 3. It is difficult to determine the parameter threshold and the rules are not perfect

Due to the variety of working conditions and the complexity of the fault mechanism, the parameter thresholds set by experts in the diagnosis process are often based on their own experience to set the diagnosis rules, which cannot be generalized to all cases under the category of similar faults; 4. Expert diagnosis There are contradictions; taking TCU as an example, different experts may make different diagnoses for the same fault based on their own experience with TCU; 5. The direction of fault diagnosis is unclear; empirical analysis based on mechanism has limitations, so in some fault cases There will be multiple labels on the fault category

The training model has the following disadvantages: it is not easy for maintenance personnel to understand, and it is not easy for iterative optimization; technicians cannot learn from the data stored in the training model

[0005] In summary, the diagnostic efficiency of the current fault diagnosis model needs to be improved

Images