CN-120430460-B - Industrial equipment fault prediction method and system based on deep learning

CN120430460BCN 120430460 BCN120430460 BCN 120430460BCN-120430460-B

Abstract

The invention discloses an industrial equipment fault prediction method and system based on deep learning, comprising the steps of obtaining historical operation data of industrial equipment to be detected in a set period, preprocessing the historical operation data to generate a sample data set, inputting the sample data set into an industrial equipment fault prediction model constructed based on a deep learning algorithm for learning and training, utilizing an adaptive optimizer to iteratively update model parameters and dynamically adjust model weights in the training process, deploying the trained fault prediction model into a local server, receiving operation data flow of the industrial equipment in real time in a model operation stage, performing fault probability prediction on the operation data flow, and outputting a corresponding fault type and occurrence probability thereof. The invention aims to solve the problems of low real-time prediction precision and poor model generalization capability of equipment operation data by combining an efficient data preprocessing technology and a self-adaptive optimization mechanism of a deep learning algorithm.

Inventors

JIN JING
ZHUANG XIAOBO
JIANG XIAOYUAN
Zhuang Bitai

Assignees

江苏太航信息科技有限公司

Dates

Publication Date: 20260508
Application Date: 20250425

Claims (6)

1. An industrial equipment fault prediction method based on deep learning is characterized by comprising the following steps: Acquiring historical operation data of industrial equipment to be detected in a set period of time, and preprocessing the historical operation data to generate a sample data set; Inputting the sample data set into an industrial equipment fault prediction model constructed based on a deep learning algorithm for learning training, and iteratively updating model parameters by using an adaptive optimizer in the training process and dynamically adjusting model weights; After model parameters are iteratively updated by using an optimizer, dynamically adjusting model weights, wherein the method comprises the steps of continuously calculating weighted precision indexes of the latest N prediction results and comparing the weighted precision indexes with a historical base line during the training period of the fault prediction model, triggering incremental online learning when detecting that performance degradation exceeds a degradation threshold value, inserting newly acquired tagged data into an experience playback buffer zone, executing fine adjustment updating of limited iteration to correct the model weights after determining data distribution drift based on a drift detection module, and recalculating the performance indexes after finishing updating, and replacing old model weights if the performance indexes are higher than the degradation threshold value, otherwise, rolling back to the historical optimal weights; deploying the trained fault prediction model into a local server, receiving an operation data stream of industrial equipment in real time at a model operation stage, performing fault probability prediction on the operation data stream, and outputting a corresponding fault type and occurrence probability thereof; Deploying the fault prediction model after training to a local server, wherein the fault prediction model comprises 8bit quantification and graph optimization for reducing reasoning delay, adopting a containerization mode to package model service and externally providing a reasoning interface through RESTful API; the method comprises the steps of receiving an operation data stream of industrial equipment in real time in a model operation stage, performing fault probability prediction on the operation data stream, outputting a corresponding fault type and occurrence probability thereof, extracting data fragments from continuously acquired original operation data stream by utilizing a fixed sliding window strategy, performing normalization processing on the data fragments to obtain a normalized tensor, inputting the normalized tensor into an inference interface deployed on a local server, calling the fault prediction model to perform vectorization inference, outputting a multidimensional probability vector, selecting a candidate fault type corresponding to a probability maximum value from the multidimensional probability vector to form a prediction result pair, returning the prediction result pair if the probability maximum value is more than or equal to a dynamic threshold value, and issuing the prediction result pair in an OPCUA event channel in a key-value mode; the candidate fault types at least comprise bearing wear, misalignment, rotor imbalance, insufficient lubrication, coil short circuit, overload operation, abnormal temperature rise and power supply fluctuation.
2. The deep learning based industrial equipment fault prediction method according to claim 1, wherein the historical operation data at least comprises a historical vibration signal, a current voltage signal, a bearing temperature signal, a rotation speed signal, a lubricating oil pressure signal and an equipment operation state of industrial equipment to be tested.
3. The deep learning based industrial equipment fault prediction method of claim 2, wherein preprocessing the historical operating data to generate the sample data set comprises: Removing random noise by adopting a wavelet threshold denoising method; filling the missing data through spline interpolation and removing the abnormal points by 3 sigma criterion; Normalizing each physical quantity according to the dimension; Segmenting the normalized data according to a fixed sliding window strategy, marking corresponding fault type labels, and summarizing the fault type labels in a set to obtain the sample data set.
4. The deep learning-based industrial equipment fault prediction method of claim 3, wherein inputting the sample dataset into an industrial equipment fault prediction model constructed based on a deep learning algorithm for learning training comprises: the fault prediction model is of a serial structure of a one-dimensional convolutional neural network two-way long-short-term memory network attention mechanism; Performing iterative training by using a cross entropy loss function and using a AdamW optimizer; Setting a Dropout random inactivation rate of 0.3 in the training process to inhibit overfitting, and adopting an early-stopping strategy to monitor the loss of the verification set; after the optimal super-parameter combination is obtained through five-fold cross validation, the model weight with the highest F1 score in the validation set is stored.
5. An industrial equipment fault prediction system based on deep learning, comprising: One or more processors; a memory storing instructions operable, when executed by the one or more processors, to cause the one or more processors to perform operations comprising the flow of the deep learning-based industrial equipment fault prediction method of any one of claims 1-4.
6. A computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising the flow of the deep learning-based industrial equipment fault prediction method of any one of claims 1-4.

Description

Industrial equipment fault prediction method and system based on deep learning Technical Field The invention relates to the technical field of industrial equipment fault prediction, in particular to an industrial equipment fault prediction method and system based on deep learning. Background The traditional prediction method adopts a traditional statistical analysis method, depends on physical parameters or expert knowledge of equipment, has poor performance on high-dimensional, strong nonlinear or unsteady data, and gradually becomes a research hot spot by using the recently developed artificial intelligence technology, particularly a fault prediction method based on machine learning and deep learning, and has the advantages of strong data driving and generalization capability and better capturing capability on complex nonlinear relations. The industrial equipment fault prediction method in the prior art has the problems of insufficient generalization, poor real-time performance and weak self-adaptation capability of the prediction model, especially for complex industrial operation data with multiple dimensions and parameters, the prediction accuracy is difficult to effectively improve, and the requirements of quick response and high reliability in actual industrial production cannot be met. CN116089870A discloses an industrial equipment fault prediction method based on meta-learning under a small sample condition, the technology utilizes meta-learning to realize cross-domain knowledge migration, so that the demand of a model on target domain data is reduced, however, the technical scheme excessively depends on the similarity of source domain and target domain data distribution, and when the difference of equipment operation working conditions is large, the generalization performance of the model is greatly reduced, and the robustness is poor. CN117349619a discloses an industrial equipment fault prediction method for constructing input features by utilizing Copula function, the method determines the input features by Spearman rank correlation and analytic hierarchy process to improve fault prediction accuracy, but the method still belongs to traditional statistical analysis framework, the extraction and selection of features are too dependent on empirical functions, and are difficult to adapt to dynamically changing equipment operation conditions, large-scale and real-time monitoring data cannot be processed efficiently, and certain limitation exists in practical industrial environment. In summary, the existing industrial equipment fault prediction technology generally has the problems of poor data adaptability, poor generalization performance and insufficient real-time prediction accuracy, and the industrial equipment fault prediction method based on deep learning provided by the invention focuses on solving the problems of low real-time prediction accuracy and poor model generalization capability of equipment operation data by combining an efficient data preprocessing technology and an adaptive optimization mechanism of a deep learning algorithm. Disclosure of Invention This section is intended to summarize some aspects of embodiments of the application and to briefly introduce some preferred embodiments, which may be simplified or omitted in this section, as well as the description abstract and the title of the application, to avoid obscuring the objects of this section, description abstract and the title of the application, which is not intended to limit the scope of this application. The present invention has been made in view of the above-described problems occurring in the prior art. In order to solve the technical problems, the invention provides the following technical proposal that the historical operation data of the industrial equipment to be detected in a set period of time is obtained and preprocessed to generate a sample data set; Inputting the sample data set into an industrial equipment fault prediction model constructed based on a deep learning algorithm for learning training, and iteratively updating model parameters by using an adaptive optimizer in the training process and dynamically adjusting model weights; And deploying the trained fault prediction model into a local server, receiving the operation data stream of the industrial equipment in real time at the model operation stage, performing fault probability prediction on the operation data stream, and outputting the corresponding fault type and occurrence probability thereof. As a preferable scheme of the deep learning-based industrial equipment fault prediction method, the historical operation data at least comprises a historical vibration signal, a current voltage signal, a bearing temperature signal, a rotating speed signal, a lubricating oil pressure signal and an equipment operation state of industrial equipment to be detected. As a preferable scheme of the deep learning-based industrial equipment fault prediction method of th