A method for abnormal detection of time series data of low pressure casting machine based on bidirectional lstm

Abstract

The present invention provides a method for detecting abnormality of low-pressure casting machine time series data based on bidirectional LSTM, which comprises the following steps: retrieving the original pressure data of the low-pressure casting machine in a normal state from a database; , and convert the data into the data format required for supervised learning; divide the obtained pressure data into training set and test set data; use the obtained training set data to train, build and save a bidirectional LSTM neural network model; according to the expected output and predicted output of the test set Calculate the error, and use the corrected error as the bidirectional LSTM neural network to predict the output error; retrieve the original pressure data to be measured, and input the processed data into the trained bidirectional LSTM neural network for prediction. The patent prediction part of the present invention uses bidirectional LSTM, which can obtain a more accurate prediction effect than the unidirectional LSTM neural network.

Description

technical field [0001] The invention belongs to the field of abnormal pressure data detection of low pressure casting machines, in particular to a method for abnormal detection of time series data of low pressure casting machines based on bidirectional LSTM. Background technique [0002] Anomaly detection is the problem of finding data patterns that do not meet expected behavior. The development of computer-related technologies has led to the improvement of industrial technology, and anomaly detection has been applied in more and more fields, such as system health monitoring, industrial fault diagnosis and intrusion detection. Wait. There are two main types of common exceptions. The first exception type is called outliers, which are values ​​outside the normal value. Another type of anomaly is anomalous behavior, which is a cyclical collapse phenomenon in a time series that is anomalous in terms of periodicity even if the anomalous behavior reaches a normal value. [0003]...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to propose a method for abnormal detection of low-pressure casting machine time series data based on a bidirectional long-short-term memory neural network for the problem of using historical pressure data of low-pressure casting machines to guide production and the need to eliminate abnormal parts

Images