Time-Delay Autoencoder for Factory Anomaly Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Factory automation systems face challenges in anomaly detection and diagnosis due to data complexity, requiring loss function adaptation, temporal information exploration, and data cleaning to improve production efficiency and quickly restore systems to normal conditions.
Innovation Solution
A pre-processed time-delay autoencoder neural network with adaptive loss function selection and data filtering is employed for anomaly detection and diagnosis, utilizing multiple loss functions for different data types and incorporating temporal data to identify anomalies caused by series of operations, while cleaning data to reduce noise and improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional autoencoder neural network is used for anomaly detection in FA systems, then the system can detect anomalies with basic functionality, but the detection accuracy is insufficient due to data complexity and lack of temporal information processing
Solution Approach 1:
The patent segments the anomaly detection task into multiple specialized components: a time-delay module for temporal information processing, a data pre-processing module for cleaning and normalization, and a multi-loss function module for handling different data types. Each component addresses specific limitations of conventional autoencoders, collectively improving detection accuracy without requiring a complete redesign of the neural network architecture.
Solution Approach 2:
The patent applies preliminary actions by implementing data pre-processing steps before the main anomaly detection process. This includes cleaning raw data to remove noise, normalizing data distributions, and extracting relevant features. By preparing the data in advance, the neural network receives high-quality input, which significantly improves detection accuracy without increasing the complexity of the network itself.
2Adaptability or versatility
If a single loss function is used in the autoencoder, then the implementation is simple, but the performance deteriorates when handling multiple data types with different characteristics
Solution Approach 1:
The patent applies local quality by assigning different loss functions to different data types based on their specific characteristics. For example, mean squared error loss is used for continuous numerical data, while cross-entropy loss is used for categorical data. This localized adaptation allows each data type to be processed with the most appropriate loss function, improving overall model adaptability without requiring a completely complex unified loss function.
Solution Approach 2:
The patent implements a universal loss function framework that can handle multiple data types through a single multi-loss function module. This module automatically selects and applies appropriate loss functions for different data types (continuous, categorical, binary), making the autoencoder versatile across various FA system data types while maintaining a unified architectural structure.
3Measurement precision
If temporal information is not considered in anomaly detection, then the processing is simpler and faster, but the ability to detect anomalies caused by series of operations is lost
Solution Approach 1:
The patent introduces a temporal dimension to the anomaly detection process by implementing a time-delay module that processes sequences of data points over time. This module captures temporal patterns and relationships between consecutive operations, enabling the detection of anomalies caused by series of operations. The time-delay mechanism adds temporal context without significantly increasing computational complexity, as it uses simple delay registers and sequence processing.
Data Source
AI summary
A computer-implemented pre-processed time-delay autoencoder based anomaly detection method are provided for detecting anomalous states of machines arranged in a factory automation (FA) system or a manufacturing production line. The method includes acquiring source signals from the machines via an interface performing a data pre-processing process for the acquired source signals by normalizing value ranges of the acquired source signals and filtering undesired features from the acquired source signals performing a time-delayed data reform process for the pre-processed source signals based on a time-delay window to generate pre-processed time-delay data submitting pre-processed time-delay testing data to a pre-processed time-delayed autoencoder (Prep-TDAE) neural network, wherein the pre-processed time-delay testing data are collected online while the machines are operated, wherein the Prep-TDAE neural network has been pre-trained by using the pre-processed time-delay training data detecting, if an anomaly state is encountered with respect to the machines, by computing anomaly scores of the pre-processed time-delay testing data, and determining, when the anomaly state is detected, anomaly occurrence time, duration and severity with respect to the anomaly state of each of the machines.