Graph Autoencoder Anomaly Detection for PLC Control Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to efficiently detect and track anomalies in automation equipment operations, leading to process interruptions, production delays, and decreased operation rates.
Innovation Solution
A method is developed to generate graph data for training an anomaly detecting model, which involves classifying contact point value changes, identifying major states, and converting log data into node and edge index data using a graph neural network (GNN) AutoEncoder.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection methods are used in automation equipment, then the system structure remains simple, but the detection precision and reliability are insufficient leading to missed anomalies
Solution Approach 1:
The patent introduces a graph neural network AutoEncoder as an intermediary component between the PLC control logic and the anomaly detection system. This intermediary transforms raw control logic data into graph representations with nodes and edges, enabling sophisticated anomaly detection without requiring complete system redesign. The GNN AutoEncoder serves as a mediator that bridges simple data collection and complex anomaly pattern recognition.
Solution Approach 2:
The patent replaces traditional mechanical or rule-based anomaly detection mechanisms with a data-driven graph neural network approach. Instead of using predefined thresholds or simple statistical methods, the system employs deep learning models (GNN AutoEncoder) to automatically learn anomaly patterns from historical control logic data, substituting conventional detection mechanics with intelligent algorithmic detection.
2Reliability
If comprehensive monitoring of all contact points is implemented, then the reliability of anomaly detection improves, but the loss of time for data processing and the complexity of the system increase
Solution Approach 1:
The patent extracts only the essential and relevant features from the comprehensive PLC control logic data. The graph neural network identifies and extracts key contact points, states, and transitions that are most indicative of anomalies, rather than processing every single data point. This extraction approach maintains detection reliability by focusing on critical features while significantly reducing processing time through dimensionality reduction.
Solution Approach 2:
The patent segments the complex control logic into discrete graph components (nodes representing contact points and edges representing transitions). By dividing the monitoring task into segmented graph elements, the system can process and analyze specific segments independently and in parallel, reducing overall processing time while maintaining comprehensive coverage through the interconnected graph structure.
3Measurement precision
If the graph neural network AutoEncoder model is trained with extensive graph data, then the anomaly detection precision improves, but the loss of time for model training and the computational resources required increase
Solution Approach 1:
The patent performs preliminary actions by pre-processing control logic data into graph representations before model training. The graph construction, including node and edge definition, is prepared in advance, allowing the GNN AutoEncoder to train more efficiently on structured data. This preliminary graph preparation reduces the computational burden during training while maintaining the precision benefits of comprehensive data analysis.
4Reliability
If continuous monitoring of operation patterns is implemented, then the reliability of detecting process interruptions improves, but the use of energy and computational resources increases
Solution Approach 1:
The patent implements continuous monitoring through the GNN AutoEncoder model that continuously processes incoming control logic data streams. The model maintains a learned understanding of normal operation patterns and continuously compares real-time data against this understanding, enabling uninterrupted detection of process interruptions. This continuous action maintains high reliability while the efficient graph-based architecture keeps energy consumption manageable.
Data Source
AI summary
A method, of training an anomaly detecting model using a plurality of pieces of graph data, includes: (a) inputting one piece of graph data that has not yet been input, among the plurality of pieces of graph data, to a graph neural network (GNN) AutoEncoder calculating a probability of each edge as input data; (b) calculating a difference value (hereinafter, “edge difference value”) between an edge probability value of reconstructed data output by the GNN AutoEncoder and an edge value of the input data; (c) calculating an average value (hereinafter, “positive edge loss”) of a positive edge and an average value (hereinafter, “negative edge loss”) of a negative edge using the edge difference value, and calculating an edge prediction loss value of the reconstructed data by summing the positive edge loss and the negative edge loss; (d) retraining the GNN AutoEncoder until the edge prediction loss value is minimized.