Industrial Machine Time-Series Modeling for Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for utilizing data from industrial machines, such as substrate processing apparatuses, are inefficient in predicting and detecting abnormalities, limiting effective control and monitoring.
Innovation Solution
An information processing method using dynamic mode decomposition to calculate parameters for transforming observed data into time evolution data, incorporating fractional differential equations to enhance prediction accuracy and detect abnormalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional data utilization methods are used for industrial machines, then the system structure remains simple, but the prediction accuracy and abnormality detection capability are insufficient
Solution Approach 1:
The patent transforms the raw observed data into dynamic mode decomposition parameters (spatial modes and temporal coefficients) that capture the essential characteristics of the system behavior. This parameter transformation enables accurate prediction and anomaly detection by focusing on the dominant modes of variation rather than processing all raw data points directly
Solution Approach 2:
The patent decomposes the complex time-series data into multiple dynamic modes, each representing a distinct pattern of behavior. By segmenting the data into separate spatial and temporal components, the system can analyze and predict each mode independently, improving overall prediction accuracy while maintaining computational efficiency
2Loss of information
If more data is collected from industrial machines, then the information available for analysis increases, but the computational burden and processing time increase
Solution Approach 1:
The patent extracts only the essential information from the observed data by computing dynamic mode decomposition parameters. Instead of processing all raw data, the system extracts the dominant spatial modes and temporal coefficients that contain the most significant information for prediction and anomaly detection, thereby reducing computational burden while maintaining information quality
3Reliability
If complex models are used to predict time evolution data, then the prediction accuracy improves, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent employs dynamic mode decomposition, which inherently adapts to the temporal dynamics of the observed data. The method automatically identifies and tracks the evolving modes and frequencies in the data, allowing the model to capture non-stationary behavior without requiring complex fixed-structure models. This dynamic approach improves prediction reliability while keeping computational complexity manageable
Data Source
AI summary
Provided are an information processing method, a computer program, and an information processing apparatus that can be expected to effectively utilize data obtained from a target apparatus. An information processing method according to the present embodiment is performed by an information processing apparatus and includes acquiring time series observed data regarding a target apparatus, and calculating a first parameter and a second parameter of a model that predicts time evolution data of the observed data based on the observed data, by dynamic mode decomposition based on the acquired observed data, in which the first parameter is a parameter relating to a transformation function for transforming the observed data into the time evolution data, and the second parameter is a parameter relating to a function of the observed data describing the time evolution data.