Causal Structure Learning for Observable Variable Fluctuation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing substrate processing technologies lack the capability to derive causal structures between observable variables, leading to inefficiencies in process control and quality prediction.
Innovation Solution
A non-transitory computer-readable recording medium and information processing apparatus that acquire observation data, derive causal relationships between observable variables, and generate prediction models to identify causes of fluctuations, using algorithms like LiNGAM and interactive user modifications to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional substrate processing control methods are used, then process control is maintained, but causal structure derivation between observable variables is lacking, leading to inefficiencies in quality prediction and process optimization
Solution Approach 1:
The patent replaces traditional mechanical/statistical process control methods with an information processing system that uses causal structure learning algorithms (such as LiNGAM - Linear Non-Gaussian Acyclic Model) to automatically derive causal relationships from observation data. This substitution enables the system to identify true causal structures rather than merely correlational patterns, significantly improving quality prediction accuracy while the automated nature of the algorithm keeps the implementation complexity manageable
Solution Approach 2:
The patent transforms the approach to process control by changing from parameter-based statistical monitoring to causal structure-based analysis. By learning the causal structure from data and representing it as a directed acyclic graph (DAG), the system can trace causal pathways from process parameters to quality outcomes, enabling more precise prediction and control while adapting to different processing conditions through data-driven structure identification
2Reliability
If comprehensive observation data is collected from all sensors, then complete process monitoring is achieved, but identifying true causal relationships becomes difficult due to complex correlations
Solution Approach 1:
The patent employs causal structure learning algorithms that go beyond traditional statistical correlation methods. These algorithms (such as LiNGAM) can distinguish true causal relationships from spurious correlations by leveraging non-Gaussian properties of the data and acyclic constraints, thereby reliably identifying causal structures even in the presence of complex inter-variable correlations and comprehensive sensor data
Solution Approach 2:
The patent extracts the essential causal structure from complex observation data by representing relationships as a directed acyclic graph (DAG). This extraction process separates true causal signals from noise and spurious correlations, identifying the minimal set of causal relationships that govern the system behavior, thereby preventing information loss while simplifying the complex data into actionable causal knowledge
3Productivity
If manual analysis of process data is performed, then detailed inspection is possible, but processing time and labor requirements increase significantly
Solution Approach 1:
The patent implements a self-service system where the information processing apparatus automatically acquires observation data from sensors, performs causal structure learning, and generates causal graphs without requiring manual intervention. The system serves itself by autonomously completing the entire causal analysis workflow, from data collection to causal relationship identification, thereby dramatically increasing productivity and eliminating time losses associated with manual data analysis
Solution Approach 2:
The patent replaces manual data analysis with an automated information processing system that uses causal structure learning algorithms. This substitution enables rapid processing of comprehensive sensor data, automatically identifying causal relationships in seconds or minutes rather than hours or days of manual inspection, thereby significantly improving analysis speed while maintaining or enhancing accuracy
Data Source
AI summary
A non-transitory computer readable recording medium storing a computer program causing a computer to execute processing of acquiring observation data corresponding to a plurality of types of observable variables from an observation system to be monitored, deriving causal relationships between the observable variables based on the acquired observation data, extracting one or more other observable variables, which are candidates for a cause of a fluctuation in one observable variable, based on the derived causal structure, and outputting a extracted result.