Surrogate Model Explanations for Industrial Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models used in industrial anomaly detection are complex and lack transparency, making it difficult to understand their global and local behavior, which hinders root-cause analysis and user trust in their outputs.
Innovation Solution
A method that generates synthetic sensor data to train a surrogate model, which is intrinsically interpretable, providing global and local explanations of the machine learning model's behavior, and calculates agreement accuracy to ensure similarity with the original model, allowing for improved decision-making and optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complex machine learning model is used for anomaly detection based on multiple sensor parameters, then the detection accuracy is improved, but the model transparency and interpretability deteriorate
Solution Approach 1:
The patent introduces an intermediary explanation system that mediates between the complex black-box ML model and the user. This explanation system generates human-understandable descriptions of model predictions, including the reasoning process and key factors influencing the anomaly detection, thereby recovering interpretability without changing the underlying complex model architecture.
Solution Approach 2:
The patent segments the model interpretation process into multiple levels: global explanations that describe overall model behavior and local explanations that detail specific prediction reasoning. This segmentation allows users to access different levels of detail depending on their needs, maintaining detection accuracy while providing structured interpretability.
2Productivity
If a black-box machine learning model is deployed for industrial monitoring, then the productivity and automation are improved, but the user trust and acceptance deteriorate due to lack of explainability
Solution Approach 1:
The patent implements feedback mechanisms where the explanation system continuously provides rationale for model predictions to users. This feedback loop allows users to understand model decisions, validate them against domain knowledge, and build trust while maintaining the automated monitoring workflow and productivity benefits.
3Device complexity
If standard anomaly detection models are used without explanation capabilities, then the device complexity is reduced, but the ability to perform root-cause analysis deteriorates
Solution Approach 1:
The patent introduces an explanation intermediary layer that sits on top of standard anomaly detection models. This intermediary generates interpretable outputs including key contributing factors and reasoning paths, enabling root-cause analysis without modifying the underlying simple detection model architecture.
Data Source
AI summary
A computer-implemented method for providing information concerning a global behavior of a machine learning model trained with measured sensor data representing technical parameters of a technical system and used to evaluate the technical system, including, receiving the machine learning model and measured sensor data generating a number of synthetic sensor data by a synthetic data generator, predicting labels for the synthetic sensor data and the measured sensor data by the result of the machine learning model when processing the synthetic sensor data and the measured sensor data as input data, training a surrogate model based on the synthetic sensor data and measured sensor data and the predicted labels, calculating an agreement accuracy indicating the similarity of a result of the surrogate model compared to a result of the machine learning model, outputting to a user interface the trained surrogate model and the agreement accuracy.