Causal Network Learning for Real-Time Industrial Root Cause Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection and diagnosis systems in industrial processes struggle with real-time root cause identification due to the complexity of nonlinear and non-stationary data, with knowledge-based methods requiring significant initial effort and data-driven methods failing to capture nonlinear relationships and non-stationarity.
Innovation Solution
A deep learning-based causal network learning technique that processes multivariate time-series data using pre-processing, deep neural networks, and stochastic proximal gradient descent to identify root cause variables and fault propagation paths in industrial processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If knowledge-based methods are used for fault diagnosis, then prior process knowledge can be utilized, but significant initial effort is required and knowledge may not be exhaustive
Solution Approach 1:
The system performs preliminary action by automatically learning and storing causal relationships from historical data during normal operation. This pre-computed causal knowledge base is then rapidly queried during fault diagnosis, eliminating the need for significant initial expert effort while maintaining comprehensive coverage of process relationships.
Solution Approach 2:
The system enables self-service by automatically acquiring causal knowledge from operational data without requiring continuous expert input. The causal discovery algorithm autonomously learns relationships between process variables, allowing the system to improve its diagnostic capability over time without increasing initial effort requirements.
2Device complexity
If data-driven methods are used for fault diagnosis, then minimum initial effort is required, but existing techniques cannot effectively identify nonlinear relationships and cannot deal with non-stationarity
Solution Approach 1:
The system applies dynamics by using a causal discovery algorithm that can adapt to changing process conditions and non-stationary data characteristics. The method dynamically learns causal relationships from operational data without requiring the process to be stationary, thereby maintaining reliability in identifying nonlinear relationships while keeping initial effort requirements minimal.
3Reliability
If manual causal network generation is performed by subject matter experts, then domain knowledge can be applied, but it is not practical for real-time fault identification in complex processes
Solution Approach 1:
The system replaces the mechanical process of manual causal network construction by experts with an automated computational algorithm. This substitution eliminates the time-consuming manual analysis while preserving the accuracy benefits of domain knowledge, enabling real-time causal network generation for complex industrial processes without sacrificing diagnostic reliability.
Data Source
AI summary
Fault diagnosis in industries typically involves identification of key variables/sensors bearing fault signature, classification of detected fault into known fault classes and detecting root causes/sources of the fault. This disclosure relates to a method and system for a deep learning based causal inference in a multivariate time series data of abnormal events and failures in industrial manufacturing processes and equipment. The system generates causal networks for non-linear and non-stationary multivariate time series data. The causal network learns for a dynamic non-stationary and nonlinear complex process or system fault using observed data without any prior process knowledge. The causal networks of faults are identified in real-time using a deep learning-based causal network learning technique. The system identifies causal connections and temporal lag information among variables to generate a directed causal graph of fault called the causal network, which is used to identify fault propagation paths and root cause variables.