Industrial Causal Network Learning for Nonlinear Fault Root Causes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for real-time fault detection and diagnosis in industrial processes struggle to identify causal networks and root cause variables effectively, especially in nonlinear and non-stationary processes, due to the need for prior knowledge and limitations in data-driven techniques.

Innovation Solution

A deep learning-based system for causal network learning that processes multivariate time-series data to identify root cause variables and fault propagation paths without prior process knowledge, using a Deep Neural Model with encoding, time lag decomposition, and forecasting layers, and stochastic proximal gradient descent for adaptive learning.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If knowledge-based methods are used for root cause identification, then prior process knowledge can be utilized, but significant initial effort is required and knowledge may not be exhaustive

Engineering Contradiction:
Improveroot cause identification accuracyVSAvoidinitial effort and knowledge gathering
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system automatically learns causal networks from process data without requiring manual knowledge gathering or expert input. The causal network learning module autonomously discovers variable relationships and fault propagation paths, eliminating the need for a priori knowledge while maintaining high identification accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces manual knowledge-based methods with data-driven machine learning algorithms. Instead of relying on expert systems and manual causal network construction, the system uses automated learning from process data to identify root causes, substituting mechanical expert analysis with computational intelligence.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If data-driven methods are used for root cause identification, then minimum initial effort is required, but existing techniques cannot effectively identify nonlinear relationships and cannot deal with non-stationarity

Engineering Contradiction:
Improveinitial effortVSAvoidhandling of nonlinear and non-stationary data
Core Design Contradiction:
Ease of manufactureVSReliability

Solution Approach 1:

The system dynamically adapts to non-stationary process conditions by continuously learning from incoming data. The causal network learning module updates its understanding of variable relationships in real-time, allowing it to handle changing process conditions and non-stationary behavior effectively.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent transforms the approach to nonlinear relationships by using advanced machine learning models that can capture complex nonlinear dependencies. The system changes from linear statistical methods to nonlinear learning algorithms that adapt to the inherent nonlinearity in industrial processes.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

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 large number of process variables

Engineering Contradiction:
Improvefault localization accuracyVSAvoidreal-time identification capability
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs automatic causal network learning and root cause identification without requiring expert intervention. The module autonomously analyzes process data, identifies fault propagation paths, and localizes root causes in real-time, eliminating the time-consuming manual analysis while maintaining high accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent accelerates the causal analysis process by using powerful machine learning algorithms that rapidly process large volumes of process data. The system quickly identifies causal relationships and fault sources, dramatically reducing the time required compared to manual expert analysis.

Inventive Principle:
Principle #38Strong oxidants (Accelerated oxidation)

Data Source

PatentEP4310618B1Method and system for causal inference and root cause identification in industrial processes
Publication Date: 2024.11.13 TATA CONSULTANCY SERVICES LTD
  • EP4310618B1 patent drawingFigure 1
  • EP4310618B1 patent drawingFigure 2
  • EP4310618B1 patent drawingFigure 3

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.