Bayesian Data Fusion for Distributed Drive-train Condition Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current condition monitoring techniques for distributed drive-trains often focus on individual components, missing the system-wide perspective, leading to false alarms and unnecessary maintenance due to the propagation of fault signatures across components, which are not adequately addressed by existing data fusion methods.
Innovation Solution
A Bayesian data fusion method that processes physical signals from sensors across the drive-train components, using two stages of data fusion processes to determine local and global posterior probabilities of fault, integrating expert knowledge and degradation models to assess the drive-train condition and trigger alarms based on maximum value calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If condition monitoring focuses on individual components, then component-level diagnosis is simplified, but system-wide false alarms increase due to propagated fault signatures
Solution Approach 1:
The patent combines data from multiple sensors across different drive-train components into a unified analysis framework. By merging vibration, current, and temperature data from motors, gearboxes, and couplings into a single probabilistic model, the system achieves system-wide fault detection that distinguishes between localized component faults and propagated vibrations, thereby reducing false alarms while maintaining manageable complexity through integrated processing.
2Measurement precision
If data fusion from multiple components is implemented, then system-wide diagnosis accuracy improves, but computational complexity and data processing requirements increase
Solution Approach 1:
The patent transforms raw sensor data into standardized probabilistic parameters using Bayesian inference. By converting diverse sensor inputs (vibration amplitudes, current harmonics, temperature readings) into unified probability distributions that represent fault likelihoods, the system achieves high diagnostic accuracy while managing computational complexity through parameter standardization and modular probability calculations.
Solution Approach 2:
The patent introduces a Bayesian probabilistic model as an intermediary layer between raw sensor data and final diagnostic conclusions. This intermediary transforms complex multi-source data into intermediate probability distributions that represent the likelihood of various fault conditions, enabling accurate system-wide diagnosis while simplifying the integration of heterogeneous sensor data through a unified probabilistic framework.
3Productivity
If traditional single-component monitoring is used, then maintenance actions are quickly triggered, but unnecessary maintenance increases due to false alarms
Solution Approach 1:
The patent implements a feedback mechanism where the Bayesian system continuously updates fault probability assessments as new sensor data arrives. The system monitors trends in probability distributions over time and only triggers maintenance alerts when the probability of actual fault exceeds predetermined thresholds, providing feedback that distinguishes between transient vibrations and genuine faults, thereby reducing unnecessary maintenance while maintaining rapid response to real issues.
Data Source
AI summary
A method for condition monitoring of distributed drive-trains using Bayesian data fusion approach for measured data includes measurement of physical signals obtained from sensors attached to the components being chosen from the drive-train which are delivered to the computer means for processing the measured data and performing data fusion processes, using a data from information database containing at least one information system. The method is characterized by comprising two stages for data fusion processes performed by using Bayesian Inference, the first one for local data fusion process and the second one for global data fusion process, and on the basis of the second stage the assessment process of the condition of the drive-train is performed by choosing the maximum value of the received data, which maximum value serves as an indicator for the most likely fault present in the drive-train.