Machine Fault Probability Modeling Using Sensor and Maintenance Logs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting and preventing faults in machinery rely on local views and assumptions, lacking a comprehensive approach to accurately model fault probabilities and prioritize maintenance tasks based on historical data from similar machines.
Innovation Solution
A method using machine learning models, such as random forest, kernel random forest, artificial neural network, or Bayesian networks, to analyze sensor and maintenance logs from similar machines to determine fault probabilities and prioritize maintenance tasks, thereby reducing the likelihood and severity of faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If local views and assumptions are used for fault prediction, then the system is simpler to implement, but the accuracy of fault probability modeling deteriorates
Solution Approach 1:
The patent merges data from multiple machines into a centralized database, combining sensor data, maintenance logs, and fault information from across the fleet. This aggregation enables comprehensive fault probability modeling that overcomes the limitations of isolated local views while maintaining manageable system complexity through standardized data collection protocols.
Solution Approach 2:
The system creates a universal fault prediction model that serves multiple machines simultaneously. The centralized database and analytics platform process data from various machine types and operational contexts, providing generalized fault probability assessments that can be applied across the entire machine fleet rather than requiring separate local systems for each machine.
2Measurement precision
If comprehensive historical data from similar machines is analyzed, then the fault probability modeling accuracy improves, but the data processing complexity increases
Solution Approach 1:
The system creates standardized data templates and structured formats for collecting machine information. By establishing uniform schemas for sensor data, maintenance logs, and fault records, the system enables efficient processing of comprehensive historical data without requiring complex custom processing logic for each machine type.
Solution Approach 2:
The patent transforms raw, unstructured historical data into standardized parameters and features suitable for statistical analysis. By converting diverse machine data into consistent parameter formats (e.g., operational hours, maintenance intervals, fault codes), the system reduces processing complexity while preserving the accuracy benefits of comprehensive data analysis.
3Reliability
If maintenance tasks are prioritized based on quantitative fault probabilities, then the effectiveness of preventative maintenance improves, but the requirement for historical data increases
Solution Approach 1:
The system performs preliminary data collection and storage by establishing centralized databases that continuously accumulate sensor data, maintenance logs, and fault information from all machines. This preliminary action ensures that sufficient historical data is already organized and available when fault probability analysis is needed, eliminating the need for extensive additional data gathering.
Solution Approach 2:
The system implements feedback loops where maintenance outcomes and actual fault occurrences are fed back into the centralized database. This continuous feedback enriches the historical data with verified information about which maintenance tasks prevented faults, progressively improving the accuracy of fault probability models and maintenance prioritization without requiring external data sources.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of determining a fault probability for a first machine, wherein the method comprises: accessing a plurality of sensor logs corresponding to a first machine, each sensor log spanning at least a first period; accessing first computer readable logs corresponding to the first machine, each computer readable log spanning at least the first period, the computer readable logs comprising a maintenance log comprising a plurality of maintenance task objects, each maintenance task object comprising a time and a maintenance task type; determining a set of statistical metrics derived from the sensor logs; determining a set of log metrics derived from the computer readable logs; and determining, using a risk model that receives the statistical metrics and log metrics as inputs, fault probabilities or risk scores indicative of one or more fault types occurring in the first machine within a second period.