Multi-dimensional Sequential Pattern Mining for Faulty Behavior Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for monitoring and analyzing event data from machines are limited in detecting faulty behavior, especially in machine fleets with different but similar machine types, as they primarily focus on single-dimensional data mining, lacking the ability to assess the probability of sequential patterns and requiring manual analysis, which is not feasible due to the magnitude and complexity of the data.
Innovation Solution
A method that transfers logged event data from machines to a central processor for mining multi-dimensional sequential rules, incorporating attributes that indicate machine properties, allowing for the detection of patterns and rules across similar machines, and storing these as reference patterns in a central database for predictive maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If single-dimensional data mining methods are used, then the analysis process is simpler, but the ability to detect faulty behavior and assess pattern probability is insufficient
Solution Approach 1:
The patent transitions from single-dimensional event sequence analysis to multi-dimensional sequential pattern mining by incorporating additional dimensions such as machine type, component type, and operational parameters. This enables more comprehensive faulty behavior detection while systematically managing the increased complexity through structured dimensional frameworks.
2Measurement precision
If manual analysis of event data is performed, then detailed inspection is possible, but the process becomes infeasible due to data magnitude and complexity
Solution Approach 1:
The patent replaces manual mechanical analysis with automated computational algorithms for sequential pattern mining. The system automatically processes large volumes of event data, identifies patterns, and assesses probabilities without human intervention, maintaining high precision while dramatically reducing analysis time.
3Measurement precision
If multi-dimensional sequential rule mining is implemented, then pattern detection accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent segments the complex multi-dimensional analysis into manageable components: event data collection, sequential pattern identification, probability assessment, and rule generation. Each segment processes specific aspects of the data independently, reducing overall computational complexity while maintaining high detection accuracy through systematic breakdown of the analysis pipeline.
Data Source
AI summary
A method for evaluation of faulty behavior of at least one event-data-generating machine including a data logging device for providing event data, comprising: transferring logged event data, the logged event data representing a sequence of events, from the event-data-generating machines to a central processor, mining multi-dimensional sequential patterns and/or mining multi-dimensional sequential rules and/or mining anomalies and/or exceptions within the transferred event data wherein the event data additionally comprises at least one dimensional attribute holding information indicating the event-data-generating machine or at least one event-data-generating machine property; the difference between timestamps of first and last events included in an identified sequential pattern and/or sequential rule and/or anomaly and/or exception may not exceed a predefined time window; and storing at least one reference sequential pattern, sequential rule and/or anomaly and/or exception as an output of the mined multi-dimensional sequential patterns and/or sequential rules, and/or anomalies and/or exceptions in a central database.