Working Machine Diagnosis Using Classification and Frequency Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional abnormality diagnosis techniques for construction machines, such as hydraulic excavators, face challenges in accurately distinguishing between normal and abnormal operational states due to strong similarity between input signals, leading to erroneous determinations.
Innovation Solution
A monitoring and diagnosing device that utilizes both operational data classification information and operational data frequency comparison information to adaptively learn operational states, reducing erroneous determinations by comparing normalized statistical distances and frequency ratios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If statistical distance method (Mahalanobis-Taguchi method) is used for abnormality determination, then the determination process is simplified and can be performed automatically, but erroneous determination occurs when strong similarity exists between normal and abnormal operational states
Solution Approach 1:
The patent merges two different diagnostic approaches: the statistical distance method (Mahalanobis-Taguchi method) and the frequency distribution comparison method. By combining these methods, the system leverages the automation advantage of statistical distance while incorporating the accuracy advantage of frequency comparison, thereby resolving the contradiction between automatic determination and determination accuracy.
Solution Approach 2:
The patent introduces frequency distribution information as an intermediary layer between the raw operational data and the final abnormality determination. This intermediary provides additional contextual information about operational patterns, enabling more accurate distinction between normal and abnormal states while maintaining automated processing.
2Productivity
If threshold values are used for abnormality detection, then the detection process is simple and fast, but it cannot adapt to various operational manners and environmental conditions
Solution Approach 1:
The patent replaces static threshold values with dynamic frequency distribution information that adapts to different operational conditions. The frequency distribution is continuously updated based on actual operational data, allowing the system to adapt to various operational manners and environmental conditions while maintaining fast detection through automated comparison.
Solution Approach 2:
The patent changes the detection parameter from fixed threshold values to dynamic frequency distribution characteristics. By monitoring changes in frequency distribution over time and comparing them with reference distributions, the system achieves both fast detection and high adaptability to varying operational conditions.
3Ease of manufacture
If diagnostic algorithms are developed based on experimental environment data, then the algorithm development is systematic, but the algorithm performs poorly when applied to actual operational environments
Solution Approach 1:
The patent collects and stores frequency distribution information during normal operational conditions as preliminary data before actual diagnosis is performed. This preliminary action of gathering real operational data creates a reference baseline that reflects actual working conditions, improving the reliability of subsequent diagnostic comparisons.
Solution Approach 2:
The system incorporates feedback by continuously monitoring operational data and updating frequency distribution information based on actual performance. This feedback mechanism allows the diagnostic algorithm to adapt to real-world conditions, bridging the gap between experimental development and actual operational reliability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A monitoring and diagnosing device for a working machine configured is provided such that it can perform diagnosis appropriately and prevent erroneous determination even when strong similarity appears between input signals under a normal operational state and an abnormal operational state. The monitoring and diagnosing device comprises: a classification information storage section (102) in which reference classification information is stored; a frequency information storage section (105) in which reference frequency information is stored; a first data classifier section (101) which reads out the reference classification information from the classification information storage section, compares the operational data, detected by the plurality of sensors and inputted in time sequence, with the reference classification information to thereby classify the operational data, and then generates operational data classification information; a frequency comparator section (104) which compiles the operational data classification information, generates operational data frequency information by adding, to the operational data classification information, appearance frequency information for each of the classifications of the operational data, reads out the reference frequency information from the frequency information storage section, and then generates operational data frequency comparison information by comparing the operational data frequency information with the reference frequency information; and an abnormality diagnosing section (103) which performs an abnormality diagnosis upon the working machine by use of the operational data classification information and the operational data frequency comparison information.