Industrial Machine Abnormality Judgment Using Energy-State Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing state judgment devices for industrial machines face challenges in accurately discerning abnormal states due to variability in machine components and production materials, leading to high costs and inefficiencies in data collection and model application across diverse industrial machines.
Innovation Solution
A state judgment device that utilizes kinetic and electric energy from various physical quantities, such as rotation speed and current, to estimate the degree of abnormality in industrial machines, allowing for a versatile and efficient application of a single learning model across different machines, and provides alerts or operational adjustments for safe operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If diverse learning conditions are prepared for each combination of machine parts and members to increase machine learning accuracy, then the accuracy of abnormality judgment is improved, but the cost and time required for data collection increase significantly
Solution Approach 1:
The patent transforms the learning approach by changing from collecting diverse physical quantity data under various operating conditions to calculating energy values (kinetic energy, potential energy, internal energy) from basic operational parameters. This parameter transformation allows a single learning model to generalize across different machine configurations without requiring extensive diverse data collection, thereby maintaining high judgment accuracy while reducing data collection time and cost.
2Device complexity
If a single learning model is applied to diverse industrial machines to reduce cost and complexity, then device complexity is reduced, but the accuracy of abnormality judgment deteriorates due to variability in machine parts and materials
Solution Approach 1:
The patent creates a universal learning model that can be applied to diverse industrial machines by using energy values as learning data. The energy calculations (kinetic energy from motion, potential energy from position, internal energy from thermal states) are universally applicable across different machine types, configurations, and materials. This allows one learning model to serve multiple machine variants accurately, eliminating the need for separate models for each machine configuration.
3Measurement precision
If physical quantities directly from sensors are used for machine learning to maintain data accuracy, then measurement precision is preserved, but the variability in machine components causes large discrepancies between measured values and learning data
Solution Approach 1:
The patent introduces energy values as an intermediary between raw sensor measurements and the learning model. Instead of directly using diverse physical quantities from different sensors and machine configurations, the system calculates standardized energy values (kinetic, potential, internal) that serve as a common language. This intermediary transformation enables consistent comparison and learning across different machine types while preserving the essential operational information.
Data Source
AI summary
A state judgment device includes: a data acquisition unit which acquires data related to an industrial machine; an energy state calculation unit which calculates an energy state related to driving of units of the industrial machine on the basis of the data related to the industrial machine acquired by the data acquisition unit; and an abnormal state estimation unit which estimates, on the basis of the energy state related to driving of the units of the industrial machine calculated by the energy state calculation unit, whether operation of the industrial machine is normal or abnormal.