Component State Detection Using Cross-Device ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting the operating states of devices are inefficient and require significant effort, and there is a need for improved reliability in identifying these states.
Innovation Solution
A computer-implemented method that subdivides devices into components, groups similar or identical components together, and trains separate machine-learning models on databases created from component-specific data, allowing for accurate and reliable detection of operating states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If separate ML models are trained for each device component using device-specific data only, then the model training effort and data collection requirements increase significantly, but the detection reliability remains limited to device-specific patterns
Solution Approach 1:
The patent merges data from multiple devices and similar components into a unified training database. Instead of collecting and training on device-specific data separately, the system combines data across devices and component types, training a single ML model on aggregated data that encompasses multiple operating scenarios and component variations, thereby reducing data collection effort while improving generalization capability
Solution Approach 2:
The patent creates a universal ML model that can detect operating states across different device types and component configurations. The model is trained on diverse data from multiple devices and component groups, enabling it to function universally across various application scenarios without requiring device-specific customization or separate training processes
2Reliability
If ML models are trained on device-specific data only, then the models can detect known operating states accurately, but they fail to reliably detect operating states under previously unknown conditions
Solution Approach 1:
The patent performs preliminary training of the ML model on diverse data from multiple devices and component groups before deployment. This advance training on varied data sets prepares the model to handle unknown operating conditions by exposing it to a broad range of patterns and variations during the training phase, improving its ability to generalize to new scenarios
Solution Approach 2:
The patent varies training parameters by using data from different devices, component types, and operating conditions during model training. By changing the input data parameters and diversity during training, the model learns to recognize patterns across different configurations and conditions, enhancing its adaptability to previously unknown operating states
3Productivity
If traditional pattern detection methods are used to identify operating states, then the implementation effort is high and the detection efficiency is low, but the approach provides useful results
Solution Approach 1:
The patent replaces traditional mechanical pattern detection methods with a machine learning-based detection system. Instead of using conventional algorithms that require extensive manual pattern matching and analysis, the system uses an ML model that automatically learns and detects operating states from training data, significantly improving detection efficiency while reducing implementation effort through automated feature extraction and classification
Data Source
AI summary
A computer-implemented method for processing machine-related data in order to obtain at least one trained machine-learning model and to a computer-implemented method for detecting the operating state of a component of a device. Provided is also a device for processing data, and designed to carry out a respective method of the computer-implemented methods or both computer-implemented methods, and to a data structure.