Electrical Component Lifetime Prediction from Virtual Temperature Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the remaining lifetime of electrical components in building management devices require temperature measurement, which is costly and space-intensive, especially when predicting for multiple components.
Innovation Solution
A method using a trained machine learning model to estimate temperatures based on physical and operation parameters, generating a temporal course of temperature without actual measurement, and computing an indicator for the remaining lifetime of electrical components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If temperature measurement is implemented to predict remaining lifetime, then prediction accuracy is improved, but costs and space requirements increase
Solution Approach 1:
The patent creates a virtual copy of the temperature measurement function through machine learning models. Instead of using physical temperature sensors to directly measure temperature, the system uses ML models that compute estimated temperatures based on electrical parameters (voltage, current, frequency) as proxies. This virtual copying eliminates the need for physical sensing hardware while maintaining prediction capability.
Solution Approach 2:
The patent replaces the mechanical/physical temperature measurement system with a computational/electrical system. Rather than using thermal sensors that require physical contact and generate heat, the system uses electrical parameter measurements (voltage, current, frequency) processed through machine learning algorithms to estimate temperature and predict component lifetime, thereby eliminating the need for physical temperature sensing hardware.
2Reliability
If temperature measurement is implemented for multiple components, then comprehensive prediction is improved, but costs and space requirements increase significantly
Solution Approach 1:
The patent creates a universal prediction system that can monitor multiple different electrical components (capacitors, resistors, transistors, motors) using the same machine learning framework and electrical parameter measurements. The system is designed to be component-agnostic, using general electrical parameters (voltage, current, frequency) that can be measured for any electrical component, thereby providing comprehensive multi-component monitoring without requiring component-specific sensors or systems.
Solution Approach 2:
The patent applies virtual copying at system level by using identical machine learning models and measurement approaches for all electrical components. Instead of implementing separate physical temperature measurement systems for each component, the system uses a unified computational approach that estimates temperatures for multiple components simultaneously based on their respective electrical parameters, achieving comprehensive coverage with a single system architecture.
3Loss of information
If physical temperature sensors are used, then direct temperature data is obtained, but the system becomes more complex and expensive
Solution Approach 1:
The patent introduces machine learning models as intermediary computational layers between electrical parameter measurements and temperature estimation. Instead of directly measuring temperature with sensors, the system uses ML models that take electrical parameters (voltage, current, frequency) as input and produce estimated temperature as output. This intermediary computational approach bridges the gap between easily measurable electrical quantities and the desired temperature information without requiring direct thermal sensing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention provides a method for predicting a remaining lifetime of an electrical component of an electrical circuit, the electrical circuit being part of a building management device. The method comprises: estimating (Si) two or more estimated temperatures of the electrical component by using a trained machine learning model of the electrical component that is trained based on training data. Further, the method comprises: generating (S2) a temporal course of temperature of the electrical component based on the two or more estimated temperatures; and computing (S3), based on the temporal course of temperature of the electrical component, an indicator for the remaining lifetime of the electrical component. The present invention also provides a method for predicting a remaining lifetime of an electrical circuit being part of a building management device.