Virtual Sensor Modeling for Multi-Point Electric Motor Thermal Sensing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies face challenges in accurately determining physical and control parameters of complex systems like electric motors, particularly in real-time, due to spatial distribution of parameters and limitations in sensor placement, leading to inaccuracies in thermal management and control.

Innovation Solution

A device utilizing trained machine learning units, combining different machine learning models such as AdaBoostRegressor and BayesianRidge, to determine parameters like temperature, state of health, and coolant flow rate at multiple target points within an electric motor, allowing for precise and efficient thermal management.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If physical sensors are used to measure parameters at different spatial locations, then measurement accuracy may be improved, but device complexity and space requirements increase

Engineering Contradiction:
Improveparameter measurement accuracyVSAvoidsensor placement complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent creates virtual copies of physical sensors through machine learning models. Instead of deploying physical sensors at every measurement point, the system trains ML models to replicate sensor behavior and predict parameters at multiple spatial locations. This virtual sensing approach maintains measurement accuracy while eliminating the need for complex physical sensor networks.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/physical sensor system with an information-processing system based on machine learning. The physical measurement system is substituted by computational models that process input data and generate parameter predictions, thereby reducing device complexity and space requirements while maintaining measurement capabilities.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If physical sensors are installed at all necessary positions, then parameter determination accuracy improves, but space restrictions and installation difficulty worsen

Engineering Contradiction:
Improveparameter determination accuracyVSAvoidsensor installation ease
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The patent develops universal machine learning models that can determine multiple different parameters (temperature, pressure, flow rate, etc.) at multiple spatial locations using a single integrated system. This multi-functional approach eliminates the need for installing different physical sensors at different positions, greatly simplifying installation while maintaining comprehensive parameter determination capability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Measurement precision

If physical sensors are used to measure spatially distributed parameters, then gradient detection improves, but the number of sensors and system complexity increase

Engineering Contradiction:
Improvegradient detection accuracyVSAvoidnumber of sensors
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent uses machine learning models to create virtual sensors that replicate the measurement capabilities of physical sensors. By training these models on data from fewer physical sensors, the system can predict parameters and detect gradients at multiple locations without proportionally increasing the number of physical sensors required.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent combines multiple measurement functions and spatial locations into a unified machine learning model. Instead of using separate physical sensors for each measurement point, the system merges the measurement tasks into a single computational framework that processes input data and generates predictions for multiple parameters and locations simultaneously.

Inventive Principle:
Principle #5Merging (Combining)

4Measurement precision

If complex analytical models are used to determine parameters, then determination accuracy improves, but computational complexity and processing time increase

Engineering Contradiction:
Improveparameter determination accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent transforms the approach from using complex analytical models with many parameters to using machine learning models that learn parameter relationships from data. By changing from physics-based analytical models to data-driven ML models, the system achieves high determination accuracy while reducing computational complexity and processing requirements.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4383019A1Virtual sensor
Publication Date: 2024.06.12 HITACHI LTD
  • EP4383019A1 patent drawingFigure 1
  • EP4383019A1 patent drawingFigure 2
  • EP4383019A1 patent drawingFigure 3

AI summary

A device 1 for determining values of parameters related to a machine, the device 1 comprising an input interface for receiving input data; one or more sets 2 of determination units 3, wherein each set 2 includes a plurality of determination units 3 and each determination unit 3 within one set 2 is configured to determine a value of a predetermined parameter for a target point based on the received input data, and wherein one or more determination units 3 from one set 2 are combined into a subset 5 of determination units 3 and each subset 5 is configured to determine a value of the predetermined parameter for a target point associated to the subset 5; and an output interface for outputting a value determined by the subset 5.