Motor Control Circuit Using AI Rotor Position Estimation

Solution Overview

Problem

Existing motor control systems face challenges in achieving precise rotor position measurement without incurring high material costs, as precise physical sensors are expensive and limited by the quality of mathematical structure-based observers.

Innovation Solution

Implementing an AI-based virtual position sensor using a neural network trained with high-accuracy laboratory data to estimate rotor position from motor currents and control voltages, allowing for improved precision and cost-effectiveness.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If precise physical sensors (encoder or resolver) are used to measure rotor position, then measurement precision is improved, but manufacturing cost increases

Engineering Contradiction:

Improve rotor position measurement precisionVSAvoidmanufacturing cost

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent creates a virtual copy of the physical position sensor by training a neural network model using data from a high-precision physical sensor during a teaching phase. The trained model then serves as a software-based position sensor that replicates the measurement function without requiring expensive hardware, thereby resolving the contradiction between measurement precision and manufacturing cost

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/physical position sensing system (encoder or resolver) with an electronic/software-based neural network model. This substitution eliminates the need for expensive physical sensors while maintaining position measurement capability through current and voltage signal processing, thus reducing manufacturing cost while preserving measurement precision

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If mathematical structure-based observers are used to determine rotor position, then manufacturing cost is reduced, but measurement precision deteriorates

Engineering Contradiction:

Improvemanufacturing costVSAvoid rotor position measurement precision

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent creates a virtual copy of the physical position sensor by training a neural network model using data from a high-precision physical sensor during a teaching phase. The trained model then serves as a software-based position sensor that replicates the measurement function without requiring expensive hardware, thereby resolving the contradiction between measurement precision and manufacturing cost

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent transforms the approach from using fixed mathematical structure-based observers to a dynamic neural network model with learned parameters. The neural network adapts its internal parameters through training on high-precision sensor data, enabling it to achieve high measurement precision while maintaining the cost advantages of software-based solutions

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If high-precision physical sensors are built into the vehicle, then control quality is improved, but energy consumption increases

Engineering Contradiction:

Improvecontrol qualityVSAvoidenergy consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent creates a virtual copy of the physical position sensor by training a neural network model using data from a high-precision physical sensor during a teaching phase. The trained model then serves as a software-based position sensor that replicates the measurement function without requiring expensive hardware, thereby resolving the contradiction between measurement precision and manufacturing cost

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The neural network model processes existing current and voltage measurements that are already available in the motor control system to infer rotor position. This self-service approach uses readily available data without requiring additional sensors or increased energy consumption, achieving high control quality while maintaining energy efficiency

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12580508B2Motor control circuit

Publication Date: 2026.03.17 INFINEON TECHNOLOGIES AG

AI summary

According to various embodiments, a motor control circuit is described having a controller configured to determine values of a plurality of control voltages for a motor. The motor control circuit includes one or more current sensors configured to measure a plurality of operation currents of the motor and a neural network having a multi-layer perceptron architecture. The neural network is trained to estimate a rotor position of the motor for a current control cycle. The controller is configured to determine values of the plurality of control voltages for the current control cycle using the estimate of the rotor position.