Neural Network Control for Sensorless Vibration-Type Actuators
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vibration-type actuators require speed and torque sensors, making them difficult to miniaturize due to the need for additional hardware components.
Innovation Solution
A control device for vibration-type actuators that utilizes a neural network to control speed and thrust without speed or torque sensors, using a vibrating body with an elastic body and electro-mechanical energy conversion element, and a contact body that moves relative to the vibrating body, receiving command values for relative speed and thrust, and outputs a command for the alternating-current voltage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a torque sensor is used for speed estimation, then speed control accuracy is improved, but device complexity increases and miniaturization becomes difficult
Solution Approach 1:
The patent extracts and eliminates the torque sensor from the system by using alternative measurement approaches. Instead of directly measuring torque with a sensor, the system uses current detection combined with neural network-based estimation to achieve speed control without the physical torque sensor, thereby reducing device complexity while maintaining control accuracy
Solution Approach 2:
The patent replaces the mechanical torque sensor with an electronic/neural network-based estimation system. By substituting the physical sensing mechanism with a computational model that processes current signals and electrical parameters, the system achieves torque estimation without mechanical sensors, enabling miniaturization
2Measurement precision
If a speed sensor is used for speed control, then speed control accuracy is improved, but device complexity increases and miniaturization becomes difficult
Solution Approach 1:
The patent extracts and eliminates the speed sensor from the system by using alternative measurement approaches. Instead of directly measuring speed with a sensor, the system uses current detection combined with neural network-based estimation to achieve speed control without the physical speed sensor, thereby reducing device complexity while maintaining control accuracy
Solution Approach 2:
The patent replaces the mechanical speed sensor with an electronic/neural network-based estimation system. By substituting the physical sensing mechanism with a computational model that processes current signals and electrical parameters, the system achieves speed estimation without mechanical sensors, enabling miniaturization
3Measurement precision
If torque sensor and speed sensor are used, then control precision is improved, but actuator size increases
Solution Approach 1:
The patent extracts and eliminates both torque and speed sensors from the system, using only current detection units. The neural network-based estimation system processes electrical parameters to infer mechanical state, removing the need for physical sensors and enabling actuator miniaturization while maintaining control precision
Solution Approach 2:
The patent replaces mechanical sensing systems with an electronic estimation system based on neural networks. By substituting physical torque and speed sensors with computational models that process electrical signals, the system achieves the same control precision without the volume occupied by mechanical sensors
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables precise control of speed and thrust without the need for speed or torque sensors, facilitating miniaturization of the actuator.
Implementation Method 1
a vibrating body (5), which includes an elastic body (1) and an electro-mechanical energy conversion element (2), and a contact body (6), which is in contact with the elastic body (1)
Data Source
AI summary
A control device for a vibration-type actuator including a vibrating body and a contact body, the contact body being moveable relative to the vibrating body by vibration excited by applying an alternating-current voltage includes a neural network including an input layer which receives, as inputs, a detected value and at least one of a command value for a relative speed of the contact body with respect to the vibrating body and a command value for a thrust occurring between the vibrating body and the contact body, an intermediate layer which performs an arithmetic operation according to a signal received from the input layer, and an output layer which outputs a command for a manipulated variable for the alternating-current voltage, wherein the detected value is a detected value of a signal related to the alternating-current voltage or a detected value of vibration of the vibrating body.