Machine controller and methods for configuring and using the machine controller

Patent metadata

Field	Value
Publication number	US-12523971-B2
Application number	US-202118015831-A
Country	US
Kind code	B2
Filing date	Jul 13, 2021
Priority date	Jul 31, 2020
Publication date	Jan 13, 2026
Grant date	Jan 13, 2026

Abstract

A machine controller, geometry data and measured physical data of a machine is provided. The geometry data and the physical data are input to a machine learning module and to a simulation module of the machine controller. By the input data, the simulation module generates first values of a first physical property of a component of the machine on a discretized grid. Furthermore, an evaluator is provided for evaluating a physical compatibility of the first values with second values of a second physical property of the component, and for generating a residual quantifying the compatibility. The evaluator evaluates the compatibility of the first values with output data of the machine learning module and generates a resulting residual. Moreover, the machine learning module is trained to minimize the resulting residual, thus configuring the machine controller for controlling the machine by the output data of the trained machine learning module.

First claim

The invention claimed is:

1 . A computer-implemented method for configuring a machine controller comprising a machine learning module and a simulation module, the method comprising:
a) receiving geometry data and measured physical data of a machine to be controlled; b) inputting the geometry data and the physical data as input data to the machine learning module and to the simulation module; c) generating by the simulation module by first values of the input data of at least one first physical property of a component of the machine on a discretized grid; d) providing an evaluator for evaluating a physical compatibility of the first values with second values of at least one second physical property of the component different from the at least one first physical property, and for generating a residual quantifying the physical compatibility; e) evaluating by the evaluator the physical compatibility of the first values with output data of the machine learning module and generating a resulting residual; f) training the machine learning module to minimize the resulting residual, thus configuring the machine controller for controlling the machine by the output data of the trained machine learning module; and g) controlling the machine by the output data of the trained machine learning module.

2 . The method as claimed in claim 1 , wherein
the at least one first physical property is a stiffness of the component, a force acting on the component, a heat input to the component, and/or an electromagnetic input to the component.

3 . The method as claimed in claim 1 , wherein
the at least one second physical property is a strain, deformation, displacement, temperature, fluid property, vibration, or electromagnetic property of the component.

4 . The method as claimed in claim 1 , wherein
the simulation module uses a finite element method for generating the discretized grid and/or the first values.

5 . The method as claimed in claim 1 , wherein
the machine learning module comprises an artificial neural network, a recurrent neural network, a convolutional neural network, a reinforcement learning model, a Bayesian neural network, an autoencoder, a deep learning architecture, a support vector machine, a data driven trainable regression model, a k-nearest-neighbor classifier, a physical model and/or a decision tree.

6 . The method as claimed in claim 1 , wherein
the training of the machine learning module is performed by a reinforcement learning method, a gradient decent method, a particle swarm optimization method, and/or a genetic algorithm.

7 . The method as claimed in claim 1 , wherein
the evaluator determines the residual as a quantified measure of a non-fulfillment of a discretized physical partial differential equation relating the at least one first physical property as known quantity to the at least one second physical property as unknown quantity.

8 . A machine controller for controlling a machine, configured to perform a method according to claim 1 .

9 . A computer program product, comprising a computer readable hardware storage device having computer readable program code stored therein, said program code executable by a processor of a computer system to implement the method according to claim 1 .

10 . A non-transient computer readable storage medium storing a computer program product according to claim 9 .

11 . A computer-implemented method for controlling a machine, the method comprising:
providing a machine controller, wherein the machine controller includes a trained machine learning module and a simulation module, wherein the machine controller is a trained machine controller configured for controlling the machine by output data of the trained machine learning module, receiving operational data from the machine and inputting them as input data to the trained machine learning module and to the simulation module, generating by the simulation module by the input data first values of the at least one first physical property, outputting by the trained machine learning module output data as predicted second values of the at least one second physical property, generating by the evaluator a resulting residual quantifying a physical compatibility of the first values with the predicted second values, and controlling the machine by the predicted second values and the resulting residual.

12 . The computer-implemented method as claimed in claim 11 , wherein
depending on the resulting residual the predicted second values are accepted or discarded for controlling the machine.

13 . The computer-implemented method as claimed in claim 11 , wherein
by the simulation module and the trained machine learning module a digital twin of a component of the machine is constituted, and the digital twin is continuously supplied with operational data of the machine, thus simulating or representing a state or behavior of the component in real-time.

Assignees

• Siemens Ag

Inventors

• Ellath Meethal Rishith
• Obst Birgit

Classifications

• G05B13/027 Primary

  using neural networks only · CPC title

• G05B13/048 Primary

  using a predictor · CPC title