Sensors for determining an output value, method for evaluating a sensor signal, and method for training an output unit to evaluate a sensor signal

Patent metadata

Field	Value
Publication number	US-11573100-B2
Application number	US-202017101994-A
Country	US
Kind code	B2
Filing date	Nov 23, 2020
Priority date	Nov 28, 2019
Publication date	Feb 7, 2023
Grant date	Feb 7, 2023

Abstract

The invention relates to a sensor for determining an output value, the sensor having: a detection unit configured to detect a sensor signal; a preprocessing unit configured to determine an intermediate signal on the basis of the sensor signal and of a predefined reference signal; and an evaluation unit that is trained in accordance with a machine learning method and that is configured to determine the output value on the basis of the intermediate signal.

First claim

The invention claimed is:

1. A sensor for determining an output value comprising:
a detection unit configured to detect a sensor signal;
a preprocessing unit configured to determine an intermediate signal on the basis of the sensor signal and of a predefined reference signal, wherein the predefined reference signal comprises a sensor signal of the sensor and/or a sensor signal of a reference sensor; and
an evaluation unit that is trained in accordance with a machine learning method and that is configured to determine the output value on the basis of the intermediate signal.

2. The sensor in accordance with claim 1 , wherein the sensor is an inductive proximity sensor.

3. The sensor in accordance with claim 1 , wherein the detection unit is configured to digitize the sensor signal.

4. The sensor in accordance with claim 1 , wherein the preprocessing unit is configured to preprocess the sensor signal on the basis of at least one of a transformation, a linear principal component analysis and a linear discriminant analysis.

5. The sensor in accordance with claim 1 , wherein a separate reference pulse is recorded and stored as a reference signal for each sensor.

6. The sensor in accordance with claim 1 , wherein the preprocessing unit is configured to determine at least one intermediate signal on the basis of the sensor signal and of at least one predefined reference signal that is selected from a plurality of reference signals.

7. The sensor in accordance with claim 1 , wherein the preprocessing unit is configured to determine the intermediate signal on the basis of a difference between the sensor signal and the predefined reference signal.

8. The sensor in accordance with claim 1 , wherein the evaluation unit performs at least one Gaussian process regression and/or has at least one support vector machine and/or at least one decision tree and/or at least one artificial neural network and/or at least one linear model.

9. The sensor in accordance with claim 1 , further comprising:
at least one coil; and
means for feeding the at least one coil.

10. The sensor in accordance with claim 1 , wherein the output value represents a distance of an object from the sensor.

11. The method for evaluating a sensor signal in accordance with claim 1 , which is a method for evaluating a sensor signal of an inductive proximity sensor.

12. A method for evaluating a sensor signal, in particular a method for evaluating a sensor signal of an inductive proximity sensor, the method comprising:
detecting a sensor signal;
determining an intermediate signal on the basis of the sensor signal and of a predefined reference signal, wherein the predefined reference signal comprises a sensor signal of the sensor and/or a sensor signal of a reference sensor; and
determining an output value on the basis of the intermediate signal by means of an evaluation unit trained in accordance with a machine learning method.

13. A method for training an evaluation unit to evaluate a sensor signal, the method comprising:
providing a plurality of training sensor signals, wherein a corresponding training output value is associated with each training sensor signal;
determining an intermediate signal for each training sensor signal, said intermediate signal being associated with the respective training sensor signal, on the basis of the respective training sensor signal and of a predefined reference signal, wherein the predefined reference signal comprises a sensor signal of the sensor and/or a sensor signal of a reference sensor;
determining an output unit output value for each training sensor signal, said output unit output value being associated with the respective training sensor signal, on the basis of the intermediate signal associated with the respective training sensor signal using the evaluation unit; and
training the evaluation unit on the basis of the output unit output values and of the training output values.

14. The method in accordance with claim 13 , which is a method for training an evaluation unit to evaluate a sensor signal of an inductive proximity sensor.

15. The method in accordance with claim 13 , wherein the training of the evaluation unit comprises determining parameters used in the evaluation unit to determine the output unit output values.

16. A method in accordance with claim 15 , wherein the parameters are determined by means of an optimization method.

Assignees

• Sick Ag

Inventors

• Peter Andreas
• Thoss Sascha
• Machul Olaf

Classifications

• G06N3/09

  Supervised learning · CPC title

• G01D5/20 Primary

  by varying inductance, e.g. by a movable armature · CPC title

• G06N20/20

  Ensemble learning · CPC title

• H03K17/94 Primary

  characterised by the way in which the control signals are generated · CPC title

• G06N3/045

  Combinations of networks · CPC title