US-12625493-B2 - Monitoring methods, computer program product, monitoring unit, gas analysis device, and use of artificial intelligence

US12625493B2US 12625493 B2US12625493 B2US 12625493B2US-12625493-B2

Abstract

A computer program product, a monitoring unit, gas analysis device equipped with the monitoring unit, use of artificial intelligence for monitoring the gas analysis device, and monitoring methods for a system having a plurality of devices which are configured to provide an associated measured value and/or a control command, wherein the monitoring methods are based on the use of a processing device or a neural network, with which measured values and/or control commands are evaluated.

Inventors

Jana EDER
Simon Weilandt
Ralf Bitter
Stephanie HOLLY
Tim OFFERMANN
Daniel Schall
Piotr Strauch

Assignees

SIEMENS AKTIENGESELLSCHAFT

Dates

Publication Date: 20260512
Application Date: 20230524
Priority Date: 20220615

Claims (11)

1 . A monitoring method for a gas analyzer having a plurality of devices, which form nodes in a neural network, which maps an interaction of the devices in the gas analyzer, the method comprising: a) providing a trained neural network having a plurality of edges which each correspond to a zero correlation or an operating correlation, a correlation value being determined during training for zero correlations, and the correlation value having a value which is below an adjustable correlation threshold value; b) placing, by a controller, the gas analyzer into an active operating state and capturing, by the controller, at least one of measured values and control commands, each measured value and control command corresponding to least one of (i) at least one node and (ii) at least one edge of the neural network; c) determining, by the controller, a network deviation parameter which combines correlation values of a plurality of edges of the neural network, the correlation values being determined utilizing at least one of the measured values and control commands captured during step b); d) identifying, by the controller, an abnormal operating state of the system, if a value of the network deviation parameter exceeds an adjustable network threshold value, a warning being issued if the abnormal operating state is identified; and e) identifying an edge of the neural network which corresponds to the zero correlation in a normal operating state as a defective edge, if a value of the correlation value associated with the identified edge exceeds the associated correlation threshold value; and f) initiating a countermeasure during the active operating state of the gas analyzer to increase an efficiency of error detections within the gas analyzer based on a reduced computational load when the value of the correlation value associated with the identified edge exceeds the associated correlation threshold value such that the identified edge of the neural network which corresponds to the zero correlation in the normal operating state is identified as a defective edge.
2 . The monitoring method as claimed in claim 1 , wherein at least one of steps b), c) and d) are performed, by the controller, with a first frequency for operating correlations and are performed with a second frequency for zero correlations.
3 . The monitoring method as claimed in claim 1 , wherein at least one of steps b), c) and d) are performed, by the controller, in a first pass for operating correlations and in a second pass for zero correlations, if an abnormal state is identified in the first pass.
4 . The monitoring method as claimed in claim 1 , wherein at least one of steps b), c) and/or d) are performed, by the controller, in a first pass for operating correlations and in a second pass for zero correlations, if an abnormal state is identified in the first pass.
5 . The monitoring method as claimed in claim 1 , wherein the correlation threshold value is established as a fixed threshold value or as a moving threshold value.
6 . The monitoring method as claimed in claim 2 , wherein the correlation threshold value is established as a fixed threshold value or as a moving threshold value.
7 . The monitoring method as claimed in claim 3 , wherein the correlation threshold value is established as a fixed threshold value or as a moving threshold value.
8 . The monitoring method as claimed in claim 1 , the monitoring method further comprising: f) performing, by the controller, a pattern matching and based on at least one identified defective edge and identifying a cause of a defect.
9 . A non-transitory computer readable medium encoded with a computer program which, when executed by processor of a monitoring unit for monitoring a gas analyzer which has a plurality of devices which interact to operate the gas analyzer, causes at least one of measured values and control commands to be captured and processed, the program comprising: a) program code for establishing a trained neural network having a plurality of edges which each correspond to a zero correlation or an operating correlation, a correlation value being determined during training for zero correlations, said correlation value have a value which is below an adjustable correlation threshold value; b) program code for placing, by a controller, the gas analyzer into an active operating state and capturing at least one of measured values and control commands, each measured value and control command corresponding to least one of (i) at least one node and (ii) at least one edge of the neural network; c) program code for determining, by the controller, a network deviation parameter which combines correlation values of a plurality of edges of the neural network, the correlation values being determined utilizing at least one of the measured values and control commands captured during step b); d) program code for identifying, by the controller, an abnormal operating state of the gas analyzer, if a value of the network deviation parameter exceeds an adjustable network threshold value, a warning being issued if the abnormal operating state is identified; and e) program code for identifying an edge of the neural network which corresponds to the zero correlation in a normal operating state as a defective edge, if a value of the correlation value associated with the identified edge exceeds the associated correlation threshold value; wherein a countermeasure is initiated during the active operating state of the gas analyzer increase an efficiency of error detections within the gas analyzer based on a reduced computational load when the value of the correlation value associated with the identified edge exceeds the associated correlation threshold value such that the identified edge of the neural network which corresponds to the zero correlation in the normal operating state is identified as a defective edge.
10 . A monitoring unit for gas analyzer which comprises the plurality of devices which interact to operate the gas analyzer and for which measured values and/or control commands can be captured, wherein the monitoring unit is configured to receive and process at least one of the measured values and control commands and is configured to issue a warning, and wherein the monitoring unit includes the non-transitory computer readable medium as claimed in claim 9 to at least one of process the measured values and control commands.
11 . The gas analyzer comprising the plurality of devices for conditioning and measuring a material sample, the gas analyzer being configured with the monitoring unit and identifying an abnormal operating state of the gas analyzer, wherein the monitoring unit is configured in accordance with claim 10 .

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This is a U.S. national stage of application No. PCT/EP2023/063947 filed 24 May 2023. Priority is claimed on European Application No. 22179239.3 filed 15 Jun. 2022, the content of which is incorporated herein by reference in its entirety. BACKGROUND OF THE INVENTION 1. Field of the Invention The invention relates to a monitoring method for monitoring a system, a computer program product, a monitoring unit which has a corresponding computer program product, a gas analysis device having such a monitoring unit, and further relates to the use of artificial intelligence for monitoring the gas analysis device. 2. Description of the Related Art Printed publication KR 2021147318 A discloses a multi-sensor-based artificial intelligence that is designed for error diagnosis in a mechanical device. Sensor data is collected therein and is processed by means of an autoencoder. US Pub. No. 2020/0310370 A1 discloses a control system having an autoencoder, in which sensor data is processed into derived sensor data. The autoencoder receives the derived sensor data and from it determines a prediction. Based on the prediction, a control intervention is triggered. In various automation technology applications, increasingly complex systems with an increasing number of individual devices are being employed. The aim is also to have a flexible operation for processing different input materials. As a result of the increasing complexity of operations and systems, monitoring such systems is becoming more demanding. SUMMARY OF THE INVENTION In view of the foregoing there is therefore a need for a facility for monitoring complex systems that offers reliable automatic error detection and that can be quickly adapted to changing operational requirements. This applies in particular for gas analysis devices, which are becoming increasingly more complex and sensitive due to increasing demands on measurement accuracy. The objects and advantages are achieved in accordance with the invention by a monitoring method for monitoring an operation of a system that comprises a plurality of devices that are each configured to provide a measured value and/or a control command. The devices can, for example, be sensors, control units, regulation units or combinations thereof. The measured values can be captured physical variables or information about the state of the respective device. The monitoring method comprises a first step, in which an input signal array is provided that comprises a plurality of cells. A cell of the input signal array is suitable for receiving a specified measured value or control command and for passing it to a further processing unit. The input signal array can be formed as a one-, two- or higher-dimensional array, i.e., a data field. Furthermore, in the first step a monitoring data array is provided, which corresponds to the input signal array in terms of dimensionality and size, i.e., the number of cells. In each case, a cell of the input signal array corresponds to a cell of the monitoring data array. The monitoring method also has a second step, in which the system is provided in an active operating state. In the active operating state, measured values or control commands are generated essentially continuously and are provided for the monitoring method for further processing. The captured measured values and/or control commands are transferred to the input signal array and are fed into the corresponding associated cells as contents. With the further processing unit, these are processed into contents of corresponding cells in the monitoring data array. In a third step, a system deviation parameter is determined based on the second step into the input signal array. The system deviation parameter is formed by a plurality of cells of the monitoring data array. The system deviation parameter can, for example, be formed as a sum of contents of cells of the monitoring data array or as a sum of squared contents of the cells of the monitoring data array. The contents of the cells of the monitoring data array can each be standardized to a reference value individually or as several together. The contents are configured to be type-compatible with contents of the corresponding cells of the input signal array. A weighting of individual cells of the monitoring data array is also possible. The weighting can be performed by an algorithm before or during the operational sequence of the monitoring method. If the system deviation parameter exceeds an adjustable system threshold value in the third step, the presence of an abnormal state is identified. The monitoring method further has a fourth step, which can be performed if an abnormal operating state of the system is identified in the third step. Here, the contents of the cells of the monitoring data array combined in the system deviation parameter are further processed separately. Using contents of the cells of the monitoring data array and contents of co