WO2023117317A1

WO2023117317A1 - Method for the automated inspection of a field device

Info

Publication number: WO2023117317A1
Application number: PCT/EP2022/083494
Authority: WO
Inventors: Thorsten Springhart; Arno Schüler; Stefan Griner
Original assignee: Endress+Hauser Process Solutions Ag
Priority date: 2021-12-22
Filing date: 2022-11-28
Publication date: 2023-06-29
Also published as: DE102021134288A1

Abstract

The invention comprises a method for the automated inspection of a field device (FG), wherein the field device (FG) comprises a multiplicity of components (KO1, KO2), in particular a sensor or actuator unit, a housing, a display operating unit, at least one electronics unit, at least one electronic connection, and/or cabling, and wherein the field device (FG) is surrounded by one or more surroundings components (UK) or is connected to the at least one surroundings component (UK), comprising: - capturing a photo (FO) of the field device (FG); - transmitting the photo (FO) to a cloud application (CL), wherein a digital twin (TW) of the field device (FG) is stored in the cloud application (CL), wherein the digital twin (TW) contains a three-dimensional model of the field device (FG); - comparing the photo (FO) with the three-dimensional model and identifying changes in components (KO1, KO2) of the field device (FG) between the photo (FO) and the three-dimensional model; and - notifying a user (BN), by way of the cloud application (CL), if there is at least one identified change in at least one of the components (KO1, KO2) of the field device (FG) or in at least one of the surroundings components (UK).

Description

Method for the automated inspection of a field device

The invention relates to a method for the automated inspection of a field device, the field device consisting of a large number of components, in particular a housing, a display/operating unit, at least one electronic unit, at least one electronic connection, and/or cabling, and the field device is surrounded by one or more environmental components or is connected to the at least one environmental component.

Field devices that are used in industrial plants are already known from the prior art. Field devices are often used in process automation technology as well as in production automation technology. In principle, all devices that are used close to the process and that supply or process process-relevant information are referred to as field devices. Thus, field devices are used to record and/or influence process variables. Measuring devices or sensors are used to record process variables. These are used, for example, for pressure and temperature measurement, conductivity measurement, flow measurement, pH measurement, level measurement, etc. and record the corresponding process variables pressure, temperature, conductivity, pH value, level, flow rate, etc. Actuators are used to influence process variables. These are, for example, pumps or valves that can influence the flow of a liquid in a pipe or the fill level in a container. In addition to the measuring devices and actuators mentioned above, field devices also include remote I/Os, wireless adapters or devices in general that are arranged at the field level.

A large number of such field devices are produced and sold by the Endress+Hauser Group.

In modern industrial plants, field devices are usually connected to higher-level units via communication networks such as fieldbuses (Profibus®, Foundation® Fieldbus, HART®, etc.). Normally, the higher-level units are control systems (DCS) or Control units such as a PLC (programmable logic controller). The higher-level units are used, among other things, for process control, process visualization, process monitoring and for commissioning the field devices. The measured values recorded by the field devices, in particular by sensors, are transmitted via the respective bus system to one (or optionally several) higher-level unit(s). In addition, data transmission from the higher-level unit via the bus system to the field devices is also required, in particular for configuring and parameterizing field devices and for controlling actuators.

In the course of Industry 4.0 or lloT (“Industrial Internet of Things”), the data generated by the field devices is often collected directly from the field using so-called data conversion units, which are referred to as “edge devices” or “cloud gateways”, for example and automatically transmitted to a central cloud-enabled database (also referred to simply as “cloud”) on which an application is located. This application, which offers, among other things, functions for the visualization and further processing of the data stored in the database, can be accessed by a user via the Internet.

Using these methods, it is possible to monitor the electronic components of the system—that is, the field devices and control units—using the data transmitted by these electronic components, for example in the sense of predictive maintenance. With these methods, however, the consideration or the influence of purely mechanical environmental components, such as pipelines, containers or cables, are left out. The change in such environmental components, for example in the event of build-up or leakage, could only be detected indirectly, for example via changed measurement characteristics of the field devices. In the event of an error, a long period of time is therefore sometimes required to determine the exact cause of the error.

If acute problems occur with a field device, this often results in diagnostic and/or maintenance messages, which the field device transmits to the control system or to the cloud. If the problem is not immediately apparent from the diagnostic and/or maintenance message, a service technician often goes to the measuring point of the field device to carry out a visual inspection of the field device. Problems occurring on the field device and/or on the surrounding components of the field device (e.g. pipes, containers, cables, etc.) are identified by comparing the optical actual status of the field device with static target images or based on the experience of the service technician.

Such target images can be product images from the manufacturer, but can also be created during the manufacture of the field device or during the commissioning of the field device of the field device.

The disadvantage here is that the service technician must be located directly at the installation site of the field device. The inspection process is often time-consuming and the result sometimes depends on the experience of the service technician. Furthermore, the target images are often not up-to-date (e.g. the appearance of the field device can be changed due to dust or dirt). In this way, it is also possible to only react to a problem instead of being able to act in a foresighted or proactive manner.

Proceeding from this problem, the invention is based on the object of presenting a method which allows optical changes on the field device to be reliably detected in a simple manner.

The object is achieved by methods for the automated inspection of a field device, wherein the field device consists of a large number of components, in particular a housing, a display/operating unit, at least one electronic unit, at least one electronic connection, and/or cabling, and wherein the Field device surrounded by one or more environmental components or connected to the at least one environmental component, comprising:

- capture a photo of the field device;

- Transmission of the photo to a cloud application, wherein a digital twin of the field device is stored on the cloud application, wherein the digital twin contains a three-dimensional model of the field device; - comparing the photo with the three-dimensional model and identifying changes in components of the field device between the photo and the three-dimensional model; and

- Notifying a user through the cloud application in the event of the presence of at least one identified change in at least one of the components of the field device or in at least one of the environmental components.

The core of the method according to the invention is that changes in the field device and its environment can be inferred from a comparison of a currently recorded photo of a field device (actual state) with a 3D model of the field device (target state). The detected changes can be such that the field device is already damaged and cannot or no longer reliably perform its intended task. However, the changes can also be such that they allow a prediction of the forthcoming operating time of the field device, ie, for example, whether the field device could fail in the future.

The procedure is carried out automatically. This means that there is no longer a need for a human service technician to carry out a visual analysis of the field device. The method according to the invention is therefore faster and more reliable (due to the different empirical values of individual service technicians) than known methods. In addition, not only direct components of the field device can be checked, but also so-called environmental components.

These environmental components are components of the system that are directly connected to the field device (e.g. a container, e.g. a tank, or a pipeline to which the field device is connected to carry out its intended purpose) or are in the immediate vicinity environment to this are located.

The 3D model of the field device is contained in a digital twin of the field device or linked to it. A digital twin, also digital An image is a virtual representation of the field device that has the identical configuration, parameter values, current device status, algorithms, etc. of the field device. The digital twin thus has all the properties of the field device that fully describe the field device for its intended purpose. It is envisaged that the field device and the digital twin will always be identical. A change in the properties of the field device leads to a synchronization (via Industry 4.0 or HoT techniques), so that the properties of the digital twin are updated accordingly.

The digital twin is integrated in a cloud application. A cloud application is a program that runs on or is integrated into a cloud-based platform (known as a cloud for short) or a server. The terms cloud-based platform, cloud and server are to be understood as synonymous in the context of this application. The cloud-based platform can be accessed via the Internet. A user can connect to the cloud-based platform via the internet and make modifications in the corresponding cloud applications of the cloud-based platform and/or operate them, ie write data to the cloud applications, read data from the cloud applications and/or edit this data.

Field devices, which are described in connection with the method according to the invention, have already been listed and defined as examples in the introductory part of the description. Within the scope of the method according to the invention, network devices such as gateways and edge devices are also referred to as field devices.

According to an advantageous development of the method according to the invention, it is provided that the cloud application or the digital twin uses an AI system, in particular based on at least one neural network and/or at least one machine learning algorithm and/or at least one deep learning algorithm , which AI system performs the steps of comparing the photo to the three-dimensional model and identifying changes. Any type of Kl algorithm can be used for this purpose, which is suitable, e.g. by cluster identification, changes in the actual state of the To recognize field device compared to the target state of the field device based on the comparison. In this case, the K1 algorithm is taught in particular with training data containing photos of field devices linked to the statement as to whether changes are present, which bring the visual damage patterns corresponding to the K1 algorithm closer.

An advantageous embodiment of the method according to the invention provides that the cloud application requests feedback from the user regarding the identified change, with the user confirming or denying for the feedback that the identified change is an actual change, with the feedback being given to Kl system is fed, with the AI system learning from any user feedback. As a result, the Kl algorithm does not necessarily have to be taught in advance, but learns independently over time. Here, however, an increased administrative effort is necessary, since more changes can be discovered, especially at the beginning of the process. But even a K1 algorithm that has already been taught can be made more precise in this way.

According to an advantageous embodiment of the method according to the invention, it is provided that the AI system accesses one or more further digital twins of field devices of the same or comparable type, which each also have an AI system, the AI system being based on the experience of the AI systems of the other digital twins learns. This makes the Kl algorithm more precise, since the wealth of experience is expanded.

In an advantageous embodiment of the method according to the invention, it is provided that the AI system determines the viewing angle of the photo on the field device, applies it accordingly to the three-dimensional model and identifies the components and the environmental components in the photo using the three-dimensional model. The advantage here is that the photo can be taken from almost any perspective. This makes it easier to use the method, since no predefined perspective has to be taken into account. According to an advantageous embodiment of the method according to the invention, it is provided that the AI system classifies the identified change into a degree of severity, which is communicated to the user in the course of the notification. The degree of severity can also be learned using training data. In particular, it is envisaged that the severity will be classified in device status, which will be communicated to the user. The device status advantageously corresponds to the Namur recommendation (in particular NE 107).

In an advantageous embodiment of the method according to the invention, it is provided that the AI system determines one or more possible causes for the identified change, the one or more causes being communicated to the user in the course of the notification.

According to an advantageous embodiment of the method according to the invention, it is provided that the one or more measures are proposed to the user in the course of the notification.

The AI algorithm can also be taught in advance using training data with regard to the causes and measures. Here, too, it is advantageous if the user gives the KI algorithm feedback as to whether the causes and proposed measures applied or were successful in eliminating the problem, so that the KI algorithm becomes more precise in this respect as well.

Provision can be made for the various detection steps (is there a change; which cause is responsible for this; which measure can remedy the situation?) to be processed by the same AI algorithm, or for several AI algorithms to be interconnected to form a system.

An advantageous embodiment of the method according to the invention provides that the three-dimensional model is created using at least two photos of the field device from different angles. For example, the three-dimensional model is already created during the manufacture of the field device (typically then without environmental components), or when commissioning the field device (typically then with the environmental components).

In an advantageous embodiment of the method according to the invention, it is provided that if no change was identified, the photograph taken is used to refine the three-dimensional model. Refinement means that the model becomes more precise, for example, when new perspectives of the field device are recorded, close-ups are created, more details are visible, etc. This also keeps the target state of the field device up to date.

According to an advantageous embodiment of the method according to the invention, it is provided that the field device is used in an industrial plant, with at least one surveillance camera being provided in the plant, and with the surveillance camera capturing the photo once, regularly or at predetermined times and sending it directly or indirectly to the Cloud application submitted. The advantage here is that in particular no additional components have to be provided since the surveillance camera is already in the system. The viewing angle of the field device typically remains the same over time, making comparison with previous photos easier.

An advantageous embodiment of the method according to the invention provides that an autonomous drone is provided which has a camera unit and the drone navigates to the field device once, regularly or at predetermined times, captures the photo using the camera unit and transmits it directly or indirectly to the cloud application . The drone is, for example, an airworthy drone (gyrocopter, etc.) or an unmanned, automated vehicle. This can be programmed so that it navigates through the system along a predetermined path and advantageously captures photos of all field devices located in the system at the same time.

Provision can also be made for a service technician who is in the vicinity of the field device, routinely takes a photo of the field device with a mobile operating unit, e.g. a smartphone or a tablet, and transmits this manually to the cloud-based platform.

The invention is explained in more detail with reference to the figure below. It shows

1: an exemplary embodiment of the method according to the invention.

A field device FG installed in a measuring point of a process engineering plant is shown in FIG. 1 . In the present case, the field device FG is a fill level measuring device, which is designed to detect the fill level in a container. Put simply, the field device consists of two externally visible components KO1, KO2. The first component KO1 is the housing of the free device FG, in which the measuring and operating electronics are located and on which connections to a fieldbus or a communication loop are located, for example. The second component KO2 is the sensor element, designed in the present case for the transmission and reception of microwaves. If the field device FG is a battery-operated device, sometimes there are no externally visible electronic connections.

In practice, the field device FG has a large number of other components, for example a display/operating element, etc. For the sake of simplicity, however, this exemplary embodiment is limited to the two components mentioned. Of course, the method according to the invention can be carried out with any type of field device.

The field device FG is in direct or indirect contact with one or more environmental components UK. In the present case, the environmental component UK shown is the tank to which the field device FG is attached. There is a digital twin TW for the field device FG, which is implemented on a cloud application CL. The digital twin ZW is a digital image of the field device FG and has the same properties as the field device FG itself, ie in particular it has the same configuration and the same parameterization.

A visual inspection of the field device FG should be carried out at regular intervals in order to check its functionality and to be able to prevent any errors that may arise at an early stage.

A surveillance camera KA is provided in the system for this purpose. This is normally used for security checks in the system, but also detects the field device FG in a sub-area of its field of view. It is now provided that the surveillance camera KA captures a photo and transmits this to the cloud application CL via the Internet, for example via mobile radio.

In addition to the properties of the field device FG described above, the digital twin TW of the field device FG contains a three-dimensional model of the field device FG. This was created, for example, when the field device FG was started up. For this purpose, two or more photos of the field device FG were taken from different viewing angles or perspectives. The three-dimensional model was then created from these photos, for example by the cloud application CL or by a program linked to the cloud application CL.

Furthermore, a Kl algorithm is implemented on the cloud application. The K1 algorithm compares the newly taken photo FO and compares it with the three-dimensional model to determine whether changes have occurred. To improve the detection, the AI algorithm can be taught in advance using training data from field devices of the same type or the same application.

In the present case, the K1 algorithm detects that the field device FG is slightly inclined. Such a slant can lead to a situation that deviates from the ideal Envelope lead, whereby false conclusions about the level of the level in the tank could be drawn.

The cloud application CL then creates a notification NO, which is transmitted to a PC of the user BN. The notification NO contains the detected change and asks the user to check it. The user is now asked to check whether it is an actual change. Since the extent of the skew is only slight but clear, the user BN confirms that it is an actual change. This feedback FB is in turn communicated to the AI algorithm, so that it can adapt its decision-making process to it in the future and learn accordingly.

The user is now informed about the change and can remedy the error accordingly. Optionally, the K1 algorithm K1 can communicate a possible cause for the change in the notification and suggest a measure to rectify the error. In the present case, the error could be in the flange attachment of the field device FG to the tank, which could have formed, for example, in the course of maintenance by a service technician or over time as a result of shocks or vibrations. As a possible measure, it is suggested to check the flange connection and adjust it if necessary. A statement about the correctness of the cause or the suggestion of a correct measure can also be communicated to the AI algorithm as feedback FB, so that it can be continuously improved.

To further increase efficiency, the K1 algorithm can draw on the wealth of experience of one or more other K1 algorithms K1' for its decision-making. Such a further K1 algorithm K1' is implemented on a further cloud application CL' and assigned to a further digital twin TW. The digital twin belongs to a field device FG' of the same type, which is used in a similar application to the field device FG described.

All access and transfers between cloud applications and physical Components, such as the surveillance camera and PC of the user BN, are made in the sense of this application, in particular via the Internet.

As an alternative to the surveillance camera KA, the photos FO can also be captured using a drone capable of flying or land, which navigates through the system at regular intervals or as required, or using a smartphone or tablet of a service technician and transmitted to the cloud application.

Claims

Claims Method for the automated inspection of a field device (FG), wherein the field device (FG) consists of a large number of components (KO1, KO2), in particular a sensor or actuator unit, a housing, a display/operating unit, at least one electronic unit, at least an electronic connection and/or cabling, and wherein the field device (FG) is surrounded by one or more environmental components (UK) or is connected to the at least one environmental component (UK), comprising:

- Capturing a photo (FO) of the field device (FG);

- Transmission of the photo (FO) to a cloud application (CL), a digital twin (TW) of the field device (FG) being stored on the cloud application (CL), the digital twin (TW) being a three-dimensional model of the field device (FG) contains;

- Comparing the photo (FO) with the three-dimensional model and identifying changes in components (KO1, KO2) of the field device (FG) between the photo (FO) and the three-dimensional model; and

- Notifying a user (BN) through the cloud application (CL) in the event of the presence of at least one identified change in at least one of the components (KO1, KO2) of the field device (FG) or in at least one of the environmental components (UK). The method of claim 1, wherein the cloud application (CL) or the digital twin (TW) via a Kl-system (Kl), in particular based on at least one neural network and / or at least one machine learning algorithm and / or at least one deep -Learning algorithm has which Kl system (Kl) performs the steps of comparing the photo (FO) with the three-dimensional model and identifying the at least one change. The method of claim 2, wherein the cloud application (CL) requires feedback (FB) from the user (BN) regarding the identified change, wherein the User (BN) for the feedback (FB) confirms or denies that the identified change is an actual change, with the feedback (FB) being fed to the KI system (Kl), the KI system (Kl ) learns from each feedback (FB) from the user (BN). Method according to claim 2 or 3, wherein the Kl system (Kl) accesses one or more further digital twins (TW) of field devices of the same or comparable type, which each also have their own Kl system (Kl'), the Kl system (Kl) from experiences of the Kl systems (Kl') of the other digital twins (TW) learns. Method according to one of claims 2 to 4, wherein the AI system (Kl) determines the viewing angle of the photo (FO) on the field device (FG), applies it accordingly to the three-dimensional model and the components (KO1, KO2) and the environmental components ( UK) identified in the photo (FO) using the three-dimensional model. Method according to one of Claims 2 to 5, in which the AI system (K1) classifies the identified change into a degree of severity which is communicated to the user (BN) in the course of the notification. Method according to at least one of claims 2 to 6, wherein the AI system (KI) determines one or more possible causes for the identified change, the one or more causes being communicated to the user (BN) in the course of the notification. Method according to Claim 7, in which the AI system (K1) determines one or more measures for remedying the determined cause, the one or more measures being proposed to the user (BN) in the course of the notification. 15

9. The method according to at least one of the preceding claims, wherein the three-dimensional model is created using at least two photos from different angles of the field device (FG).

10. The method according to at least one of the preceding claims, wherein in the event that no change was identified, the recorded photo (FO) is used to refine the three-dimensional model.

11 . Method according to at least one of the preceding claims, wherein the field device (FG) is used in an industrial plant, wherein at least one surveillance camera (KA) is provided in the plant, and the surveillance camera captures the photo once (KA), regularly or at predetermined times (FO) recorded and transmitted directly or indirectly to the cloud application (CL).

12. The method according to at least one of the preceding claims, wherein an autonomous drone is provided which has a camera unit and wherein the drone navigates to the field device (FG) once, regularly or at specified times, captures the photo (FO) using the camera unit and transmitted directly or indirectly to the cloud application (CL).