Method for diagnosing a situation in a process automation plant

Abstract

The invention relates to a method for diagnosing a situation in a process automation system (1). A field device (2) is installed in the system (1), and data about the system (1) is provided. Based on this data and starting from a prototype model assigned to the field device (2), a system model is generated using artificial intelligence. Based on a measured value generated by the field device (2), a statement about a situation in the system (1) is generated using the system model.

Description

[0001] The invention relates to a method for diagnosing a situation in a process automation system.

[0002] In modern technology, it is common practice to monitor and control processes in plants using field devices. These field devices are therefore designed as either sensors or actuators. The sensors detect various measurement signals, which are either relevant process variables themselves or allow for the derivation of a value from the process variable. These include, for example, fill level, distance, pressure, pH value, and temperature. From these primary process variables, further process information can potentially be inferred. This includes, for example, foam or buildup detection, whether an agitator is moving, or whether a container is being filled or emptied. Statements about such situations, which can be derived from the primary or actual measurement or process values, are classically generated, for example, by windowing with limit values ​​or by evaluating the temporal progression of the measured values.However, this usually requires that the field devices used are adapted to the specific application. For example, suitable measurement or evaluation parameters must be found or entered. This is very time-consuming and prone to errors.

[0003] The object underlying the invention is therefore to propose a method for generating the most accurate and application-specific diagnostic messages possible from measured or determined process values ​​of a process plant.

[0004] The invention solves the problem by a method for diagnosing a situation in a process automation system, wherein the method comprises at least the following steps: installing at least one field device in the system; providing data about the system; generating a system model based on the data about the system and starting from a prototype model assigned to the field device using artificial intelligence; generating at least one measured value from the field device; and generating a statement about a situation in the system based on the at least one generated measured value using the system model.

[0005] According to the invention, diagnosis is enabled by adapting a prototype model, associated with the field device, to the existing measurement situation in the process plant using artificial intelligence. The prototype model includes, for example, measurement parameters, value ranges, or relationships between measured values. The prototype model makes it possible, for example, to recognize from specific measured values ​​whether foam is present or whether, for example, a container is being emptied. However, the prototype model may be applicable to a broad range of situations or may only include those situations that can be identified based on measured values ​​from the field device itself. The plant model contains the data for the specific case and can, for example, enable higher precision in diagnosis. Alternatively or additionally, the plant model can also include the processing of measurement data from additional sensors.Alternatively, diagnostic functions not normally included in the field device's capabilities can be implemented. Depending on the configuration, the data required to generate the plant model is entered directly by a user on the field device itself. However, it can also be provided by a control center or a digital twin.

[0006] The process unfolds as follows: The field device is installed. Starting with a prototype model and in conjunction with data assigned to the plant, an artificial intelligence generates a plant model. This plant model is preferably tailored to the field device, enabling a reliable and precise diagnosis of a plant's situation or condition based on the field device's measurements. Overall, this increases diagnostic precision and allows for automatic application-specific adaptation.

[0007] One embodiment involves storing the plant model—specifically, only the plant model—within the field device. In this embodiment, preferably only the field device has access to the plant model. For this purpose, the plant model is stored, for example, within the field device itself and is therefore only possessed by the field device. Consequently, in one embodiment, the field device performs the diagnostics using the plant model itself. Alternatively, it is envisaged that the field device provides the plant model to another unit for diagnostic purposes.

[0008] One implementation involves evaluating the generated statement and then using artificial intelligence to adjust the investment model based on this evaluation. In this implementation, the investment model is improved through a form of feedback. The evaluation might consist of commenting on whether the statement is true or false. Alternatively or additionally, the statement is modified to better reflect the actual situation. This evaluation is then used to further refine the investment model.

[0009] The following details relate to which unit the investment model generates.

[0010] One design involves the field device generating the plant model.

[0011] One design provides that the plant model is generated by an edge device that is separate from the field device and located in the area of ​​the plant.

[0012] One design involves generating the investment model through the use of a cloud.

[0013] The following two configurations each utilize a digital twin.

[0014] One design envisages that the plant model is generated using a digital twin of the field device.

[0015] A supplementary or alternative design involves creating the plant model using a digital twin of the plant.

[0016] Another issue is which unit the plant model uses and performs the diagnosis in conjunction with at least one measured value from the field device.

[0017] In one configuration, this could be a control room.

[0018] In an alternative configuration, the diagnosis is carried out by the field device and the generated diagnostic message is transmitted, for example, to a control center.

[0019] An alternative or supplementary configuration provides that the field device generates the information about the situation in the plant. In a further configuration, this implies that only the field device has access to the generated plant model, at least after it has been generated.

[0020] The following details deal with the models used. This concerns the datasets, the underlying relationships, or the models generated by artificial intelligence that capture or describe dependencies.

[0021] One design involves optimizing the plant model and / or the prototype model by considering the capabilities relevant to generating the plant model—particularly storage space and / or computing power—of the device that will generate it. This design takes into account that different performance parameters—such as data storage or computing power—may be available for generating the plant model. Therefore, the plant model and / or the prototype model is adapted accordingly, for example, by scaling it down or up. Adapting the prototype model can be advantageous to ensure that it is not more demanding than the plant model.

[0022] One embodiment provides that the plant model is designed taking into account the capabilities relevant for generating the statement about the situation in the plant – in particular, storage space and / or computing power – of a device that performs the generation of the statement about the situation in the plant. This embodiment considers the unit that is to use the plant model for diagnostic purposes.

[0023] One embodiment includes the generation of at least one measured value by the field device relating to a fill level, a temperature, a pressure, a pH value or a flow rate.

[0024] One design provides that the statement about the situation refers to whether foam, build-up or corrosion is present in an area of ​​the field device.

[0025] Furthermore, the invention relates to a device configured to perform the method according to at least one of the aforementioned or following embodiments. Therefore, the embodiments and explanations also apply accordingly to the device. The device is, for example, a field device.

[0026] The invention is explained in more detail with reference to the following figure. It shows: Fig. 1: A schematic representation of a process automation system.

[0027] The Fig. Figure 1 shows a schematic system 1 of process automation in which a tank or the contents of the tank are monitored with three different field devices 2, 2', 2".

[0028] Each field device 2, 2', 2" is installed with an associated prototype model that describes the field device 2, 2', 2", its measurement capabilities, and related diagnostic options with regard to situations within the system. A system model is then generated from this prototype model in conjunction with data describing the system. The system model is thus the prototype model adapted to the application of the field device 2, 2', 2". The system data describes, for example, parameters of the environment, the processes taking place, any situations that may occur, the characterization of these situations, etc.

[0029] The investment model is generated using artificial intelligence. This includes machine learning and large language models.

[0030] The three field devices 2, 2', 2" differ from each other in which unit the plant model is determined.

[0031] A field device 2 is capable, with regard to its data storage and computing power, of generating the plant model itself.

[0032] The two other field devices 2', 2" each depend on other units 3, 4. This is indicated by the double arrows. For example, one field device 2' accesses an edge device 3, which is located in the area or vicinity of the plant 2. In one embodiment, the edge device 3 can also access a cloud. The third field device 2" has access to a cloud 4, in which, in this embodiment, a digital twin of the field device 2" and the plant 1 is used.

[0033] If the plant model is available, the field devices 2, 2', and 2" each acquire measurement data and, based on the plant model, determine, for example, whether a specific situation exists. This could be, for instance, the detection of foam. This diagnosis is then transmitted, for example, to a control room (not shown here) for further processing. In one variant, the affected diagnosis is evaluated by the users. This evaluation is then used to improve the plant model.

Claims

[1] Method for diagnosing a situation in a plant (1) of process automation, the procedure includes at least the following steps, that at least one field device (2, 2', 2") is installed in the plant (1), that data about the plant (1) is made available, that, based on the data about the plant (1) and starting from a prototype model assigned to the field device (2, 2', 2"), a plant model is generated using artificial intelligence, that at least one measured value is generated by the field device (2, 2', 2"), and that, starting from the at least one generated measurement value, a statement about a situation in the plant (1) is generated using the plant model. [2] Method according to claim 1, wherein the plant model is stored - in particular only - in the field device (2, 2', 2"). [3] Method according to claim 1 or 2, where the generated statement is evaluated, and the investment model adapts based on the evaluated statement using artificial intelligence. [4] Method according to any one of claims 1 to 3, wherein the plant model is generated by the field device (2). [5] Method according to any one of claims 1 to 3, wherein the plant model is generated by an edge device (3) that is separate from the field device (2') and is located in the area of ​​the plant (1). [6] Method according to any one of claims 1 to 3, wherein the investment model is generated by applying a cloud (4). [7] Method according to any one of claims 1 to 6, wherein the plant model is generated using a digital twin of the field device (2, 2', 2"). [8] Method according to any one of claims 1 to 7, wherein the plant model is generated using a digital twin of the plant (1). [9] Method according to any one of claims 1 to 8, wherein the field device (2, 2', 2") generates the statement about the situation in the plant (1). [10] Method according to any one of claims 1 to 9, wherein the plant model and / or the prototype is designed taking into account capabilities relevant for the generation of the plant model - in particular storage space and / or computing power - of a device (2, 3, 4) performing the generation of the plant model. [11] Method according to any one of claims 1 to 10, wherein the plant model is designed taking into account capabilities relevant for generating the statement about the situation in the plant (1) - in particular storage space and / or computing power - of a device (2, 2', 2") performing the generation of the statement about the situation in the plant (1). [12] Method according to any one of claims 1 to 11, wherein the field device (2, 2', 2") generates at least one measured value relating to a fill level, a temperature, a pressure, a pH value or a flow rate. [13] Method according to any one of claims 1 to 12, wherein the statement about the situation refers to whether foam, build-up or corrosion is present in an area of ​​the field device (2, 2', 2").

Images