Sensor fusion for classifying technical objects

Abstract

The invention proposes a system comprising a first sensor (5), preferably a second sensor, and an evaluation device (7), the system being designed to use the first sensor (5) to determine a first value of a first physical variable (4, 9, 11) in connection with a technical object (1) at a first location (A), the evaluation device (7) being designed to use the first value to determine a state of the technical object (1). The system is characterized in that it is designed to use the first sensor (5) or the second sensor at a second location (B, C) different from the first location (A) to determine a second value of the first physical variable (4, 9, 11) or of a second physical variable different from the first physical variable (4, 9, 11) in connection with the technical object (1), the evaluation device (7) being designed to additionally use the second value to determine the state of the technical object (1).

Description

[0001] Description

[0002] Sensor fusion for classifying technical objects

[0003] The invention relates to a method for determining the state of a technical object. Furthermore, the invention relates to a system comprising a first sensor, a second sensor, and an evaluation unit.

[0004] To monitor the condition of a technical object or asset, particularly using autonomous mobile robots, the asset can be detected by sensors, for example, using a directional microphone. The asset's condition can then be determined using a machine learning (ML) model. However, reliably distinguishing between similar states is often challenging. For instance, the vibrations caused by a centrifugal pump blocked on either the pressure or suction side are similar to those produced during normal operation, provided the centrifugal pump operates at a low head.

[0005] Another difficulty in determining states is to reliably distinguish unknown states, which were not considered in the training of the ML model, from known states.

[0006] For the classification or determination of a specific state of technical objects, so-called "Supervised Machine Learning" methods such as "Random Forest", "Gradient Boosting Machine" or neural networks are known.

[0007] However, all methods have the disadvantage that states with similar physical effects are difficult or impossible to distinguish.

[0008] In DE 10 335 126 A1, a device for monitoring the condition of an object that can assume a closed state and an open state is disclosed.

[0009] DE 11 2020 005 513 T2 discloses a valve state detection system and a valve state detection method.

[0010] The invention is based on the objective of providing a method for determining the state of a technical object that exhibits significantly higher reliability. This objective is achieved by a method for determining the state of a technical object with the features of claim 1. Furthermore, the objective is achieved by a system according to claim 8. Advantageous embodiments are described in the dependent claims.

[0011] The inventive method for determining the state of a technical object, which is arranged in a technical plant, in particular a process or manufacturing plant, in which a first value of a first physical quantity is determined by means of a first sensor in connection with the technical object at a first location, wherein the first value is used for determining the state of the technical object, is characterized in that a second value of the first physical quantity or of a second physical quantity different from the first physical quantity is determined by means of the first sensor or a second sensor at a second location different from the first location in connection with the technical object, wherein the second value is additionally used for determining the state of the technical object.

[0012] The technical installation can be a plant from the process industry, such as a chemical, pharmaceutical, petrochemical, or food and beverage industry plant. This also includes any plant from the manufacturing industry, such as factories where cars or goods of all kinds are produced. Technical installations suitable for carrying out the process according to the invention can also originate from the field of energy generation. Wind turbines, solar power plants, or power plants for energy generation are likewise included in the term "technical installation."

[0013] The technical objects can include, for example, compressors, tanks, motors, valves, actuators, sensors, burners, or robots. Preferably, the technical object is a pump, a motor, a compressor, or a valve.

[0014] The state of a technical object is generally determined by the totality of the values ​​of all its attributes. In the inventive method, a first sensor determines a first value of a physical quantity. This determined value is used to determine the state of the technical object. If, for example, the physical quantity is the pressure inside a technical object designed as a boiler, and the value is 10 bar, then a "normal" state of the boiler can be determined by the fact that the value of 10 bar is below a certain threshold, e.g., 20 bar. However, under certain circumstances, the single determined value may not be sufficient to unambiguously determine the state of the boiler, since other possible states would also involve a pressure inside the boiler below the threshold value.

[0015] According to the invention, a second value of the physical quantity is therefore determined at a second location using the first sensor or a further, second sensor. It is essential that the second location differs from the first location. The term "different" here means that the second measurement is recognizable to a person skilled in the art and is deliberately carried out at a location distant from the first location. The second location can be at least as far from the first location as the largest spatial extent of the technical object.

[0016] Alternatively, a value of a second physical quantity, different from the first, can be determined. In the example above, in addition to the pressure inside the boiler, a flow rate (e.g., in liters / minute) through a boiler outlet could be determined. This flow rate measurement takes place at a (second) location, which is obviously remote from the (first) location of the pressure measurement. Both sensor measurements are performed "in connection with the technical object." They therefore relate to parameters that can be used to characterize the technical object or its behavior.

[0017] The second value is also used to determine the state of the technical object. By expanding the determination of the parameters relevant to the technical object, the individual possible states of the technical object can be more easily distinguished from one another, thus enabling a more reliable determination of the actual, current state of the technical object.

[0018] The first and / or second sensor can be mounted on an autonomous, mobile robot designed to move to the first and second locations. In this way, the sensor(s) can be easily and automatically moved to the locations where the values ​​of the physical quantities are to be determined. Especially in more complex technical systems, a robot can reach difficult-to-access or hazardous locations more efficiently than a human operator. Furthermore, using such an autonomous, mobile robot eliminates the need for a static sensor setup. Instead, the robot enables flexible measurements at varying locations without requiring a large number of sensors.

[0019] The robot's navigation route can be manually specified by an operator or, more generally, by personnel of the technical facility. Preferably, however, the first and second locations are automatically determined, taking into account the topology of the technical facility, particularly a piping and instrumentation diagram (PID), and transmitted to the robot for navigation. From the PID, for example, using isometric drawings or 3D plans, the measurement locations can be automatically determined, converted into the robot's coordinate system, and transmitted to the robot as inspection points.

[0020] If, for example, a pump is to be monitored as a technical object, the (first) location of the pump, as well as all valves, all orifices in pipelines connected to the pump and changes in their pipe diameters, can be determined from the planning documents, which are suitable as the (second) location in the sense of the present invention, since valves, orifices and changes in pipe diameter can lead to cavitation or flow separation.

[0021] In a preferred embodiment of the invention, a first state component is determined using a first machine learning model and the first value, and a second state component is determined using a second machine learning model and the second value. The first and second state components are then used to determine the state of the technical object. In other words, the first measured value (from the first location) is fed into a first model, and the first state component is determined, while the second measured value (from the second location) is fed into a second model, and the second state component is determined. By combining the two (partial) results from the two machine learning models, the determination of the (overall) state of the technical object can be made more robust and reliable.

[0022] Alternatively, the two measurements can be combined and fed into a single machine learning model, which ultimately determines the (overall) state of the technical object. In other words, the state of the technical object can be determined using a machine learning model based on the first and second values. This may reduce the modeling effort.

[0023] Preferably, the spatial relationship between the two locations is taken into account when determining the state of the technical object. For example, in the case of measuring acoustic vibrations, background noise can be identified and excluded from the determination of the object's state. This further improves the accuracy of the state determination.

[0024] Information regarding the condition of the technical object is preferably transmitted to a control system of the technical plant. There, it serves, for example, to operate and monitor the technical plant and can trigger maintenance measures such as the replacement of the technical object.

[0025] The previously formulated task is also solved by a system comprising a first sensor, a second sensor and an evaluation device, wherein the system is configured to determine a first value of a first physical quantity by means of the first sensor in connection with a technical object at a first location, wherein the evaluation device is configured to use the first value for determining a state of the technical object.The system is characterized in that it is designed to determine a second value of the first physical quantity or of a second physical quantity different from the first physical quantity by means of the first sensor or a second sensor at a second location different from the first location in connection with the technical object, wherein the evaluation device is designed to additionally use the second value for determining the state of the technical object.

[0026] The system can include an autonomous, mobile robot comprising the first and second sensors. This robot is designed to establish a communication link with the evaluation unit to transmit the measured first and second values ​​to the evaluation unit for determining the state of the technical object. The actual state determination thus takes place on a separate, external evaluation unit, for example, in a cloud-based environment. Alternatively, the robot can also include the evaluation unit itself, in which case the robot can also perform the state determination.

[0027] The system is preferably designed to transmit the determined state to a control system of the technical plant for further processing of the state information.

[0028] The properties, features and advantages of this invention described above, as well as the manner in which they are achieved, will become clearer and more easily understood in connection with the following description of exemplary embodiments, which are explained in more detail in conjunction with the drawings.

[0029] The figure depicts a technical object, designed as a pump 1 of a process plant. A pipe 2 leads into the pump 1 and a pipe 3 leads out. The pump 1 is designed, in a manner known per se, to draw a fluid from the pipe 2 and pump it through the pipe 3. During this process, noise 4 is generated in the pump 1, primarily by a motor and a bearing. This noise 4 is detected by a noise sensor 5 of an autonomous, mobile robot 6 as a first value of a first physical quantity (oscillation amplitude, oscillation frequency), with the robot 6 being located at a first location A during the measurement.

[0030] In the present embodiment, the autonomous, mobile robot 6 has an evaluation unit 7 in which a first machine learning model is implemented. The first machine learning model is configured to determine the state of the pump based on the recorded sounds 4. The first model distinguishes between the two states "normal speed" and "deviant speed". The first model thus determines a first state component of the overall state of the technical object 1, which is designed as a pump.

[0031] Further pump states regarding valve position or potential cavitation cannot be reliably determined solely using the first measurement and the first model. Therefore, the autonomous, mobile robot 6 moves to a second location B, which is located away from the first location A. There, the noise sensor 5 detects noises 9 emanating from a section 8 of the outgoing pipe 3 as a second measurement of the first physical quantity. This second measurement is fed into a second machine learning model, which uses it to determine another state of pump 1: The second model distinguishes between the two states "cavitation" and "no cavitation." Thus, the second model determines a second component of the overall state of pump 1.

[0032] The autonomous, mobile robot 6 then moves to a third location C, which is located away from the first location A and the second location B. There, the noise sensor 5 detects noises 11 emanating from another section 10 of the outgoing pipe 3 as a third measurement of the first physical quantity. This third measurement is fed into a third machine learning model, which uses it to determine another state of the pump 1: The third model distinguishes between the two states "blockage" and "no blockage". Thus, the third model determines a third component of the overall state of the pump 1.

[0033] The autonomous, mobile robot 6 determines the location of places A, B and C using a piping and instrumentation diagram of the process plant, or this information is automatically transmitted to it - for example by a control system of the process plant.

[0034] The following table lists possible combinations of individual state components at the three locations A, B, C.

[0035] In the first row of the table, a logical 0 at location A indicates that the state is "normal rotational speed". A logical 1 at location B indicates that the state is "no cavitation". A logical 1 at location C indicates that the state is "no blockage". The overall state of pump 1, determined by the evaluation unit 7 of the autonomous, mobile robot 6, is "normal".

[0036] In the second row of the table, a logical 1 at location A indicates that the state "speed deviation" exists. A logical 0 at location B indicates that the state "cavitation" exists. A logical 1 at location C indicates that the state "no blockage" exists. The overall state of pump 1, determined by the evaluation unit 7 of the autonomous, mobile robot 6, is "cavitation".

[0037] In the third row of the table, a logical 1 at location A indicates that the state "Speed ​​Deviates." A logical 1 at location B indicates that the state "No Cavitation" is present. A logical 0 at location C indicates that the state "Blockage" is present. The overall state of pump 1, determined by the evaluation unit 7 of the autonomous, mobile robot 6, is "Blockage." In the fourth row of the table, a logical 1 at location A indicates that the state "Speed ​​Deviates." A logical 1 at location B indicates that the state "No Cavitation" is present. A logical 1 at location C indicates that the state "No Blockage" is present. The overall state of pump 1, determined by the evaluation unit 7 of the autonomous, mobile robot 6, is "Excessive Speed, No Cavitation."

[0038] Although the invention has been illustrated and described in detail by the preferred embodiment, the invention is not limited by the disclosed examples and other variations can be derived by the person skilled in the art without leaving the scope of protection of the invention.

Claims

Patent claims 1. Method for determining the state of a technical object (1) arranged in a technical plant, in particular a process or manufacturing plant, in which a first value of a first physical quantity (4, 9, 11) is determined by means of a first sensor (5) in connection with the technical object (1) at a first location (A), wherein the first value is used for determining the state of the technical object (1), characterized in that a second value of the first physical quantity (4, 9, 11) or of a second physical quantity different from the first physical quantity (4, 9, 11) is determined by means of the first sensor (5) or a second sensor at a second location (B, C) different from the first location (A) in connection with the technical object (1), wherein the second value is additionally used for determining the state of the technical object (1).

2. Method according to claim 1, wherein the first sensor (5) and / or the second sensor are arranged on an autonomous, mobile robot (6) which is configured to move to the first location (A) and the second location (B, C).

3. Method according to claim 2, wherein the first location (A) and the second location (B, C) are automatically determined taking into account a topology of the technical plant, in particular a piping and instrumentation diagram, and transmitted to the autonomous mobile robot (6) for navigation.

4. Method according to one of the preceding claims, wherein a first state part is determined using a first machine learning model and the first value, and a second state part is determined using a second machine learning model and the second value, wherein the first state part and the second state part are used to determine the state of the technical object (1).

5. Method according to one of claims 1 to 3, wherein the state of the technical object (1) is determined by means of a machine learning model using the first value and the second value.

6. Method according to one of the preceding claims, wherein a spatial location of the two places (A, B, C) relative to each other is taken into account when determining the state of the technical object (1).

7. Method according to any of the preceding claims, wherein the technical object (1) is a pump, a motor, a compressor or a valve.

8. System comprising a first sensor (5), preferably a second sensor, and an evaluation device (7), wherein the system is configured to determine a first value of a first physical quantity (4, 9, 11) by means of the first sensor (5) in connection with a technical object (1) at a first location (A), wherein the evaluation device (7) is configured to use the first value for determining a state of the technical object (1), characterized in that the system is configured to determine a second value of the first physical quantity (4, 9, 11) or a second physical quantity different from the first physical quantity (4, 9, 11) by means of the first sensor (5) or the second sensor at a second location (B, C) different from the first location (A) in connection with the technical object (1), wherein the evaluation device (7) is configured toto use the second value additionally for determining the condition of the technical object (1).

9. System according to claim 8, comprising an autonomous mobile robot (6) which includes the first sensor (5) and the second sensor, wherein the autonomous mobile robot (6) is configured to establish a communication link to the evaluation unit (7) in order to transmit the determined first and second values ​​to the evaluation unit (7) for determining the state of the technical object (1).

10. System according to claim 9, wherein the autonomous mobile robot (6) comprises the evaluation device (7).

Images