Method and system for detecting anomalies in a robotic system

Abstract

The present invention relates to a method and system for detecting anomalies in a robotic system in an industrial plant. The robotic system is associated with a computing system configured to detect anomalies in the robotic system. The computer system monitors configuration parameters of the robotic system and process parameters associated with the robotic system. In addition, the computing system detects an association between at least one configuration parameter and at least one process parameter for obtaining optimal configuration parameters and optimal process parameters. The optimal configuration parameters and optimal process parameters are analyzed for detecting anomalies. At least one parameter among the configuration parameters and process parameters that causes an anomaly is identified. Thereafter, the detected anomaly is verified, an effective setpoint is estimated, and the estimated effective setpoint is updated in an analytics model. The updated analytics model is then used to accurately detect anomalies.

Description

Technical Field

[0001] The present invention generally relates to industrial robots and, more particularly, to efficiently detecting anomalies in a robotic system in an industrial plant. Background Art

[0002] Industrial robots are widely used in various applications, such as automotive painting industries to supply chain management. A robotic system can include a robotic arm, a controller, and a computer. Generally, a robotic painting system is associated with process equipment (e.g., a pump for supplying paint to the robotic arm) for performing operations (painting a vehicle, picking and placing an object, etc.).

[0003] In a typical industrial plant, the process equipment and the robotic system are monitored by an operator in the industrial plant, and specific parameters that may cause anomalies are marked. Additionally, if an operator detects a parameter that causes an unusual effect during operation, such a parameter is also marked. Models are used to analyze the marked parameters and detect anomalies. However, such models are built for specific parameters and cannot be used with existing models when new parameters are used. Moreover, manually generating models for different parameters is a tedious task. Additionally, a plant operator may not know the function of the model, so the plant operator cannot precisely select the parameters that must be uploaded for analysis.

[0004] Generally, analytics models are implemented in a server local to the industrial plant or in a remote server (cloud server). Typically, an analytics model requires raw data for monitoring and analysis. Usually, a plant operator selects all the raw data for uploading to the server. Therefore, a huge data storage device is required to store the raw data, thereby increasing the cost of operating the industrial plant.

[0005] Typically, a plant operator may request an analytics model to be updated due to a change in the parameters in the industrial plant. However, a third-party vendor may not be available to update the model. Moreover, a huge cost is associated with updating the model.

[0006] In view of the above, it is necessary to address at least one of the above limitations and propose methods and systems for overcoming the above problems. Summary of the Invention

[0007] In an embodiment, the present invention relates to a method and a system for detecting anomalies in a robotic system in an industrial plant. The robotic system may include at least one robot and one or more controllers for controlling the at least one robot to perform operations on an object. The one or more controllers may be part of a distributed control system (DCS) configured in the industrial plant. The DCS may include one or more sensors for measuring configuration parameters of the at least one robot and process parameters associated with the robotic system. The DCS further includes a database for storing the measured configuration parameters and process parameters. The robotic system is associated with a computing system configured to detect anomalies in the robotic system. The computer system monitors the configuration parameters and process parameters of the at least one robot. In an embodiment, the configuration parameters and process parameters are obtained from the one or more sensors. Further, the computing system detects an association between at least one configuration parameter and at least one process parameter to obtain optimal configuration parameters and optimal process parameters. The optimal configuration parameters and optimal process parameters are analyzed by the computing system to detect anomalies. At least one parameter among the configuration parameters and process parameters that causes an anomaly is identified. Thereafter, the detected anomaly is verified by a plant operator. Based on the verification, an error in the analysis is determined. The error indicates that a set point associated with the at least one parameter is invalid. Thereafter, a valid set point is estimated and the estimated valid set point is updated in an analytics model. The updated analytics model is then used to detect anomalies for the at least one parameter.

[0008] In one embodiment, the configuration parameters include data related to the sprayer settings of the at least one robot, the path traversed by the at least one robot, and the size of the sprayer. In one embodiment, the process parameters include data related to the movement pattern of the at least one robot, the size of the object, one or more substances required by the process, and parameters related to the operations to be performed on the object.

[0009] In an embodiment, a computing unit uses machine learning techniques to determine the optimal configuration parameters and optimal process parameters. In an embodiment, the detected anomaly is presented to a plant operator. The plant operator verifies the detected anomaly. In one embodiment, the plant operator verifies the detected anomaly as one of "success" or "unsuccess".

[0010] In one embodiment, for each identified parameter that causes an anomaly, a corresponding analytics model is generated and stored in a memory. When one of the identified parameters is subsequently used, the corresponding analytics model is used to detect anomalies for that parameter. In an embodiment, a single analytics model may be stored and the single analytics model is updated for each identified parameter that causes an anomaly.

[0011] A system with a varying range is described herein. In addition to the aspects and advantages described in this overview, other aspects and advantages will become apparent by referring to the accompanying drawings and to the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The subject matter of the present invention will be explained in more detail hereinafter with reference to the preferred exemplary embodiments shown in the accompanying drawings, in which:

[0013] Figure 1 A simplified block diagram of a robotic system in an industrial plant according to an embodiment of the present disclosure is shown;

[0014] Figure 2A A block diagram showing a conventional arrangement for detecting an abnormality in a robotic system is shown;

[0015] Figure 2B An exemplary block diagram for detecting an abnormality in a robotic system according to an embodiment of the present disclosure is shown; and

[0016] Figure 3 An exemplary flowchart for detecting an abnormality in a robotic system according to an embodiment of the present disclosure is shown; and

[0017] Figure 4 and Figure 5 An exemplary scenario for detecting an abnormality in a robotic system according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0018] The present invention discloses methods and systems for detecting an abnormality in a robotic system.

[0019] Figure 1 A simplified diagram of a painting workshop (100) in an industrial plant is shown. In an embodiment, a process plant may be considered to replace an industrial plant. The above disclosure is described with respect to an industrial plant. However, it should not be construed as limiting. Those of ordinary skill in the art will understand that aspects applicable to a process plant fall within the scope of the present invention. The present invention is described with respect to the painting workshop (100). The present invention is not limited to the painting workshop (100), and the present invention finds application in various industrial plants and process plants. The painting workshop (100) includes a robotic system for automatically painting an object. For example, in the automotive industry, robots are used to paint vehicles. The robots are programmed to spray a specific amount of paint on specific areas of the vehicle. As in Figure 1As seen in, the painting workshop (100) includes a vehicle (101), at least one robot (102), a controller (103), and a computing system (104). The vehicle (101) is placed in a painting line, where at least one robot (102) is configured to spray paint on the vehicle (101). In an exemplary embodiment, at least one robot (102) can be a robotic arm that has a sprayer configured to spray paint.

[0020] The controller (103) is configured to control the robotic arm (102). The controller (103) receives one or more inputs and adjusts its output to operate the robotic arm (102) in an expected manner. The controller (103) may be able to control each part of the robotic arm (102). For example, the controller (103) can operate the robotic arm (102) to move in a specific direction. In another example, the controller (103) can operate the sprayer of the robotic arm (102) ( Figure 1 not shown in) to tilt and spray paint at a specific air pressure.

[0021] The computing system (104) is configured to analyze configuration parameters and process parameters. Configuration parameters are parameters related to the configuration of the robotic arm (102). For example, configuration parameters can include sprayer settings, brush sizes, the path that the robotic arm (102) travels to spray paint on the vehicle (101). Process parameters can include air pressure, pump settings, paint characteristics, vehicle (101) dimensions, robotic arm (102) movement patterns, etc. In one embodiment, the computing system (104) includes an analytics model for studying the characteristics of configuration parameters and process parameters. Moreover, the analytics model is used to detect anomalies in the configuration parameters and process parameters.

[0022] Figure 2A A conventional computing system is shown. As shown, the conventional computing system includes an analytics model, a memory, a monitoring module, an analytics module, and a decision-making module. The monitoring module is configured to monitor configuration parameters and process parameters. All configuration parameters and process parameters are stored in a conventional manner for analysis. Configuration parameters and process parameters are raw data that consume a huge amount of storage space. Therefore, storing such a large amount of data results in high costs and additional resources. In addition, the conventional method has one analytics model built for analyzing configuration parameters and process parameters. The analytics module uses a single analytics model to analyze configuration parameters and process parameters. A single analytics model is built for specific configuration parameters and process parameters. Therefore, if one of the configuration parameters and process parameters is changed, the analytics model cannot be used. The existing method often causes false alarms due to inaccurate anomaly detection. Usually, the factory operator has to reset the set points in the computing system to avoid false alarms.

[0023] Figure 2B A proposed computing system for accurately detecting anomalies in a robotic system is shown. The proposed computing system (104) includes a monitoring module (202), an analysis module (203), a determination module (204), a memory (205), and one or more analytics models (201a, ... 201n).

[0024] In an embodiment, the monitoring module (202) is configured to monitor configuration parameters and process parameters. Additionally, the monitoring module is configured to detect an association between the configuration parameters and the process parameters. For example, if a sprayer is not spraying paint evenly, the monitoring module determines a plausible reason for the uneven spraying of the sprayer. One reason could be due to varying air pressure. Thus, the sprayer as a configuration parameter and the air pressure as a process parameter, an association exists between the two. Such an association is detected by the monitoring module (202), and the optimal configuration parameters and process parameters are obtained. In the example given above, the sprayer data and the air pressure data can be detected as the optimal configuration parameters and process parameters. The optimal configuration parameters and process parameters are stored in the memory (205). In one embodiment, the monitoring module (202) can use correlation analysis to determine the association between the configuration parameters and the process parameters. Similarly, any analysis can be used to detect the association.

[0025] In an embodiment, the analysis module (203) is configured to analyze the optimal configuration parameters and process parameters. The analysis is performed to study the nature and characteristics of the optimal configuration parameters and process parameters. The analysis module (203) can use unsupervised machine learning techniques to perform the analysis. Examples of unsupervised analysis can include correlation analysis, clustering, and dimensionality reduction of data. The nature and characteristics of the optimal configuration parameters and process parameters are used for feature selection. The analysis module utilizes one or more analytics models (201a... 201n) to perform the analysis.

[0026] For example, when analyzing an anomaly in a pump, parameters such as the power of the pump, current consumption, torque output of the pump, and the flow rate of paint from the output of the pump are monitored. The analysis module (203) notifies that the power consumption is correlated with the torque of the pump, and only records any one of these parameters while maintaining the correlation coefficient. The analysis module (203) then only selects the paint flow rate and the power consumption as the main features of the pump for analysis based on the variation of the parameters. If the analysis module (203) notifies a sharp change in the variation of the parameters or the correlation coefficient, then this results in a re-evaluation of the features used for analysis and reports it to the determination module (204). The determination module (204) can then change the model to perform calculations or re-train the model for anomaly detection.

[0027] In an embodiment, each analytics model (201) is constructed for a specific set of configuration parameters and process parameters. In an embodiment, machine learning techniques are used to update the analytics model (201). In an embodiment, the analytics model (201) is a learning model that is autonomously updated based on a learning process. For example, an analytics model (201a) is constructed for high-viscosity paint. The analytics model (201a) includes a setpoint for the viscosity of the paint. In an embodiment, if the paint is replaced with a thinner having a relatively low viscosity, the analytics model (201a) can detect the change in the parameters, update the analytics model (201a), and generate a separate analytics model (201b) for the thinner. The analytics model (201b) includes setpoints related to the thinner. Similarly, the analytics model (201) can generate separate models for each parameter. In an embodiment, existing analytics parameters can be updated with setpoints of new parameters.

[0028] In an embodiment, the determination module (204) is configured to recommend an action to be performed based on an analysis. For example, the recommendation can include: identifying important parameters among the configuration parameters and process parameters for continuous monitoring, based on load prediction, data related to retraining the analytics model (201), more data required for the analysis, etc. In addition, the determination module (204) can recommend archiving the monitored data in the memory (205) for subsequent analysis. The determination module (204) can also influence the change or tuning of the monitoring setup of the application by determining the semantics for triggering the sending of data for further analysis. Machine learning techniques (such as change point or anomaly detection) can be used to configure the triggering semantics. The determination module (204) can also rank the triggers and recommend sampling data based on the relevance of a particular alarm / trigger.

[0029] In an embodiment, the computing system (104) can include, but is not limited to, a server, a supercomputer, a workstation, a laptop computer, or any other electronic device capable of performing the method described below in Figure 3 as set forth.

[0030] Figure 3 A flowchart (300) is shown for detecting anomalies in an industrial plant. In step 301, the computing system (104) monitors the configuration parameters and process parameters of the robotic arm (102). The monitoring module (202) monitors the configuration parameters and process parameters of the robotic arm (102).

[0031] Let us consider a first scenario where a pump stores a first liquid (paint) to be sprayed on a vehicle (101). Refer to Figure 4, showing the painting workshop (100). Let's consider that the pump (401) contains a liquid (paint) with high viscosity. The pump (401) pumps the paint to the robotic arm (101) for spraying the paint on the vehicle (101). The analytics model (201a) is configured to store the viscosity set point related to the paint. The sprayer of the robotic arm (101) sprays the paint at a certain air pressure. Referring back to Figure 3 , in step 302, the monitoring module (202) is configured to detect the association between the configuration parameters and the process parameters, and in step 303, the monitoring module (202) is configured to obtain the optimal configuration parameters (such as sprayer settings) and process parameters (characteristics of the pump (401)) based on the association between the configuration parameters and the process parameters.

[0032] Now let's consider a second scenario where the pump stores a second liquid (thinner) to be sprayed on the vehicle (101). Referring to Figure 5 , showing the painting workshop (100) where the thinner is sprayed on the vehicle (101). In one embodiment, the thinner has a relatively low viscosity compared to the paint. As Figure 5 shown, the pump (501) pumps the thinner to the robotic arm (102) for spraying on the vehicle (101). The sprayer can spray the thinner at a specific air pressure, which can be different from that for spraying the paint. The monitoring module (202) is configured to monitor the configuration parameters and process parameters of the robotic system as Figure 5 shown. As shown in step 302, the monitoring module (202) can detect the association between the air pressure of the sprayer and the characteristics of the pump (501). As shown in step 303, the monitoring module obtains the optimal configuration parameters and process parameters based on the association and stores them in the memory (205).

[0033] In step 304, the analysis module (203) analyzes the optimal configuration parameters and the optimal process parameters to detect anomalies. Considering the first scenario and referring to Figure 4 , the analysis module (203) uses the analytics model (201a) for analysis. Referring back to Figure 3 , in step 304, the analysis module (203) can use the viscosity set point related to the paint to accurately detect anomalies. Viscosity can be considered as at least one parameter causing anomalies.

[0034] Considering the second scenario and referring to Figure 5 , the analysis module (203) uses the analytics model (201a) for analysis. Referring back to Figure 3, as shown in step 304, the analysis module (203) cannot accurately detect anomalies using the viscosity set points related to paint. Additionally, the detected anomalies may cause false alarms because the viscosity set points for thinner may be different from those for paint. Since the analytics model (201a) is specific to paint, using the analytics model (201a) to analyze thinner parameters cannot yield accurate analysis. Therefore, when an anomaly is detected, it is verified by the factory operator. For example, a display unit associated with the computing system (104) can display the anomaly to the factory operator. The factory operator can verify the anomaly by entering one of "success" or "unsuccessful". Those skilled in the art will understand that the verification can be performed in different ways and may not be limited to the techniques described in the present invention. If the operator has verified the anomaly as "unsuccessful", the analysis module (203) detects an error by comparing the output of the first scenario and the output of the second scenario, as shown in Figure 3 step 305. The error can be used to determine that the set points used are invalid. In step 306, valid set points are estimated for the thinner. In one embodiment, historical data related to the thinner can be used to estimate the valid set points. In another embodiment, the factory operator can input the valid set points for the thinner. In an embodiment, the analytics model (201a) can be updated with the valid set points related to the thinner. For subsequent analysis, the analytics model (201a) can be used to analyze parameters related to the thinner. In another embodiment, a new analytics model (201b) related to the thinner can be generated. The new analytics model (201b) can be stored in the memory (205). When the thinner is used in the robotic system, the analytics model (201b) can be used to analyze parameters related to the thinner.

[0035] In an embodiment, machine learning techniques can be used to generate multiple analytics models (201a...201n). In an embodiment, using machine learning techniques to update the analytics model (201) or generate a new analytics model (201) increases the accuracy of detecting anomalies in the robotic system. Additionally, the interaction between the factory operator and the analytics model (201) is reduced.

[0036] In an embodiment, the determination module (204) estimates the valid set points. The determination module (204) can be connected to a network, which can be local to the industrial plant or can be remote from the industrial plant. The determination module (204) can obtain data related to the thinner from different industrial plants. The obtained data can be used to update the analytics model (201a) or generate a new analytics model (201b).

[0037] This written description uses examples to describe the subject matter herein (including the best mode), and also enables any person skilled in the art to make and use the subject matter. The scope of patentability of the subject matter is defined by the claims, and may include other examples that occur to those skilled in the art. If such other examples have structural elements that are not different from the literal language of the claims, or if such other examples include equivalent structural elements that have insubstantial differences from the literal language of the claims, then such other examples are intended to fall within the scope of the claims.

[0038] Reference Numerals:

[0039] 101 - Vehicle / Object;

[0040] 102 – Robot Arm;

[0041] 103 - Controller;

[0042] 104 - Computing System;

[0043] 201 - Analytics Model;

[0044] 202 - Monitoring Module;

[0045] 203 - Analysis Module;

[0046] 204 - Decision Module;

[0047] 205 - Database;

[0048] 401 - First Pump;

[0049] 501 - Second Pump.

Claims

1. A method for detecting anomalies in a robotic system in an industrial plant, wherein, The robot system includes at least one robot and one or more controllers of a distributed control system (DCS), the one or more controllers being configured to control the at least one robot to perform operations on an object, wherein the DCS includes an analytics model for detecting anomalies in the robot system for predefined configuration parameters of the at least one robot and predefined process parameters associated with the robot system. The method includes: Monitoring configuration parameters of the at least one robot and process parameters associated with the robot system, wherein the configuration parameters are among the configuration parameters obtained from at least one of one or more sensors of the DCS and a database associated with the DCS; Detecting an association between at least one configuration parameter and at least one process parameter; Obtaining optimal configuration parameters and optimal process parameters based on the association, wherein the optimal configuration parameters and the optimal process parameters are analyzed for detecting anomalies, wherein at least one parameter causing the anomaly is identified from the optimal configuration parameters and the optimal process parameters, and wherein the anomaly is verified by a plant operator; Determining an error in the analysis based on the verification, wherein the error indicates an invalid set point for the at least one parameter; and Estimating a valid set point for the at least one parameter based on the error, wherein the valid set point is updated in the analytics model, and wherein the updated analytics model is used to accurately detect anomalies.

2. The method according to claim 1, wherein the configuration parameters at least include data related to a sprayer setting of the at least one robot, a path traversed by the at least one robot, and a size of the sprayer.

3. The method according to claim 1, wherein the process parameters include data related to a movement pattern of the at least one robot, a size of the object, one or more substances required for the process, and parameters related to the operation to be performed on the object.

4. The method according to claim 1, wherein machine learning techniques are used to obtain the optimal configuration parameters and the optimal process parameters.

5. The method according to claim 1, wherein the updated analytics model corresponding to the at least one parameter is stored in a memory of the DCS, wherein the memory includes an analytics model corresponding to each parameter of a plurality of at least one parameters, and wherein the corresponding analytics model is used to detect anomalies based on the at least one parameter.

6. A computing system for detecting anomalies in a robotic system in an industrial plant, wherein, The robot system includes at least one robot and one or more controllers of a distributed control system (DCS), the one or more controllers being configured to control the at least one robot to perform operations on an object, wherein the computing system includes an analytics model for detecting anomalies in the robot system for predefined configuration parameters of the at least one robot and predefined process parameters associated with the robot system, wherein the computing system includes: A processor configured to: Monitor configuration parameters of the at least one robot and process parameters associated with the robot system, where the configuration parameters are among the configuration parameters obtained from one or more sensors of the DCS and at least one of databases associated with the DCS; Detect an association between at least one configuration parameter and at least one process parameter; Obtain optimal configuration parameters and optimal process parameters based on the association, where the optimal configuration parameters and the optimal process parameters are analyzed for detecting anomalies, where at least one parameter causing the anomaly is identified from the optimal configuration parameters and the optimal process parameters, and where the anomaly is verified by a plant operator; Determine an error in the analysis based on the verification, where the error indicates an invalid set point for the at least one parameter; and Estimate a valid set point for the at least one parameter based on the error, where the valid set point is updated in the analytics model, and where the updated analytics model is used to accurately detect anomalies; And A memory configured to store the updated analytics model, the optimal configuration parameters, and the optimal process parameters.

7. The computing system according to claim 6, wherein the processor is configured to implement machine learning techniques to obtain the optimal configuration parameters and the optimal process parameters.

8. The computing system according to claim 6, wherein the computing system is associated with a user interface UI, and wherein the UI enables a plant operator to verify the analyzed anomalies.

9. The computing system according to claim 6, wherein the memory includes an analytics model corresponding to each of a plurality of at least one parameters, and wherein based on the at least one parameter, the corresponding analytics model is used to detect anomalies.

Images