A structured data model for the control of manufacturing equipment and its propagation.

Abstract

After start-up of the first machine, a standardized structured data model is instantiated in the first machine's controller to describe the first machine according to predefined categories with predefined labels indicating measured parameters of the first machine, components of the first machine, and subsystems of the first machine.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Application No. 18 / 308,853, filed April 28, 2023, which claims the benefit of U.S. Provisional Application No. 63 / 336,597, filed April 29, 2022, the contents of all of which are incorporated herein by reference in their entireties. [Technical field]

[0002] The present disclosure relates to control of manufacturing equipment. [Background technology]

[0003] Manufacturing control systems may respond to input signals by generating output signals that cause controlled equipment to operate in a particular manner. Summary of the Invention

[0004] The method includes instantiating, after startup of the first machine, on a controller of the first machine a standardized structured data model describing the first machine according to predefined categories populated with predefined labels indicative of measured parameters of the first machine, components of the first machine, and subsystems of the first machine, respectively. The predefined labels have a parent-child relationship defined by the predefined categories, such that the predefined labels indicative of the measured parameters are classified by predefined labels indicative of the components, and the predefined labels indicative of the components are classified by predefined labels indicative of the subsystems. The predefined categories and predefined labels of the first machine correspond to categories and labels describing the second machine such that their parent-child relationships correlate to the parent-child relationships of the labels describing the second machine. The method also includes instantiating on the controller of the first machine a version of a machine learning model trained on the second machine and in communication with the standardized structured data model, and controlling operation of the first machine according to an output of the machine learning model. [Brief description of the drawings]

[0005] [Figure 1] FIG. 1 is a block diagram of a manufacturing system. [Diagram 2] FIG. 1 is a block diagram of a manufacturing system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0006] An embodiment is disclosed herein. However, it should be understood that the disclosed embodiment is merely exemplary, and other embodiments may take various alternative forms. The figures are not necessarily to scale. Some features may be exaggerated or minimized to show details of specific components. Therefore, specific structural and functional details disclosed herein should not be construed as limiting, but merely as a representative basis for teaching those skilled in the art.

[0007] Various features illustrated or described with reference to any one embodiment may be combined with features illustrated or described in one or more other embodiments to produce embodiments not expressly illustrated or described. The illustrated combinations of features provide representative embodiments for typical applications. However, various combinations and modifications of features consistent with the spirit of the present disclosure may be desired for particular applications or implementations.

[0008] The standardized structured data model proposed here allows the modeling of a manufacturing line and all the machines therein. These models can be defined in a database. The use of the word "model" suggests that there is the ability to create multiple levels of parent-child relationships of machines and their characteristics. For example, given an extruder, the same model would include a screw, which would have temperature control zones, speed control devices, etc. These characteristics of the machine are important not only to collect and analyze data, but also to allow a common labeling language for use with other extruders. This, continuing with the example, could be useful to create software tools that can compare one extruder to another, communicate capabilities / efficiency in a common way, and create machine learning models that can apply what has been learned from one line to another. With regard to the latter, an example would be the discovery that an increase in the temperature zone of the screw of one extruder results in a corresponding change in the product dimensions in that extruder. This insight could be used to increase the learning rate of other extruders if all extruders are modeled in the same way.

[0009] A standardized structured modeling process typically looks like this: A user enters the machines of a production line and their characteristics into the database using a series of drop-down boxes provided by the front-end user interface to the database. The database model instantiates these as objects in the database, creating object definitions for all the equipment of the desired line. Once the modeling process is complete, the architecture uses the instantiated object definitions to connect to data sources and store the data from there in a structured way, i.e. by associating data points with the characteristics of each appropriate machine. Storing the data in this standardized structure allows for standard reporting such as machine efficiency and anomaly detection. As we will see later, machine learning processes are also improved using this method.

[0010] Continuing with the above example, and referring to FIG. 1, an extruder 10 may have multiple temperature sensors 14a, 14b, 14c, humidity sensors 18, motors 20, and controllers 22 corresponding to the screw 12, zones 16a, 16b, 16c, respectively. Upon initial installation, a standardized structured data model of the extruder 10 may be constructed. The table below illustrates one such possible configuration. For ease of discussion, this example has relatively few categories and labels. However, the ideas disclosed herein may be applied to more complex environments having hundreds, if not thousands, of categories and labels.

[0011] [Table 1]

[0012] In this example, three categories are used to characterize the extruder 10: abstraction, component, and feature. Features (measured parameters in this example) belong to components in a parent-child relationship, components belong to abstractions (subsystems in this example) in a parent-child relationship, and features belong to abstractions. The abstraction category, as its name suggests, characterizes the extruder 10 at a more conceptual level. The component category characterizes the extruder 10 by the actual equipment, grouped according to the corresponding abstraction. Thus, the numbered elements in FIG. 1 are assigned to component categories. Thus, single or multiple components can belong to a specific abstraction category. For example, only the screw 12 belongs to the abstraction category "mechanism", while the temperature sensor 14a, zone 16a, and humidity sensor 18 belong to the abstraction category "area 1". The same component can belong to multiple specific abstraction categories. For example, the humidity sensor 18 belongs to the abstraction categories "area 1", "area 2", and "area 3". As will be discussed in more detail below, the use of abstraction categories in addition to component categories can facilitate faster machine learning.

[0013] The feature categories characterize the extruder 10 by measurable parameters grouped according to the corresponding components. A single feature or multiple features can be used to describe a particular component. For example, the measured screw 12 RPM is used to describe the screw 12, and the measured temperature and temperature change are used to describe zone 16a.

[0014] In this example, the labels carry other contextual information that further helps to accelerate the machine learning process, since this information does not have to be learned from the training data. Some of the feature labels identify acceptable ranges of operation data and / or identify whether signals related to the same operation data are correlated. The feature “RPM” belonging to “Motor 20” indicates an acceptable operating range of 1000 rpm to 1150 rpm. The feature “Power Consumption” belonging to “Motor 20” indicates that its signal value is correlated with the signal value of “Temperature Change”. The feature “Temperature” belonging to “Temperature Sensor 14a” indicates that its signal value is affected by the operation of “Motor 20”. The feature “Humidity” belonging to “Area 1”, “Area 2”, and “Area 3”, respectively, indicates an acceptable range of less than 80% relative humidity. Some of the component labels include an identifier that indicates whether the device associated with the particular component label is a sensor or an active driver. The “Screw 12” indicates that it is an active driver, and the “Temperature Sensor 14a” indicates that it is a sensor. Some of the labels include a mark that indicates that the data corresponding to that label should definitely be included when creating a training dataset. The abstract label "Driver" is marked as being related to key data (e.g., data from "Motor 20" such as "RPM" and "Power Consumption"). The component label "Zone 16a" is also marked as such. The context information need not be limited to the examples given here. Thus, any relevant information can be included when creating a label.

[0015] The mapping detailed by Table 1, including the context information, can be strategically created by one or more users based on experience and other factors. Also, in contrast to existing mapping strategies, the concept of abstraction and context information is introduced through Table 1. This can improve the rate of machine learning since the collective associations of temperature sensors, zones, and humidity sensors in terms of “areas” in the abstraction categories do not need to be learned during the training phase, as described above. For example, to the extent that the measured parameters associated with the temperature sensors, zones, and humidity sensors collectively represent operating patterns under certain conditions that correlate with and / or impact the performance of the extruder 10, no learning time and data is required to recognize those correlations and / or impacts, since such correlations and / or impacts are already predefined. The same is true for the context information, etc.

[0016] Furthermore, if a second extruder is later introduced, either in the same or a different manufacturing facility, and conventional techniques are used to create a model of the extruder for machine learning purposes, that model may be created ad-hoc, ignoring the model in Table 1 that describes extruder 10. In such a situation, even though both extruders are the same, what has been machine learned about extruder 10 cannot be easily applied / transferred to the second extruder because the underlying mappings used to describe each are not the same.

[0017] Continuing with the example, leveraging the concept of a standardized structured data model described above, a mapping corresponding to the mapping used to describe extruder 10 is used to describe a second extruder. Referring to FIG. 2, extruder 110 comprises a screw 112, a number of temperature sensors 114a, 114b, 114c corresponding to zones 116a, 116b, 116c, respectively, a humidity sensor 118, a motor 120, and a controller 122. Table 2 shows that the mapping (i.e., also showing the parent-child relationship) and context information of extruder 110 correlates with the mapping of extruder 10. In this example, they are the same, but they do not need to be the same (nor do they need to be one-to-one) as long as the correspondence is clear.

[0018] [Table 2]

[0019] When applying machine learning to manufacturing, a template of a physical model corresponding to the relevant machine is typically selected and trained with training data before being used to control the machine. Training the template model involves, among other things, changing the weighting coefficients between nodes so that the template model performs better on the training data (and manufacturing data). Therefore, training the template model requires data and time. If the data models between the same machine are different (typically the case when models are created ad-hoc), each physical model linked to the corresponding data model must be trained with data and time, since one physical model cannot be directly applied to the other.

[0020] Now, if the machine data models (such as Tables 1 and 2) are standardized, the second machine can be initialized with a learned version of the first machine's physics model, greatly shortening the learning of the second machine's physics model. This is because the machine learning aspects associated with the second machine only need to be learned once by the first machine, and the second machine only needs to learn about its unique environment and its impact on its operation. Thus, the controller 122 can instantiate a version of the machine learning model trained on the extruder 10 to communicate with the standardized structured data model of Table 2, thereby, among other things, the machine learning model recognizes the parent-child relationships captured therein and the contextual information provided thereby (e.g., the machine learning model has access to the information held by the standardized structured data model, etc.).

[0021] The machine learning model of the controller 122 may then generate predicted parameter values ​​associated with parts output by the extruder 110 based on a live streaming feature set derived via pre-processing (e.g., data cleansing, principal component analysis, etc.) of the live data including outputs from the temperature sensors 114a, 114b, 114c, the humidity sensor 118, and the motor 120. The controller 122 can further direct control actions to the extruder 110 (e.g., the motor 120) based on the predicted parameter values ​​to change control settings (e.g., RPM settings, power consumption settings, etc.) in order to keep the predicted feature parameter values, and thus the actual values, at or near the target values.

[0022] If the extruders 10, 110 are of the same configuration and have the same standardized structured data model and the same machine learning model version with a common learning history (each communicating with one of the same corresponding structured data models), the control aspects learned for one of the extruders 10, 110 can be applied to the other of the extruders 10, 110. For example, when it is learned from the extruder 10 that the characteristic of the component produced by the extruder 10 falls out of its target range (e.g., the length of the part is greater than the target length) only when the speed of the screw 12 and the temperature in the zone 16b detected by the temperature sensor 14b are at a certain value and the relative humidity detected by the humidity sensor 18 is greater than a certain threshold (e.g., 70%), then the controller 22 can communicate control settings reflecting the same to the controller 122, which may be in the same or another location or facility. The controller 122 can automatically implement these received settings when certain conditions occur in order to avoid component parameters going outside of target ranges and control the extruder 110. Furthermore, considering that the extruders 10, 110 have the same standardized structured data model, analysis, comparison, reporting, and visualization of data relating to the same can be more efficient compared to when each extruder has a different structured data model.

[0023] The algorithms, methods, or processes disclosed herein may be implemented or performed on a computer, controller, or processing device, and may comprise any dedicated or programmable electronic control unit. Similarly, the algorithms, methods, or processes may be stored in many forms, including but not limited to, as data and instructions executable by a computer or controller, permanently stored in a non-writable storage medium, such as a read-only memory device, or mutably stored in a writable storage medium, such as a compact disk, random access memory device, or other magnetic and optical medium. The algorithms, methods, or processes may also be implemented as software executable objects. Alternatively, the algorithms, methods, or processes may be implemented in whole or in part using suitable hardware components, such as application specific integrated circuits, field programmable gate arrays, state machines, or other hardware components or devices, or a combination of firmware, hardware, and software components.

[0024] Although exemplary embodiments have been described above, it is not intended that these embodiments describe all possible forms encompassed by the scope of the claims. It should be understood that the words used herein are words of description rather than limitation, and that various changes may be made without departing from the spirit and scope of the present disclosure. For example, in this specification, the word controller may refer to the singular or the plural.

[0025] As previously mentioned, features of various embodiments may be combined to form further embodiments of the invention not expressly described or illustrated. Although various embodiments may be described as offering advantages or being preferred over other or prior art embodiments with respect to one or more desired characteristics, those skilled in the art will recognize that one or more features or characteristics may be compromised to achieve desired overall system attributes depending on the particular application and embodiment. These attributes include, but are not limited to, cost, strength, durability, life cycle cost, marketability, appearance, packaging, size, maintainability, weight, manufacturability, ease of assembly, and the like. Thus, embodiments described as less desirable than other or prior art embodiments with respect to one or more characteristics are not outside the scope of this disclosure and may be desirable for particular applications.

Claims

1. After booting up the first machine, instantiating in a controller of the first machine a standardized structured data model describing the first machine according to predefined categories having predefined labels indicative of measured parameters of the first machine, components of the first machine, and subsystems of the first machine, respectively, the predefined labels having a parent-child relationship defined by the predefined categories, such that the predefined labels indicative of the measured parameters are classified by the predefined labels indicative of the components, and the predefined labels indicative of the components are classified by the predefined labels indicative of the subsystems, the predefined categories and the predefined labels correspond to categories and labels describing the second machine such that the parent-child relationship correlates with a parent-child relationship of labels describing a second machine; instantiating on the controller a version of a machine learning model that was trained on the second machine and that communicates with the standardized structured data model; and controlling operation of the first machine according to an output of the machine learning model.

2. 10. The method of claim 1, further comprising receiving data from the second machine defining a configuration of the second machine, updating a version of the machine learning model with the data, and executing the configuration by the controller.

3. The method of claim 1 , wherein one or more of the predefined labels comprises information defining a target value range for a corresponding one or more of the measured parameters.

4. The method of claim 1 , wherein one or more of the predefined labels comprises information indicating that corresponding signals are correlated.

5. The method of claim 1 , wherein one or more of the predefined labels comprises information indicating that a corresponding signal value is affected by an operation of at least one of the components.

6. The method of claim 1 , wherein one or more of the predefined labels include information identifying whether a corresponding one or more of the components is a sensor.

7. 10. The method of claim 1, wherein one or more of the pre-defined labels comprises information that identifies whether corresponding data should be included in a dataset used to train the machine learning model.

8. The method of claim 1 , wherein a plurality of predefined labels indicative of the measured parameters are categorized according to one of the predefined labels indicative of the component.

9. The method of claim 1 , wherein one of the predefined labels representing the measured parameter is classified with a plurality of predefined labels representing the component.

10. The method of claim 1 , wherein a plurality of predefined labels representing the components are categorized according to one of the predefined labels representing the subsystems.

Images