Threshold determination for predictive process control of factory processes, equipment and automated systems

Abstract

The deep learning process receives desired process values ​​associated with one or more process stations. The deep learning processor receives desired target values ​​for one or more key performance indicators of the manufacturing process. The deep learning processor simulates the manufacturing process to generate predicted process values ​​and predicted target values ​​for the one or more key performance indicators for optimizing the one or more key performance indicators. The simulating includes generating proposed state changes for at least one process parameter of the initial set of process parameters. The deep learning processor determines that the predicted process values ​​and predicted target values ​​are within tolerance limits of the desired process values ​​and desired target values. Based on this determination, the deep learning process generates changes to the initial set of process parameters based on the proposed state changes.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Patent Application No. 17 / 931,442, filed September 12, 2022, which is incorporated herein by reference in its entirety.

[0002] The present disclosure relates generally to systems, apparatus and methods for predictive process control of factory processes, equipment and automation systems. Summary of the Invention [Problem to be solved by the invention]

[0003] Constant monitoring and adjustments to the manufacturing process are required to safely, timely, and minimize waste production of products that consistently meet desired design specifications. [Means for solving the problem]

[0004] In some embodiments, a manufacturing system is disclosed herein. The manufacturing system includes one or more process stations, a station control system, and a controller. The one or more process stations are configured to perform a manufacturing process. The manufacturing process includes an initial set of processing parameters. The station control system is configured to control the one or more process stations. The controller is in communication with the one or more process stations and the station control system. The controller is configured to perform an operation. The operation includes receiving desired process values ​​associated with the one or more process stations. The operation further includes receiving desired target values ​​for one or more key performance indicators of the manufacturing process. The operation further includes simulating, with a deep learning processor, the manufacturing process to generate expected process values ​​and expected target values ​​for the one or more key performance indicators for optimizing the one or more key performance indicators. The simulating includes generating proposed state changes for at least one process parameter of the initial set of processing parameters. The operation further includes determining, with the deep learning processor, that the expected process values ​​and the expected target values ​​are within tolerance limits of the desired process values ​​and the desired target values. The operations further include causing, based on the determination, a change to the initial set of processing parameters based on the proposed state change.

[0005] In some embodiments, a computer-implemented method is disclosed herein. A computing system receives desired process values ​​associated with one or more process stations, the one or more process stations being configured to perform a manufacturing process. The manufacturing process includes an initial set of process parameters. The computing system receives desired target values ​​for one or more key performance indicators for the manufacturing process. A deep learning processor of the computing system simulates the manufacturing process to generate expected process values ​​and expected target values ​​for the one or more key performance indicators for optimizing the one or more key performance indicators. The simulating includes generating proposed state changes for at least one process parameter of the initial set of process parameters. The deep learning processor of the computing system determines that the expected process values ​​and expected target values ​​are within tolerance limits of the desired process values ​​and desired target values. Based on this determination, the computing system generates changes to the initial set of process parameters based on the proposed state changes.

[0006] In some embodiments, a computer-implemented method is disclosed herein. A computing system receives actual output values ​​corresponding to a manufacturing process of a manufacturing system. The receiving includes receiving stable data corresponding to normal production data during a normal production phase of the manufacturing process and receiving unstable data corresponding to abnormal production data outside the normal production phase of the manufacturing process. The computing system generates a training data set including the stable data and the unstable data. The computing system trains a machine learning model to generate predicted process values ​​and predicted target values ​​for one or more key performance indicators. The computing system determines that the machine learning model has achieved a threshold level of accuracy. In response to this determination, the computing system deploys the machine learning model within the manufacturing system.

[0007] A more particular description of the present disclosure briefly outlined above may be had by reference to embodiments, some of which are illustrated in the accompanying drawings, in a manner that allows the above-listed features of the disclosure to be understood in detail. It should be noted, however, that the accompanying drawings illustrate only typical embodiments of the present disclosure and therefore should not be considered as limiting the scope of the present disclosure, since the present disclosure may embrace other equally effective embodiments. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 illustrates an exemplary manufacturing environment in accordance with an illustrative embodiment.

[0009] [Figure 2] FIG. 1 is a block diagram illustrating a deep learning processor in accordance with an exemplary embodiment.

[0010] [Figure 3] FIG. 1 is a block diagram illustrating a deep learning processor in accordance with an exemplary embodiment.

[0011] [Figure 4] 1 is a flow diagram illustrating a method for training a deep learning processor to generate predicted values, according to an example embodiment.

[0012] [Figure 5] 1 is a flowchart illustrating a method for generating recommended state changes for a manufacturing process in accordance with an illustrative embodiment.

[0013] [Figure 6] 1 is a flowchart illustrating a method for generating recommended state changes for a manufacturing process in accordance with an illustrative embodiment.

[0014] [Figure 7] 1 is a flowchart illustrating a method for generating recommended state changes for a manufacturing process in accordance with an illustrative embodiment.

[0015] [Figure 8] FIG. 1 illustrates an exemplary dashboard in accordance with an exemplary embodiment.

[0016] [Figure 9A] FIG. 1 illustrates an architecture of a computing system in accordance with an exemplary embodiment.

[0017] [Figure 9B] FIG. 1 illustrates a computer system having a chipset architecture according to an exemplary embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0018] For ease of understanding, where possible, like reference numerals have been used to designate like elements common to the figures, and it is contemplated that elements disclosed in one embodiment may be beneficially utilized on other embodiments without specific recitation.

[0019] Manufacturing within a manufacturing environment (e.g., a factory) may include various processes for controlling manufacturing components (e.g., machines, control systems) within the environment. In particular, the components enabling manufacturing within a factory may be complex and may include various process stations (or "stations") that process raw ingredients to produce a final output product. The various process stations may receive inputs for processing and output intermediate outputs that are passed to one or more subsequent (downstream) processing stations for further processing. A final process station may receive inputs for processing and output a final output.

[0020] Each process station may include one or more tools / equipment configured to perform a set of process steps on raw material or intermediate output obtained from another process station. Examples of process stations may include, but are not limited to, conveyor belts, injection molding presses, cutting machines, die stamping machines, extruders, CNC mills, grinders, assembly stations, 3D printers, robotic devices, quality control and verification stations, etc. Exemplary process steps may include, but are not limited to, transporting output from one location to another (as performed by a conveyor belt), feeding material to an extruder, melting the material, and injecting the material into a mold cavity (as performed by an injection molding press), cutting material to a specific shape or length (as performed by a cutting machine), pressing material to a specific shape (as performed by a die stamping machine), mixing, bubble columns, vapor deposition, etc.

[0021] In a manufacturing process, process stations can operate in parallel or in series. When operating in parallel, a single process station can send its intermediate output to two or more stations (e.g., 1 to N stations), and a single process station can receive and combine intermediate outputs from 2 to N stations. Furthermore, a single process station can perform the same or different process steps, sequentially or non-sequentially, on the received raw materials or intermediate outputs during a single iteration of the manufacturing process.

[0022] The operation of each process station can be controlled by one or more process controllers. In some embodiments, each process station has one or more process controllers (referred to herein as "station controllers") programmed to control the operation of the process station (programming algorithms referred to herein as "control algorithms"). However, in some aspects, a single process controller may be configured to control the operation of more than one process station. One example of a station controller is a programmable logic controller (PLC). PLCs can be programmed to operate manufacturing processes and systems. PLCs or other controllers can receive information from connected sensors or input devices, process the data, and generate outputs (e.g., control signals for controlling associated process stations) based on pre-programmed parameters and instructions.

[0023] An operator or control algorithm can provide station controller setpoints (or "setpoints" or "controller setpoints" or CSPs) to a station controller, representing a desired single value or range of values ​​for each control value. Values ​​that can be measured during operation of equipment or during a process at a station can be classified as control values ​​or station values. Values ​​controlled by a station controller can be classified as control values ​​herein, and other measured values ​​are classified as station values ​​herein. Examples of control values ​​and / or station values ​​include, but are not limited to, speed, temperature, pressure, vacuum, rotations, current, voltage, power, viscosity, materials / resources used at the station, extrusion rate, power outage duration, hazardous fumes, and the type and sequence of steps performed at the station. Whether a measured value is classified as a control value or a station value can depend on the particular station and whether the measurement is controlled by the station controller or is simply a by-product of the operation of the station. During the manufacturing process, control values ​​can be provided to a station controller, while station values ​​may not be provided to the controller.

[0024] The control algorithm may further include instructions for monitoring the control value, comparing the control value with a corresponding setpoint, and determining what action to take when the control value is not equal to the corresponding station controller setpoint (or is not within a specified range of the corresponding station controller setpoint). For example, if the current measured temperature of a station is lower than the setpoint, a signal may be sent by the station controller to increase the temperature of the station's heat source until the station's current temperature value equals the setpoint.

[0025] Many process controllers used in manufacturing processes to control stations can be limited as such controllers follow static algorithms (e.g., on / off control, PI control, PID control, lead / lag control) to dictate what action to take when a control value deviates from a set point.

[0026] Each process station may include one or more sensors or may be coupled to one or more sensors. These sensors may be physical or virtual sensors present in the manufacturing process that are unrelated to the operation of the deep learning processor, as well as new sensors that may be added to perform additional measurements required by the deep learning processor. Sensors may be used to measure values ​​generated by the manufacturing process, such as station values, control values, intermediate and final output values, etc. Exemplary sensors may include, but are not limited to, rotary encoders for detecting position and velocity, sensors for detecting proximity, pressure, temperature, level, flow, current, and voltage, and limit switches for detecting conditions such as presence or end-of-travel limits.

[0027] The sensors described herein can include both sensing devices and signal conditioning. For example, sensing devices respond to station or control values, and signal conditioners convert the response into a signal that can be used and interpreted by a deep learning processor or station controller. Examples of sensors that respond to temperature are RTDs, thermocouples, and platinum resistance probes. Strain gauge sensors respond to changes in pressure, vacuum, weight, and distance, among other things. Proximity sensors respond to objects when they are within a certain distance or threshold from one another. In all of these examples, the response must be converted into a signal that can be used by a station controller or deep learning processor. In many cases, the sensor's signal conditioning function generates a digital signal that is interpreted by the station controller. Signal conditioners can also generate analog or TTL signals, among other things. Virtual sensors, also known as soft sensors, smart sensors, or estimators, include system models that can receive and process data from physical sensors.

[0028] As used herein, a process value refers to a station value or control value that is aggregated or averaged across the entire set of stations (or a subset of stations) that are part of a manufacturing process. Process values ​​may include, for example, total output time, total resources used, average temperature, average speed.

[0029] In addition to station and process values, various characteristics of the product output (i.e., intermediate or final output) of a process station can be measured, such as temperature, weight, product dimensions, mechanical, chemical, optical, and / or electrical properties, the number of design defects, and the presence or absence of defect types. Generally, the various characteristics that can be measured are referred to as "intermediate output values" or "final output values." An intermediate / final output value can reflect a single measured characteristic of the intermediate / final output or an overall score based on a specified set of characteristics related to the intermediate / final output that are measured and weighted according to a predetermined formula.

[0030] Mechanical properties may include hardness, compression, viscosity, density, and weight. Optical properties may include absorption, reflection, transmission, and refraction. Electrical properties may include electrical resistivity and electrical conductivity. Chemical properties may include enthalpy of formation, toxicity, chemical stability in a given environment, ignition ability (ability to burn), preferred oxidation state, pH (acidity / alkalinity), chemical composition, boiling point, vapor point). The disclosed mechanical, optical, chemical, and electrical properties are merely examples and are not intended to be limiting.

[0031] Processes are typically controlled by a process controller using a control algorithm such as PID. Overriding this process controller is another monitoring and control layer called statistical process control (SPC). The purpose of SPC is to provide an alert when a particular process deviates from normal operation. A basic tenant of SPC is that the process controller should not be adjusted unless the process variable being monitored by the SPC indicates an out-of-control condition.

[0032] New methodologies for controlling processes are known as predictive process control (PPC), such as those described in U.S. Patent Nos. 11,156,991 and 11,156,992, which are incorporated herein by reference in their entireties. These methods use machine learning to model processes and implement changes to processes in order to optimize process or end-product characteristics. Such techniques may allow users to specify a confidence threshold that must be achieved before implementing or recommending a process change.

[0033] One or more techniques described herein refine the use of predictive process control by providing additional threshold methodologies that enable users to better understand when and how to have a deep learning processor implement or recommend changes to a process. Such techniques can inform users that changes to a process will have a positive or negative impact on the process, depending on a combination of the deep learning processor's confidence score, known process limits, user-assigned specification limits, and known SPC rules. Such techniques can be deployed in real time and communicated to users through interactive dashboards and notifications.

[0034] 1 illustrates an exemplary manufacturing system 100 in accordance with an exemplary embodiment. As shown, the manufacturing system 100 may include a factory data system 102, a station control system 104, a first process station 106, a second process station 108, and a deep learning processor 118.

[0035] Factory data system 102 may represent any factory data system, such as, but not limited to, a supervisory control and data acquisition (SCADA) system. A SCADA system may include software and hardware elements that enable industrial entities to control industrial processes locally or remotely, monitor, collect, and process real-time data, interact directly with devices such as sensors, valves, pumps, motors, etc. through human-machine interface (HMI) software, record events in log files, etc.

[0036] The station control system 104 may be configured to control one or more process stations, such as the first process station 106 and the second process station 108. In some embodiments, the station control system 104 may include multiple individual station controllers. Each station controller may be configured to control one or more process stations. For example, a first station controller may be configured to control the first process station 106, and a second station controller may be configured to control the second process station 108.

[0037] In operation, operational instructions may be provided from the factory data system 102 to the station control system 104 to monitor or control various aspects of the process stations. For example, as shown in FIG. 1 , the factory data system 102 may provide operational instructions 101 to the station control system 104. The station control system 104 may then forward those operational instructions in the form of one or more control signals to a corresponding one or more process stations (e.g., a first process station 106 and / or a second process station 108) to control the processes at the process stations. For example, as shown, the station control system 104 may provide a control signal 115 to the first process station 106 based on the operational instructions 101, and similarly, the station control system 104 may provide a control signal 116 to the second process station 108 based on the operational instructions 101.

[0038] As an example, an operational command sent from the factory data system 102 to the station control system 104 may instruct the station control system 104 to set the temperature of a process station to 150° C. To achieve the desired set temperature, the station control system 104 may output a fixed voltage (e.g., a control signal) to a heating element in the corresponding process station such that the temperature is maintained at 150° C. In this example, the station control system 104 may simply change the output voltage according to the change indicated by the new operational command provided by the factory data system 102.

[0039] In some embodiments, the operational instructions 101 may refer to a set of process parameters associated with the manufacturing process. For example, the process parameters may include, but are not limited to, set points, algorithms, initial inputs, and operational instructions, and a local or central data processing system, such as the factory data system 102, may provide system and process updates and other control inputs to the station control system 104. In some embodiments, such processes may be performed manually by an operator. In some embodiments, the factory data system 102 may also receive data outputs generated by the station control system 104, as well as data generated by sensors coupled to or within the first and second process stations 106, 108, or data from independent sensors. In some embodiments, the data output may include, but is not limited to, (i) data generated during the manufacturing process (e.g., data logs coupled to physical sensors, process station components, or station controller components), (ii) data received by or transmitted from each process station or station controller, and (iii) data communication and data generation patterns (e.g., high data volume, low data volume, inconsistent data volume, anomalous data communication or data generation based on time, origin, or destination of data) for individual process stations or station controllers or any number of process stations or station controllers. In some embodiments, the factory data system 102 may receive the response data. In some embodiments, the factory data system 102 may also receive data from associated manufacturing processes running at geographically remote locations.

[0040] The deep learning processor 118 may be configured to simulate a manufacturing process. In some embodiments, the deep learning processor 118 may represent a simulator, which may be based on predictions, physical laws, statistical regression, or other methods. For example, the deep learning processor 118 may undergo a training process during which the deep learning processor 118 may learn to simulate a manufacturing process performed by the manufacturing system 100. In this manner, the deep learning processor 118 may be configured to determine predicted values ​​of intermediate values, final output values, or key performance indicators based on input information. For example, the deep learning processor 118 may be configured to determine predicted values ​​of intermediate values, final output values, or key performance indicators based on control signals provided by the station control system 104 to the first process station 106 and / or the second process station 108. If the predicted values ​​deviate from the actual values, the deep learning processor 118 may be trained to recommend changes or adjustments to the manufacturing system to achieve the desired values.

[0041] As shown, the deep learning processor 118 may receive various information from the station control system 104, the first process station 106, the second process station 108, and / or one or more external sources to determine the expected value. For example, the deep learning processor 118 may receive a replicated control signal 165 from the station control system 104. The replicated control signal 165 may correspond to a control signal sent to the first process station 106 (e.g., control signal 115) and / or a control signal sent to the second process station 108 (e.g., control signal 116). In some embodiments,

[0042] In some embodiments, the deep learning processor 118 may receive inputs from one or more process stations. For example, as shown, the deep learning processor 118 may receive inputs 140 from the first process station 106. The inputs 140 may represent all values ​​associated with the first process station 106. For example, the deep learning processor 118 may receive control signals, control values, process values, independent sensor values, intermediate output values, and / or final output values ​​from the first process station 106.

[0043] Similarly, the deep learning processor 118 may receive inputs 141 from the second process station 108. The inputs 141 may represent all values ​​associated with the second process station 108. For example, the deep learning processor 118 may receive control signals, control values, process values, independent sensor values, intermediate output values, and / or final output values ​​from the second process station 108.

[0044] In some embodiments, as briefly described above, the control signal may correspond to a signal from the station control system 104 to a given process station. For example, a plastic extruder may require heat to condition the plastic during extrusion. In such an example, the station control system 104 may provide a temperature setpoint to the first process station 106 and / or the second process station 108 that is a target temperature for the extruder.

[0045] In some embodiments, a control value may refer to an output value from a given process station that may be used as feedback to the station control system 104. Continuing with the example above, a thermocouple may be located near the extruder barrel, and the output of this thermocouple may be sent back to the station control system 104 to provide feedback to the control loop. In some embodiments, such as those discussed above, these control values ​​may be sent to the deep learning processor 118 for further analysis.

[0046] In some embodiments, a station value may refer to a value that is associated with a process station but is not used for control feedback. Continuing with the example above, a thermocouple may be positioned so that the temperature of the plastic after it exits the extruder may be measured.

[0047] In some embodiments, the intermediate value 130 may refer to a value associated with a given process station, generally related to a work-in-process or sample characteristic, but not directly related to the value used for feedback. Continuing with the example above, plastic after exiting the extruder may exhibit die swell. This die swell may be measured by non-contact means and may represent an intermediate value.

[0048] In some embodiments, the final output value 131 may refer to a value associated with a given process station that generally relates to a characteristic of the final product, but is not directly related to the value used for feedback. Continuing with the example above, the plastic after exiting the extruder may exhibit a die swell. This die swell may be measured by non-contact means and may represent the final output value. The final output value may be related to product specifications, which determine whether the product is acceptable for its end use and intended application.

[0049] During training of the deep learning processor 118, the deep learning processor 118 may be trained to predict expected intermediate values ​​and / or expected final output values ​​based on various inputs. For example, the deep learning processor 118 may be trained to generate expected intermediate values ​​based on one or more of the control signals, control values, or station values. Similarly, the deep learning processor 118 may be trained to generate expected final output values ​​based on one or more of the control signals, control values, station values, or intermediate values.

[0050] As will be appreciated by those skilled in the art, a manufacturing process may be evaluated by key process indicators (KPIs). Managers may set targets for certain KPIs to be achieved during processing. For example, common KPIs may include cost and yield (e.g., usable product / raw materials). The deep learning processor 118 may be trained against such KPIs to optimize the manufacturing process to achieve maximum yield at minimum cost. KPIs may broadly refer to process values ​​of a manufacturing process, whether intrinsic or measured. For example, in some embodiments, a KPI may be a calculated value based on one or more intermediate or final output values. For example, yield may be calculated from a measurement of the final output value to determine whether the product is acceptable for its end use or intended use.

[0051] Similar to the process described above, the deep learning processor 118 may be trained to predict expected KPI values ​​based on various inputs, for example, the deep learning processor 118 may be trained to generate expected KPI values ​​based on one or more of a control signal, a control value, or a station value.

[0052] After the deep learning processor 118 has been properly trained (i.e., when the predicted values ​​generated during training are within a threshold tolerance of the associated actual values), the deep learning processor 118 may be deployed to monitor the manufacturing system. In some embodiments, once deployed, the deep learning processor 118 may be further configured to optimize KPIs for a given manufacturing process. For example, the deep learning processor 118 may utilize predictive process control techniques to determine whether to change manufacturing process conditions to achieve desired KPIs or improve existing KPIs.

[0053] As will be appreciated by those skilled in the art, a manufacturing process may include one or more process stations, each of which may be controlled or monitored by one or more parameters. The characteristics of these parameters may represent the "state" of the process. For example, temperature is a parameter of an extrusion process. This temperature is controlled by the station control system 104, which may change this temperature by adjusting the setpoint of the corresponding process station. In another example, the state of one of the extruders may be the temperature. The setpoint may be changed to change the state of the manufacturing process.

[0054] However, there is a subtle difference between a controller setpoint and a manufacturing process state. For example, a process parameter can be a station value or a control value. In the extruder example, for the temperature of the extrudate, the station value may be higher than the control value due to heat generated by the workpiece placed in the extrudate during processing. Although the control value and the station value are related, the station value is the parameter of interest. In this case, the temperature of the extrudate can be referred to as a manufacturing process state.

[0055] In some embodiments, the state of a process may be evaluated by parameters that a user deems important to the workability or product. These parameters may be of variable or attribute data type. Variable data may refer to information that is continuously measured. Examples of variable data in a manufacturing context may include, but are not limited to, temperature, pressure, weight, and energy. Attribute data may refer to information that is discretely observed, but not measured. Examples of attribute data in a manufacturing context may include, but are not limited to, pass / fail, conform / nonconform, good / bad. Attribute data may broadly refer to qualitative data, while variable data may refer to quantitative data.

[0056] After the deep learning processor 118 has been properly trained to generate predicted values, it may be deployed to recommend state changes to the manufacturing process. For example, the deep learning processor 118 may be configured to run a background process in which the state of process stations is changed in an attempt to optimize or improve KPIs. During such a process, the deep learning processor 118 may be further configured to recommend changes that meet the equipment specifications 170 and the user specifications 180.

[0057] The equipment specifications 170 may refer to the limits of a process station. For example, if a given process station cannot exceed a certain temperature, the equipment specifications 170 for that process station may be an upper temperature limit. Generally, such equipment specifications may be considered "unchangeable." In other words, the equipment specifications may be fixed and cannot be changed. Using another example, an extruder specification for screw speed has a minimum and maximum speed that cannot be exceeded.

[0058] User specifications 180 may refer to user specifications for each process station. For example, a user or administrator may specify a processing range for a particular process station. Such user specifications 180 may be a subset of equipment specifications 170. For example, a press may have an operating range of 100°C to 200°C, but the user specifications may be narrower, such as 125°C to 150°C. In some embodiments, a user may further specify requirements for intermediate or final output values. For example, a user may specify that an intermediate output has a specification having a weight between 500 and 510 grams.

[0059] In some embodiments, user specifications 180 may be variable specifications. In other words, user specifications 180 may include a subset of user specifications that can be changed. Continuing with the example above, the user's narrower 125°C to 150°C range can be changed to a wider or narrower range, as long as the range is within the fixed limits of the device.

[0060] In some embodiments, deep learning processor 118 may be in communication with display output 190. In some embodiments, display output 190 may represent a display device in wired or wireless communication with deep learning processor 118. In some embodiments, display output 190 may represent a computing system (e.g., a desktop, laptop, mobile device, tablet, etc.) in wired or wireless communication with deep learning processor 118.

[0061] In some embodiments, a user interface may be displayed via the display output 190. The user interface may present a dashboard or report corresponding to the manufacturing process. For example, the interface may provide one or more of historical data for each process station, real-time data for each process station (e.g., actual values ​​for each process station), forecast values ​​for each process station, actual KPIs, and forecast KPIs.

[0062] 2 is a block diagram of the deep learning processor 118 according to an example embodiment. As shown, the deep learning processor 118 may include a repository 202. The repository 202 may include one or more data storage nodes, which may include instructions. The repository 202 may interact with processors 206a, 206b, and / or 206c to implement the predictive process control platform 204 as described herein.

[0063] The predictive process control platform 204 may be configured to train a model to generate predicted values ​​using a training data set. In some embodiments, the predicted values ​​may include predicted control values, predicted station values, predicted intermediate values, predicted KPIs, and / or predicted final output values. The predictive process control platform 204 may include an intake module 208 and a training module 210. The intake module 208 and the training module 210 may each comprise one or more software modules. The one or more software modules may be a collection of code or instructions stored on a medium (e.g., memory of the deep learning processor 118) representing a series of machine instructions (e.g., program code) that implement one or more algorithm steps. Such machine instructions may be actual computer code that a processor interprets to implement the instructions, or alternatively, may be a higher-level encoding of instructions that are interpreted to obtain the actual computer code. The one or more software modules may further include one or more hardware components. One or more aspects of the exemplary algorithm may be performed by the hardware component (eg, circuitry) itself, rather than as a result of instructions.

[0064] The intake module 208 can acquire or receive a training data set. In some embodiments, the training data set can include actual values. Exemplary actual values ​​may include, but are not limited to, actual control values, actual station values, actual intermediate values, actual KPI targets, and / or actual final output values. In some embodiments, the intake module 208 can be configured to generate a robust data set. To generate a robust data set for tuning a machine learning model, the setpoints (or other control inputs) corresponding to each control value for each process station can be adjusted in a systematic manner (e.g., from minimum to maximum) for every value (or subset of values) that provides an in-specification final output. In additional embodiments, the setpoints (or other control inputs) corresponding to each control value for each process station can be adjusted in a systematic manner (e.g., from minimum to maximum) for every value (or subset of values) at which the process station can operate (i.e., for the entire range of values ​​at which the process station can operate, not limited to only values ​​that provide an in-specification final output). Furthermore, any number and combination of setpoints can be adjusted for training purposes. The set point (or other control input) can be adjusted manually, by pre-programmed algorithms, or by predictive process control.

[0065] The training module 210 may be configured to train the machine learning model 212 to generate predicted values ​​based on actual values ​​to generate a fully trained model 214. For example, the machine learning model 212 may be trained to evaluate control / station / process values, intermediate values, and final output values ​​and determine adjustments to station controller inputs. As will be appreciated by those skilled in the art, the machine learning-based techniques can be modified depending on the desired implementation without departing from the disclosed technology. For example, the machine learning techniques can utilize one or more of the following, alone or in combination: hidden Markov models; recurrent neural networks; convolutional neural networks (CNNs); deep learning; Bayesian symbolic methods; reinforcement learning, generative adversarial networks (GANs); support vector machines; image registration methods; and adaptive rule-based systems.

[0066] The machine learning model may also be based on a clustering algorithm (e.g., Minibatch K-means clustering algorithm), a recommendation algorithm (e.g., Miniwise Hashing algorithm or Euclidean Locality Sensitive Hashing (LSH) algorithm), and / or anomaly detection algorithm, e.g., local outlier factor method. The machine learning model may be based on supervised and / or unsupervised methods.

[0067] In some embodiments, a Transformer architecture may be used to predict expected values ​​over short periods of time in the future, for example, based on observed values ​​(e.g., station values, control values, KPIs, etc.) over a time interval. In some embodiments, a mask may be used to limit the attention of the machine learning model 212 to data from a recent time interval, such as the past 20 seconds, and the mask may be used to predict a subsequent time interval, such as the next 20 seconds. Additionally, a fully trained model 214 may be evaluated to predict the outcome of proposed changes to observed values ​​over short future time intervals.

[0068] In some embodiments, training and testing may be completed based on a generated confidence level for the prediction at a single moment in time or over a particular time period. Such confidence levels may provide a measure of the statistical reliability of the prediction. In some embodiments, the confidence level may be expressed as a numerical probability of the prediction's accuracy. In some embodiments, the confidence level may be expressed as an interval or probability range.

[0069] 3 is a block diagram of a deep learning processor 118 according to an example embodiment. As shown, after the fully trained model 214 achieves a threshold confidence level, the fully trained model 214 may be deployed to generate proposed state changes.

[0070] As shown, the fully trained model 214 may receive as inputs a desired value for a process value 302 and a KPI target 304. In some embodiments, the desired value for the process value 302 may include, but is not limited to, a desired station value, an intermediate output value, a final output value, a control value, an independent sensor value, or generally any other data that may be associated with a process station of the manufacturing system 100.

[0071] Based on the desired process values ​​302 and KPI targets 304, the well-trained model 214 may be able to simulate the manufacturing process to generate expected process values ​​and expected targets for a given KPI or a given set of KPIs. For example, the deep learning processor 118 may simulate the manufacturing process using a set of set points, algorithms, initial inputs, and / or operational instructions that would normally be provided to the station control system 104 from the factory data system 102 via the well-trained model 214. The deep learning processor 118 may generate a set of set points, algorithms, initial inputs, and / or operational instructions for the simulation. In some embodiments, the generated set of set points, algorithms, initial inputs, and / or operational instructions may include state changes to the manufacturing process. The state changes may be changes to one or more parameters of a process station or set of process stations as compared to “normal” or existing manufacturing parameters. Using the set of settings, algorithms, initial inputs, and / or operating instructions, the deep learning processor 118 may simulate the manufacturing process to generate predicted process values ​​and predicted targets for a set of KPIs.

[0072] In some embodiments, as an output, the deep learning processor 118 may provide the suggested state change as an alert 306 to a user. In some embodiments, as an output, the deep learning processor 118 may provide the suggested state change directly to the factory data system 102 and / or the station control system 104.

[0073] 4 is a flow diagram illustrating a method 400 for training a deep learning processor to generate predictions, according to an example embodiment. Method 400 may begin at step 402.

[0074] In step 402, a manufacturing process may be initiated. For example, the manufacturing process may be initiated within the manufacturing system 100 by the factory data system 102. The factory data system 102 may provide operational instructions to the station control system 104. The station control system 104 may then provide appropriate control signals to the respective process stations based on the operational instructions.

[0075] Actual output data from the manufacturing process may be received by the deep learning processor 118 at step 404. In some embodiments, the actual output data from the manufacturing process may include, but is not limited to, control values ​​received from the station control system 104, station values ​​received from the first process station 106 and / or the second process station 108, intermediate values ​​generated by the first process station 106 and / or the second process station 108, final output values, and / or data indicative of KPIs.

[0076] In some embodiments, the actual output may include actual output from various stages of the manufacturing process. For example, the deep learning processor 118 may receive actual output data during startup, ramp-up, shutdown, ramp-down, idle, normal production, etc. Additionally, in some embodiments, the process setpoints of the manufacturing system may be adjusted or changed to accommodate the additional actual output to generate a more robust training set for the deep learning processor 118. In this manner, the deep learning processor 118 may be trained on both stable data (e.g., normal production data) and unstable data (e.g., startup, ramp-up, shutdown, etc.). In some embodiments, stable data may refer to data acquired during the manufacturing process when the manufacturing system 100 is producing in-spec product. In some embodiments, stable data may refer to data acquired during the manufacturing process when the manufacturing process is running within statistical process control.

[0077] In step 406, the deep learning processor 118 may be trained on the generated forecast values. In some embodiments, the deep learning processor 118 may be trained based on actual output data, equipment specifications, to generate the forecast values. In some embodiments, the forecast values ​​may include forecast control values, forecast station values, forecast intermediate values, forecast KPIs, and / or forecast final output values.

[0078] At step 408, the deep learning processor 118 is tested to determine a confidence level for the prediction. For example, the deep learning processor 118 may provide a confidence level for the prediction at a single moment in time or over a particular time period, e.g., to provide a measure of the statistical reliability of the prediction. In some embodiments, the confidence level may be expressed as a numerical probability of the accuracy of the prediction. In some embodiments, the confidence level may be expressed as an interval or probability range.

[0079] If the predictions generated by the deep learning processor 118 are correct over a specified period of time and within a predetermined threshold confidence level, the deep learning processor 118 can assume control of the manufacturing process.

[0080] 5 is a flow diagram illustrating a method 500 for generating recommended state changes for a manufacturing process, according to an example embodiment. Method 500 may begin at step 502. As described above, after the deep learning processor 118 has been trained to generate predictions within a predetermined threshold confidence level, the deep learning processor 118 may be deployed to recommend state changes for the manufacturing process.

[0081] In step 502, the deep learning processor 118 may receive a process value and a desired value for a KPI target. In some embodiments, the desired value for the process value may include, but is not limited to, a desired station value, an intermediate output value, a final output value, a control value, an independent sensor value, or generally any other data that may be associated with a process station of the manufacturing system 100. In some embodiments, the desired KPI target may be a desired target value for a given KPI. For example, for an "increased throughput" KPI, the desired target value for this KPI may be a "10%" increase.

[0082] In step 504, the deep learning processor 118 may simulate the manufacturing process to generate expected process values ​​and predicted targets for a given KPI or a given set of KPIs. For example, the deep learning processor 118 may simulate the manufacturing process using a set of set values, algorithms, initial inputs, and / or operational instructions that would typically be provided to the station control system 104 from the factory data system 102. The deep learning processor 118 may generate the set of set values, algorithms, initial inputs, and / or operational instructions for the simulation. In some embodiments, the generated set of set values, algorithms, initial inputs, and / or operational instructions may include state changes to the manufacturing process. The state changes may be changes to one or more parameters of a process station or set of process stations as compared to “normal” or existing manufacturing parameters. Using the set of set values, algorithms, initial inputs, and / or operational instructions, the deep learning processor 118 may simulate the manufacturing process to generate predicted targets for the set of expected process values ​​and KPIs.

[0083] In step 506, the deep learning processor 118 may determine whether the expected target values ​​for the set of expected process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values. For example, the deep learning processor 118 may compare the desired values ​​for the process values ​​and the desired target values ​​for the set of KPIs to the expected targets for the expected process values ​​and given KPIs.

[0084] If, in step 506, the deep learning processor 118 determines that the expected target values ​​for the set of expected process values ​​and KPIs are not within the tolerance limits of the desired process values ​​and desired target values, the method 500 may return to step 504, and the deep learning processor 118 may implement state changes to simulate the manufacturing process.

[0085] However, if at step 506 the deep learning processor 118 determines that the expected target values ​​for the set of expected process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values, then the method 500 may proceed to step 508. At step 508, the deep learning processor 118 may determine whether the state change is autonomous. In other words, the deep learning processor 118 may determine whether the state change can be performed with or without user interaction.

[0086] If, at step 508, the deep learning processor 118 determines that the state change is not autonomous, then the method 500 may proceed to step 510. At step 510, the deep learning processor 118 provides the recommended state change to an end user or administrator. If, at step 508, the deep learning processor 118 determines that the state change is autonomous, then the method 500 may proceed to step 512. At step 512, the deep learning processor 118 may provide the recommended state change to the factory data system 102 and / or the station control system 104.

[0087] 6 is a flow diagram illustrating a method 600 for generating recommended state changes for a manufacturing process, according to an example embodiment. Method 600 may begin at step 602.

[0088] In step 602, the deep learning processor 118 may receive a process value and a desired value for a KPI target. In some embodiments, the desired value for the process value may include, but is not limited to, a desired station value, an intermediate output value, a final output value, a control value, an independent sensor value, or generally any other data that may be associated with a process station of the manufacturing system 100. In some embodiments, the desired KPI target may be a desired target value for a given KPI. For example, for a "throughput increase" KPI, the desired target value for this KPI may be a "10%" increase.

[0089] In step 604, the deep learning processor 118 may simulate the manufacturing process to generate expected process values ​​and predicted targets for a given KPI or a given set of KPIs. For example, the deep learning processor 118 may simulate the manufacturing process using a set of set values, algorithms, initial inputs, and / or operational instructions that would typically be provided to the station control system 104 from the factory data system 102. The deep learning processor 118 may generate the set of set values, algorithms, initial inputs, and / or operational instructions for the simulation. In some embodiments, the generated set of set values, algorithms, initial inputs, and / or operational instructions may include state changes to the manufacturing process. The state changes may be changes to one or more parameters of a process station or set of process stations as compared to “normal” or existing manufacturing parameters. Using the set of set values, algorithms, initial inputs, and / or operational instructions, the deep learning processor 118 may simulate the manufacturing process to generate predicted targets for the set of expected process values ​​and KPIs.

[0090] During such a simulation, the deep learning processor 118 may apply state changes to the manufacturing process, which may be changes to one or more parameters of a process station or set of process stations as compared to "normal" or existing manufacturing parameters.

[0091] In step 606, the deep learning processor 118 may determine whether the expected target values ​​for the set of expected process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values. For example, the deep learning processor 118 may compare the desired values ​​for the process values ​​and the desired target values ​​for the set of KPIs to the expected targets for the expected process values ​​and given KPIs.

[0092] If, in step 606, the deep learning processor 118 determines that the predicted target values ​​for the set of predicted process values ​​and KPIs are not within the tolerance limits of the desired process values ​​and desired target values, the method 600 may return to step 604, and the deep learning processor 118 may implement state changes to simulate the manufacturing process.

[0093] However, if in step 606 the deep learning processor 118 determines that the predicted target values ​​for the set of predicted process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values, then the method 600 may proceed to step 608.

[0094] In step 608, the deep learning processor 118 may determine whether the state change is within the equipment limits of the manufacturing system 100. Typically, each piece of manufacturing equipment may have a set of specifications that define upper and lower limits for various attributes of the equipment (e.g., temperature, speed, etc.). For example, if a proposed state change is to increase the temperature of a process station to 400°C and the temperature limit for the process station is 200°C, it would not be possible to implement that state change on the manufacturing process.

[0095] If, at step 608, the deep learning processor 118 determines that the state change is not within the equipment limits of the manufacturing system 100, the method 600 may return to step 604, and the deep learning processor 118 may implement the state change to simulate the manufacturing process. However, if, at step 608, the deep learning processor 118 determines that the state change is within the equipment limits of the manufacturing system 100, the method 600 may proceed to step 610.

[0096] In step 610, the deep learning processor 118 may determine whether the expected values ​​(e.g., expected control values ​​and / or expected target values ​​for a set of KPIs) are within user specifications. For example, as described above, the user specifications may refer to user specifications for each process station. Such user specifications may be a subset of equipment specifications. For example, a press may have an operating range of 100°C to 200°C, but the user specifications may be narrower, such as 125°C to 150°C.

[0097] If, at step 610, the deep learning processor 118 determines that the expected value is not within the user specifications, then the method 600 may return to step 604, and the deep learning processor 118 may implement state changes to simulate the manufacturing process. However, if, at step 610, the deep learning processor 118 determines that the expected value is within the user specifications, then the method 600 may proceed to step 612.

[0098] At step 612, the deep learning processor 118 may determine whether the expected value is within statistical process control. For example, if the expected value is greater than + / - 3 sigma from the average process value, the expected value may be considered to be out of control. If at step 612, the deep learning processor 118 determines that the expected value is not within statistical process control, the method 600 may return to step 604, and the deep learning processor 118 may implement state changes to simulate the manufacturing process. However, if at step 610, the deep learning processor 118 determines that the expected value is within statistical process control, the method 600 may proceed to step 614.

[0099] In step 614, the deep learning processor 118 may accept the state change. In some embodiments, accepting the state change may include the deep learning processor 118 providing a recommended state change to an end user or administrator. In some embodiments, accepting the state change may include the deep learning processor 118 providing a recommended state change to the plant data system 102 and / or the station control system 104.

[0100] 7 is a flow diagram illustrating a method 700 for generating recommended state changes for a manufacturing process, according to an example embodiment. Method 700 may begin at step 702.

[0101] In step 702, the deep learning processor 118 may receive a process value and a desired value for a KPI target. In some embodiments, the desired value for the process value may include, but is not limited to, a desired station value, an intermediate output value, a final output value, a control value, an independent sensor value, or generally any other data that may be associated with a process station of the manufacturing system 100. In some embodiments, the desired KPI target may be a desired target value for a given KPI. For example, for a "throughput increase" KPI, the desired target value for this KPI may be a "10%" increase.

[0102] In step 704, the deep learning processor 118 may simulate the manufacturing process to generate expected process values ​​and predicted targets for a given KPI or a given set of KPIs. For example, the deep learning processor 118 may simulate the manufacturing process using a set of set values, algorithms, initial inputs, and / or operational instructions that would typically be provided to the station control system 104 from the factory data system 102. The deep learning processor 118 may generate the set of set values, algorithms, initial inputs, and / or operational instructions for the simulation. In some embodiments, the generated set of set values, algorithms, initial inputs, and / or operational instructions may include state changes to the manufacturing process. The state changes may be changes to one or more parameters of a process station or set of process stations as compared to “normal” or existing manufacturing parameters. Using the set of set values, algorithms, initial inputs, and / or operational instructions, the deep learning processor 118 may simulate the manufacturing process to generate predicted targets for the set of expected process values ​​and KPIs.

[0103] In step 706, the deep learning processor 118 may determine whether the expected target values ​​for the set of expected process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values. For example, the deep learning processor 118 may compare the desired values ​​for the process values ​​and the desired target values ​​for the set of KPIs to the expected targets for the expected process values ​​and given KPIs.

[0104] If, in step 706, the deep learning processor 118 determines that the expected target values ​​for the set of expected process values ​​and KPIs are not within the tolerance limits of the desired process values ​​and desired target values, the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process.

[0105] However, if in step 706 the deep learning processor 118 determines that the predicted target values ​​for the set of predicted process values ​​and KPIs are within tolerance limits of the desired process values ​​and desired target values, then the method 700 may proceed to step 708.

[0106] In step 708, the deep learning processor 118 may determine whether the state change is within the equipment limits of the manufacturing system 100. Typically, each piece of manufacturing equipment may have a set of specifications that define upper and lower limits for various attributes of the equipment (e.g., temperature, speed, etc.). For example, if a proposed state change is to increase the temperature of a process station to 400°C and the temperature limit for the process station is 200°C, it would not be possible to implement that state change on the manufacturing process.

[0107] If, at step 708, the deep learning processor 118 determines that the state change is within the equipment limits, then the method 700 may proceed to step 714, which is discussed in more detail below. However, if, at step 708, the deep learning processor 118 determines that the state change is not within the equipment limits of the manufacturing system 100, then the method 700 may proceed to step 710.

[0108] In step 710, the deep learning processor 118 may determine whether equipment limits are variable. For example, in general, such equipment specifications may be considered "unchangeable." In other words, equipment specifications may be fixed and cannot be changed. In some embodiments, some equipment specifications may be variable. For example, a manufacturer may set minimum and maximum screw speeds, but the manufacturer may adjust those minimum and maximum speeds to exceed the initial minimum and maximum values.

[0109] If, at step 710, the deep learning processor 118 determines that the equipment limits are not variable, then the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process. However, if, at step 710, the deep learning processor 118 determines that the equipment limits are variable, then the method 700 may proceed to step 712.

[0110] At step 712, the deep learning processor 118 may determine whether the user accepts the recommended state change. For example, at step 712, the deep learning processor 118 may inform the user that the state change requires a change to an equipment limit. If at step 712, the deep learning processor 118 determines that the user does not accept the recommended state change, the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process. However, if at step 712, the deep learning processor 118 determines that the user does accept the recommended state change, the method 700 may proceed to step 714.

[0111] In step 714, the deep learning processor 118 may determine whether the expected values ​​(e.g., expected control values ​​and / or expected target values ​​for a set of KPIs) are within user specifications. For example, as described above, the user specifications may refer to user specifications for each process station. Such user specifications may be a subset of equipment specifications. For example, a press may have an operating range of 100°C to 200°C, but the user specifications may be narrower, such as 125°C to 150°C.

[0112] If, at step 714, the deep learning processor 118 determines that the expected value is within the user specifications, then the method 700 may proceed to step 720. However, if, at step 714, the deep learning processor 118 determines that the expected value is not within the user specifications, then the method 700 may proceed to step 716.

[0113] At step 716, the deep learning processor 118 may determine whether the user specifications are variable. If at step 716, the deep learning processor 118 determines that the user specifications are not variable, the method 700 may return to step 704, and the deep learning processor 118 may implement different state changes to simulate the manufacturing process. However, if at step 716, the deep learning processor 118 determines that the user specifications are variable, the method 700 may proceed to step 718.

[0114] At step 718, the deep learning processor 118 may determine whether the user accepts the recommended state change. For example, at step 718, the deep learning processor 118 may inform the user that the state change requires a modification to the user specifications. If at step 718, the deep learning processor 118 determines that the user does not accept the recommended state change, the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process. However, if at step 718, the deep learning processor 118 determines that the user does accept the recommended state change, the method 700 may proceed to step 720.

[0115] At step 720, the deep learning processor 118 may determine whether the expected value is within statistical process control. If at step 720, the deep learning processor 118 determines that the expected value is within statistical process control, the method 700 may proceed to step 726. However, if at step 720, the deep learning processor 118 determines that the expected value is not within statistical process control, the method 700 may proceed to step 722.

[0116] At step 722, the deep learning processor 118 may determine whether the statistical process control is variable. If at step 722, the deep learning processor 118 determines that the statistical process control is not variable, the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process. However, if at step 722, the deep learning processor 118 determines that the statistical process control is variable, the method 700 may proceed to step 724.

[0117] At step 724, the deep learning processor 118 may determine whether the user accepts the recommended state change. For example, at step 724, the deep learning processor 118 may inform the user that the state change requires a change to the statistical process control. If at step 724, the deep learning processor 118 determines that the user does not accept the recommended state change, the method 700 may return to step 704, and the deep learning processor 118 may implement a different state change to simulate the manufacturing process. However, if at step 724, the deep learning processor 118 determines that the user does accept the recommended state change, the method 700 may proceed to step 726.

[0118] In step 726, the deep learning processor 118 may accept the state change. In some embodiments, accepting the state change may include the deep learning processor 118 providing a recommended state change to an end user or administrator. In some embodiments, accepting the state change may include the deep learning processor 118 providing a recommended state change to the factory data system 102 and / or the station control system 104.

[0119] 8 illustrates an example dashboard 800 according to an example embodiment. Dashboard 800 may be generated by deep learning processor 118. Dashboard 800 may visually illustrate a comparison of various predicted values ​​(i.e., values ​​generated by deep learning processor 118) and actual values. For example, dashboard 800 may provide one or more of historical data for each process station (a date range may be specified), real-time data for each process station (e.g., actual values ​​for each process station), predicted values ​​for each station, actual KPI targets, and predicted KPI targets.

[0120] FIG. 9A illustrates the architecture of a computing system 900 according to an exemplary embodiment. The system 900 may represent at least a portion of a deep learning processor 118. One or more components of the system 900 may be in electrical communication with each other using a bus 905. The system 900 may include a processing unit (CPU or processor) 910 and a system bus 905 coupling various system components to the processor 910, including system memory 915 such as read-only memory (ROM) 920 and random access memory (RAM) 925. The system 900 may include a cache of high-speed memory directly connected to the processor 910, located in close proximity to the processor 910, or integrated as part of the processor 910. The system 900 may copy data from the memory 915 and / or storage device 930 to a cache 912 for quick access by the processor 910. In this manner, the cache 912 may provide a performance boost that prevents the processor 910 from being delayed while waiting for data. These and other modules may control the processor 910 to perform various operations or may be configured to control the processor 910 to perform various operations. Other system memory 915 may be available. The memory 915 may include multiple different types of memory with different performance characteristics. The processor 910 may include any general-purpose processor, as well as hardware or software modules, such as service 1 932, service 2 934, and service 3 936 stored in storage 930, configured to control the processor 910, and special-purpose processors with software instructions embedded in the actual processor design. The processor 910 may essentially be a completely self-contained computing system incorporating multiple cores or processors, buses, memory controllers, caches, etc. Multi-core processors may be symmetric or asymmetric.

[0121] To enable a user to interact with the computing system 900, the input device(s) 945 may represent any number of input mechanisms, such as a microphone for audio, a touch-sensitive screen for gesture or graphic input, a keyboard, a mouse, motion input, voice, etc. The output device(s) 935 (e.g., a display) may also be one or more of several output mechanisms known to those skilled in the art. In some examples, a multimodal system may allow a user to provide multiple types of input to communicate with the computing system 900. The communications interface 940 may generally control and manage user input and system output. It is not limited to operating on any particular hardware device, and therefore, as improved hardware or firmware devices are developed, the basic features herein may be readily substituted for those arrangements.

[0122] Storage device 930 may be non-volatile memory, such as a hard disk or other type of computer-readable medium that may store data that can be accessed by a computer, such as a magnetic cassette, a flash memory card, a solid-state memory device, a digital versatile disk, a cartridge, random access memory (RAM) 925, read-only memory (ROM) 920, and hybrids thereof.

[0123] The storage device 930 may include services 932, 934, and 936 for controlling the processor 910. Other hardware or software modules are also contemplated. The storage device 930 may be connected to the system bus 905. In one aspect, a hardware module that performs a particular function may include software components stored on a computer-readable medium along with the necessary hardware components, such as the processor 910, the bus 905, and the output device(s) 935, to perform that function.

[0124] FIG. 9B illustrates a computer system 950 having a chipset architecture, which may represent at least a portion of the deep learning processor 118, according to an exemplary embodiment. The computer system 950 may be one example of computer hardware, software, and firmware that may be used to implement the disclosed techniques. The system 950 may include a processor 955, which represents any number of physically and / or logically distinct resources capable of executing software, firmware, and hardware configured to perform the identified computations. The processor 955 may communicate with a chipset 960, which may control input to and output from the processor 955. In this example, the chipset 960 outputs information to an output 965, such as a display, and the chipset 960 may read and write information to a storage device 970, which may include, for example, magnetic and solid-state media. The chipset 960 may also read data from and write data to a RAM 975. A bridge 980 may be provided to interface with the chipset 960 for interfacing with various user interface components 985. Such user interface components 985 may include a keyboard, a microphone, touch detection and processing circuitry, a pointing device such as a mouse, etc. In general, input to system 950 may come from any of a variety of machine- and / or human-generated sources.

[0125] Chipset 960 may further connect to one or more communication interfaces 990, which may have different physical interfaces. Such communication interfaces may include interfaces for wired and wireless local area networks, broadband wireless networks, and personal area networks. Some applications of the methods for generating, displaying, and using GUIs disclosed herein may involve receiving an ordered data set across a physical interface, or may be generated by the machine itself by processor 955 analyzing data stored in storage device 970 or RAM 975. Additionally, the machine may receive inputs from a user through user interface components 985 and perform appropriate functions, such as browsing functions, by interpreting these inputs using processor 955.

[0126] It may be understood that the exemplary systems 900 and 950 may have more than one processor 910, or that the exemplary systems 900 and 950 may be part of a group or cluster of computing devices networked together to provide greater processing power.

[0127] While the foregoing description is directed to the embodiments described herein, other and additional embodiments may be devised without departing from the basic scope of those embodiments. For example, aspects of the present disclosure may be implemented in hardware or software, or a combination of hardware and software. One embodiment described herein may also be implemented as a program product for use with a computer system. The program of the program product defines the functions of the embodiments (including the methods described herein) and can be stored on various computer-readable storage media. Exemplary computer-readable storage media include, but are not limited to, (i) non-writable storage media on which information is permanently stored (e.g., a read-only memory (ROM) device in a computer, such as a CD-ROM disk readable by a CD-ROM drive, flash memory, a ROM chip, or any type of non-volatile solid-state memory), and (ii) writable storage media on which changeable information is recorded (e.g., a floppy disk or hard disk drive in a diskette drive, or any type of solid-state random access memory). Such computer-readable storage media, when containing computer-readable instructions that direct the functions of the disclosed embodiments, are embodiments of the present disclosure.

[0128] Those skilled in the art will understand that the foregoing examples are illustrative and not limiting. All permutations, enhancements, equivalents, and improvements of these examples will be apparent to those skilled in the art upon reading this specification and studying the drawings, and it is intended that such permutations, enhancements, equivalents, and improvements fall within the true spirit and scope of this disclosure. Accordingly, the following appended claims are intended to cover all such modifications, permutations, and equivalents that fall within the true spirit and scope of these teachings.

Claims

1. one or more process stations configured to perform a manufacturing process, the manufacturing process including an initial set of processing parameters; a station control system configured to control the one or more process stations; a controller in communication with the one or more process stations and the station control system; wherein the controller is configured to perform an operation, the operation comprising: receiving a desired process value associated with the one or more process stations; receiving desired target values ​​for one or more key performance indicators of the manufacturing process; simulating, with a deep learning processor, the manufacturing process to generate predicted process values ​​and predicted target values ​​for the one or more key performance indicators for optimizing the one or more key performance indicators, wherein the simulating includes generating proposed state changes for at least one process parameter of the initial set of process parameters; determining, by the deep learning processor, that the predicted process value and the predicted target value are within tolerance limits of the desired process value and the desired target value; causing a change to the initial set of processing parameters based on the proposed state change based on the determination; and a manufacturing system.

2. determining whether the proposed state change is within equipment limits of the manufacturing system; The manufacturing system of claim 1 further comprising:

3. causing the changes to the initial set of process parameters based on the proposed state changes upon determining that the proposed state changes are within the instrument limits. The manufacturing system of claim 2 further comprising:

4. determining whether the instrument limits are variable when determining that the proposed state change is not within the instrument limits; causing the change to the initial set of process parameters based on the proposed state change upon determining that the equipment limits are variable; and The manufacturing system of claim 2 further comprising:

5. determining whether the proposed state change is within a user specification of the manufacturing system; The manufacturing system of claim 1 further comprising:

6. causing the changes to the initial set of processing parameters based on the proposed state changes upon determining that the proposed state changes are within the user specifications. The manufacturing system of claim 5 further comprising:

7. determining whether the user specification is variable when determining that the proposed state change is not within the user specification; causing the changes to the initial set of processing parameters based on the proposed state changes when determining that the user specifications are variable; and The manufacturing system of claim 5 further comprising:

8. determining whether the proposed state change is within statistical process control; The manufacturing system of claim 1 further comprising:

9. causing the changes to the initial set of process parameters based on the proposed state changes upon determining that the proposed state changes are within the statistical process control. The manufacturing system of claim 8 further comprising:

10. determining whether the statistical process control is variable when determining that the proposed state change is not within the statistical process control; causing the change to the initial set of process parameters based on the proposed state change when the statistical process control determines that a change is present; and The manufacturing system of claim 8 further comprising:

11. receiving, by a computing system, desired process values ​​associated with one or more process stations, the one or more process stations configured to perform a manufacturing process, the manufacturing process including an initial set of processing parameters; receiving, by the computing system, desired target values ​​for one or more key performance indicators of the manufacturing process; simulating, by a deep learning processor of the computing system, the manufacturing process to generate predicted process values ​​and predicted target values ​​for the one or more key performance indicators for optimizing the one or more key performance indicators, wherein the simulating includes generating proposed state changes for at least one process parameter of the initial set of process parameters; determining, by the deep learning processor of the computing system, that the predicted process value and the predicted target value are within tolerance limits of the desired process value and the desired target value; causing, based on the determination, changes to the initial set of processing parameters based on the proposed state change by the computing system; 20. A computer-implemented method comprising:

12. determining whether the proposed state change is within equipment limits of the one or more process stations; The computer-implemented method of claim 11 further comprising:

13. causing the changes to the initial set of process parameters based on the proposed state changes upon determining that the proposed state changes are within the instrument limits. The computer-implemented method of claim 12 further comprising:

14. determining whether the instrument limits are variable when determining that the proposed state change is not within the instrument limits; causing the change to the initial set of process parameters based on the proposed state change upon determining that the equipment limits are variable; and The computer-implemented method of claim 12 further comprising:

15. determining whether the proposed state change is within user specifications for the manufacturing process; The computer-implemented method of claim 11 further comprising:

16. causing the changes to the initial set of processing parameters based on the proposed state changes upon determining that the proposed state changes are within the user specifications. The computer-implemented method of claim 15 further comprising:

17. determining whether the user specification is variable when determining that the proposed state change is not within the user specification; causing the changes to the initial set of processing parameters based on the proposed state changes when determining that the user specifications are variable; and The computer-implemented method of claim 15 further comprising:

18. determining whether the proposed state change is within statistical process control; The computer-implemented method of claim 11 further comprising:

19. causing the changes to the initial set of process parameters based on the proposed state changes upon determining that the proposed state changes are within the statistical process control.

20. The computer-implemented method of claim 18, further comprising:

20. receiving, by the computing system, actual output values ​​corresponding to the manufacturing process of the manufacturing system; said receiving receiving stability data corresponding to normal production data during a normal production phase of the manufacturing process; and receiving instability data corresponding to abnormal production data outside the normal production stage of the manufacturing process; generating, by the computing system, a training data set including the stable data and the unstable data; training, by the computing system, a machine learning model to generate predicted process values ​​and predicted target values ​​for one or more key performance indicators; determining, by the computing system, that the machine learning model has achieved a threshold level of accuracy; deploying, by the computing system, the machine learning model within the manufacturing system in response to the determining.

20. A computer-implemented method comprising:

Images