Predictive Process Management for Manufacturing Processes

Abstract

To provide a method and system for utilizing a deep learning controller for monitoring and improving a manufacturing process.SOLUTION: In some aspects, a method of the disclosed technique includes the steps of: receiving a plurality of control values from two or more stations at a deep learning controller, wherein the control values are generated at the two or more stations deployed in a manufacturing process; predicting an expected value for an intermediate or final output of an article of manufacture based on the control values; and determining if the predicted expected value for the article of manufacture is in-specification. In some aspects, the process can further include a step of generating control inputs if the predicted expected value for the article of manufacture is not in-specification. Systems and computer-readable media are also provided.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Cross - reference to Related Applications This application claims priority to U.S. Patent Application No. 16 / 519,102, entitled "PREDICTIVE PROCESS CONTROL FOR A MANUFACTURING PROCESS," filed on July 23, 2019, which claims the benefit of U.S. Provisional Patent Application No. 62 / 865,859, entitled "SYSTEMS, APPARATUS AND METHODS FOR PREDICTIVE PROCESS CONTROL OF THE MANUFACTURING PROCESS," filed on June 24, 2019. The entire contents of these are hereby incorporated by reference in their entirety into this specification.

Background Art

[0002] 1. Technical Field The present disclosure generally relates to systems, apparatuses, and methods for predictive process control (PPC) of manufacturing processes. More particularly, the technology provides improvements to manufacturing processes and, in particular, includes systems and methods for adaptively managing various stations in a manufacturing process based on predictions made using machine learning models and for optimizing final manufactured products and processes. As will be described in more detail below, some aspects of the technology include systems and methods for training machine learning models.

Summary of the Invention

Problems to be Solved by the Invention

[0003] 2. Introduction Manufacturing products that always meet desired design specifications safely, in a timely manner, and with minimal waste requires a certain amount of monitoring and adjustment of the manufacturing process.

Means for Solving the Problems

[0004] In some aspects, the disclosed technology relates to the use of a deep learning controller for monitoring and improving manufacturing processes. In some embodiments, the disclosed technology includes a computer-implemented method comprising: receiving, by a deep learning controller, a plurality of control values from two or more stations, the control values being generated at two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an intermediate or final output of a manufactured article based on the control values; determining, by the deep learning controller, whether the predicted intermediate or final output specification for the manufactured article is within specifications; and generating, by the deep learning controller, one or more control inputs when the predicted intermediate or final output for the manufactured article is not within specifications, the one or more control inputs being configured to cause the intermediate or final output for the manufactured article to be within specifications.

[0005] In another embodiment, the disclosed technology is a system comprising one or more processors and a non-transitory memory storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising: receiving, by a deep learning controller, a plurality of control values from two or more stations, the control values being generated at two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an intermediate or final output of a manufactured article based on the control values; determining, by the deep learning controller, whether the predicted intermediate or final output specification for the manufactured article is within specifications; and generating, by the deep learning controller, one or more control inputs when the predicted intermediate or final output for the manufactured article is not within specifications, the one or more control inputs being configured to cause the intermediate or final output for the manufactured article to be within specifications.

[0006] In yet another embodiment, the disclosed technology is a non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations, the operations including receiving, by a deep learning controller, a plurality of management values from two or more stations, wherein the management values are generated at the two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an intermediate or final output of a manufactured article based on the management values; determining, by the deep learning controller, whether the predicted intermediate or final output of the manufactured article is within specifications; and generating, by the deep learning controller, one or more management inputs when the predicted intermediate or final output of the manufactured article is not within specifications, wherein the one or more management inputs are configured to cause the intermediate or final output of the manufactured article to be within specifications.

Brief Description of the Drawings

[0007] To explain the manner in which the above and other advantages and features of the present disclosure can be obtained, a more detailed description of the principles briefly described above will be given by reference to the specific embodiments shown in the accompanying drawings. It is to be understood that these drawings illustrate only exemplary embodiments of the present disclosure and should not be considered limiting with respect to its scope, and that the principles described herein will be described and explained with additional specificity and detail using the accompanying drawings.

[0008]

Figure 1

[0009]

Figure 2

[0010]

Figure 3

[0011]

Figure 4

[0012]

Figure 5

[0013]

Figure 6

[0014]

Figure 7

[0015]

Figure 8

[0016]

Figure 9

Best Mode for Carrying Out the Invention

[0017] The following detailed description is intended to explain various configurations of the present technology and is not intended to represent the only configuration in which the present technology can be implemented. The accompanying drawings are incorporated into this specification and form part of the detailed description. The detailed description includes specific details for the purpose of providing a more thorough understanding of the present technology. However, it is clear and obvious that the present technology is not limited to the specific details described herein and can be implemented without these details. In some examples, structures and components are shown in block diagram form to avoid obscuring the concept of the present technology.

[0018] Additional details regarding the use of image classification in manufacturing management are provided by U.S. Provisional Patent Application No. 62 / 836,202, entitled "DEPLOYMENT OF AN IMAGE CLASSIFICATION MODEL ON A SINGLE-BOARD COMPUTER FOR MANUFACTURING CONTROL", which is hereby incorporated by reference in its entirety.

[0019] Additional details regarding users of computational models for optimizing assembly / manufacturing operations are provided by U.S. Provisional Patent Application No. 62 / 836,192, entitled "A COMPUTATION MODEL FOR DECISION-MAKING AND ASSEMBLY OPTIMIZATION IN MANUFACTURING", and U.S. Patent Application No. 16 / 289,422, entitled "DYNAMIC TRAINING FOR ASSEMBLY LINES", both of which are hereby incorporated by reference in their entirety.

[0020] Manufacturing processes are complex and include the processing of raw materials by different process stations (or "stations") until a final product (referred to herein as the "final output") is produced. Except for the final process station, each process station receives an input for processing and outputs an intermediate output, which is passed to a subsequent (downstream) process station for additional processing. The final process station receives an input for processing and outputs the final output.

[0021] Each process station may include one or more tools / devices that perform a set of process steps on the received raw materials (which are applicable to either the first station or any subsequent station within the manufacturing process) and / or the output received from the previous station (which is applied to any subsequent station within the manufacturing process). Examples of process stations may include, but are not limited to, conveyor belts, injection molding machines, cutting machines, die pressing machines, extrusion molding machines, CNC milling machines, grinding machines, assembly stations, 3D printers, quality control and verification stations. Exemplary process steps may include transporting the output from one location to another (as performed by a conveyor belt); feeding the material to an extrusion molding machine, melting the material, injecting the material through a mold cavity, and cooling and curing the material in the form of the cavity there (as performed by an injection molding machine); cutting the material to a specific shape or length (as performed by a cutting machine); and pressing the material into a specific shape (as performed by a die pressing machine).

[0022] In some manufacturing processes, several process stations may operate in parallel. In other words, a single process station can send its intermediate output to one or more stations (e.g., 1 to N stations), and a single process station can receive and combine intermediate outputs from 1 to n stations. Further, a single process station can perform the same or different process steps sequentially or non-sequentially on the raw materials or intermediate outputs received during a single iteration of the manufacturing process.

[0023] The operation of each process station can be controlled by one or more process controllers. In one implementation, each process station has one or more process controllers (referred to herein as "station controllers") programmed to manage the operation of that process station (the programming algorithm thereof is referred to herein as the "management algorithm"). However, in some embodiments, a single process controller may be configured to manage the operation of two or more process stations.

[0024] The operator, or the management algorithm, can provide the station controller with station controller setpoints (or "setpoints" or "controller setpoints" or CSPs (controller setpoints)) representing the desired value or range of values for each management value. Attributes / parameters associated with the tools / instruments / process steps of the station that can be measured during the operation of the station are either management values or station values. Measured attributes / parameters are also "management values" if they are used to manage the station. Otherwise, the measured attributes / parameters are "station values". Examples of management values or station values include, but are not limited to, speed, temperature, pressure, vacuum, rotation, current, voltage, power, viscosity, materials / resources used at the station, throughput rate, downtime, toxic gases, type of step executed at the station, and order of steps. The examples are the same, but whether a measured attribute / parameter is considered a management value or a station value depends on the individual station and whether the measured attribute / parameter is used to manage the station or is merely a byproduct of the operation of the station.

[0025] The management algorithm may also include instructions for monitoring a management value, comparing the management value to a corresponding set point, and determining what action should be taken when the management value is not equal to (or not within a predetermined range of) the corresponding station controller set point. For example, if the measured current value of the temperature for a station is lower than the set point, a signal may be sent by the station controller to increase the temperature of the heat source for the station until the current value of the temperature for the station is equal to the set point. Conventional process controllers used within a manufacturing process to manage stations are limited because they follow static algorithms (e.g., on / off control, PI control, PID control, lead / lag control) for adjusting set points and defining what action should be taken when a management value deviates from a set point. Conventional process controllers also have limited, if any, ability to analyze non-management values such as ambient conditions (e.g., external temperature, humidity, exposure, wear of the station), station values, intermediate or final output values, feedback from other process stations, and make dynamic adjustments to the set points of the station controller or management algorithm that manages the operation of the associated station.

[0026] As used herein, a process value refers to a station value or management value that has been aggregated or averaged across an entire series of stations (or a subset of stations) that make up a manufacturing process. Process values can include, for example, total throughput time, total resources used, average temperature, average speed.

[0027] In addition to the station values and process values, various characteristics of the product output of the process station (i.e., intermediate output or final output) 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. The various measurable characteristics are generally referred to as "intermediate output values" or "final output values". The intermediate / final output value can reflect an overall score based on a single measured characteristic of the intermediate / final output value or a specified set of characteristics associated with the intermediate / final output value measured and quantified according to a predetermined formula.

[0028] Mechanical properties can include hardness, compression, adhesiveness, density, and weight. Optical properties can include absorption, reflection, transmission, and refraction. Electrical properties can include electrical resistivity and conductivity. Chemical properties can include enthalpy of formation, toxicity, chemical stability in a given environment, flammability, preferred oxidation state, pH (acidity / alkalinity), chemical composition, boiling point, and vapor point. The disclosed mechanical, optical, chemical, and electrical properties are merely examples and are not intended to be limiting.

[0029] As shown in FIG. 1, the values for the intermediate output of a process station and the values for the final output produced by a manufacturing process can be evaluated according to statistical process control (SPC). Using statistics, SPC tracks, for each station, (1) the values of the intermediate or final output generated by the station over a period of time, and (2) the average of the values of the intermediate or final output and the standard deviation from the average. For example, FIG. 1 shows the intermediate output values for a particular station. Each dot represents the intermediate output produced by the station and its value, and the distance of the dot from the average (represented by the black line in the middle) indicates how much the intermediate output value deviates from the average for that particular station. Upper and lower control limits (UCL: upper control limit and LCL: lower control limit) can be defined for a particular station (e.g., one or more standard deviations above or below the average). FIG. 1 shows that the upper and lower control limits are set at three standard deviations above and below the average (represented by the dashed lines). The upper and lower control limits are typically narrower than the upper and lower specification limits (USL: upper specification limit and LSL: lower specification limit) defined for the intermediate / final output values.

[0030] In some aspects, statistical values for the intermediate output and / or final output can be used to determine when a manufactured article is "in-specification," i.e., when the output meets certain pre-specified design requirements. Thus, in-specification can refer to a manufactured article or a characteristic of a manufactured article that meets or exceeds a specified set of design requirements. As an example, a manufactured article considered to be in-specification may be one that meets specified statistical requirements, such as having an acceptable deviation from an ideal or average (mean) value.

[0031] As long as the values for the intermediate / final output are within the range of the upper and lower control limits, the process station / overall process is considered to be in control, and typically, no intervention or corrective action is taken. Intervention or corrective action is typically taken when the value for the intermediate / final output exceeds the upper or lower control limit defined for the measurement. However, since SPC control only intervenes / corrects when the upper / lower control limit is exceeded, its effect on improving or optimizing the manufacturing process is limited. When the process is in control, adjustments are usually not made. Furthermore, SPC evaluates a single station independently and does not consider trends across multiple stations or the influence of several stations on the final product.

[0032] Therefore, there is a need for a new mechanism that can better manage the operation of related stations by highly dynamically adjusting the input to the station controller, considering the inputs and outputs of each station individually as well as the other inputs and outputs of other stations within the manufacturing process. In particular, there is a need for a new mechanism to predict the inputs that optimize the manufacturing process to produce a final output within specifications. There is also a need for a new mechanism to predict the inputs that improve the design and manufacturing process of the final output. Furthermore, there is a need for a new mechanism to reduce variations in the manufacturing process and the final output.

[0033] Aspects of the disclosed technology address the aforementioned limitations of conventional manufacturing processes by providing a mechanism (which may include systems, methods, devices, apparatuses, etc.) for incrementally improving manufacturing processes and the resulting manufactured products without interrupting ongoing manufacturing processes (referred to herein as predictive process management). Thus, the disclosed technology can be retrofitted and integrated with existing manufacturing systems and infrastructure without causing interruptions to ongoing manufacturing processes. The improvements are realized by providing dynamic management to one or more process stations within the manufacturing process via a conventional station controller in order to (1) steadily produce a final output within specifications, (2) optimize the design and manufacturing processes of the final output, and (3) reduce variations in the manufacturing process and the final output.

[0034] A deep learning controller based on a machine learning / artificial intelligence (AI) model may be used to evaluate the management / station / process values as well as the intermediate and final output values and determine adjustments to the inputs of the station controller. As will be understood by those skilled in the art, machine learning-based techniques may vary depending on the desired implementation without departing from the disclosed technology. For example, machine learning techniques can utilize one or more of the following, alone or in combination: hidden Markov models; regression neural networks; convolutional neural networks (CNNs); deep learning; Bayesian symbolic methods; reinforcement learning; generative adversarial networks (GANs); support vector machines; image registration methods; applicable rule-based systems.

[0035] The machine learning model can be based on clustering algorithms (e.g., mini-batch K-means clustering algorithm), recommendation algorithms (e.g., minwise hashing algorithm, or Euclidean locality-sensitive hashing (LSH) algorithm), and / or anomaly detection algorithms such as local outlier factor. The machine learning model can be based on supervised and / or unsupervised methods.

[0036] Also, using the machine learning model described herein, it is possible to determine the process stations, management / station / process values, and intermediate output values ("key influencers") that have the most influence on the final output value, and optimize the manufacturing process by targeting the key influencers.

[0037] FIG. 2 shows an exemplary deep learning controller 218 configured to manage any number (represented herein as "n") of processing stations in a manufacturing process using predictive process management, which will be described later in connection with FIGS. 4-7. In FIG. 2, the n processing stations of the manufacturing process are represented by process stations 222 and 242. The process stations can operate in series or in parallel. Universal input 236, empirical prior information 239, functional prior information 238, and inputs from each of the n stations (e.g., 222 and 242) can be provided to the deep learning controller 218.

[0038] The functional pre-information used in this specification refers to information regarding the functions and known limitations of each process station within a manufacturing process, both individually and collectively. All specifications for tools / devices used at a process station are considered as functional pre-information. Exemplary functional pre-information can include, but is not limited to, a screw-driven extrusion machine having a minimum and maximum speed at which a screw can rotate; a temperature control system having a highest and lowest temperature achievable based on its heating and cooling capabilities; a pressure vessel having a maximum pressure it can contain before bursting; a combustible liquid having a highest temperature achievable before combustion. Functional pre-information can also include the order in which individual stations making up a manufacturing process execute their functions. In some embodiments, several stations operate in parallel and their intermediate outputs can be combined at subsequent stations.

[0039] The empirical prior information used in this specification refers to information obtained from prior experience, for example, by performing the same or similar manufacturing processes; operating the same or similar stations; producing the same or similar intermediate / final outputs. In some embodiments, the empirical prior information may include acceptable final output values or unacceptable final output values. An acceptable final output value refers to the upper limit, lower limit, or range of final output values for which the final output is considered "within specifications". In other words, an acceptable final output value describes the parameters for a final output value that meets the design specifications, i.e., is within the specifications. Conversely, an unacceptable final output value refers to the upper limit / lower limit or range of final output values for which the final output is "out of specifications" (i.e., describes the parameters for a final output value that does not meet the design specifications). For example, based on prior experience, it may be known that a seal is made only when an O-ring used to seal a pipe has certain compression characteristics. This information can be used to determine the acceptable / unacceptable compression values for the O-ring final output. In other words, all O-ring final outputs with acceptable compression values can perform their sealing function, while all O-ring final outputs with unacceptable compression values cannot perform their sealing function. An acceptable intermediate output value can be defined for each station and refers to the upper limit / lower limit or range of intermediate output values that define the parameters for an intermediate output that can ultimately result in a final output within the specifications without the need for corrective action by other stations. An unacceptable intermediate output value can also be defined for each station and refers to the upper limit / lower limit or range of intermediate output values that define the parameters for an intermediate output that will ultimately result in a final output not within the specifications without corrective action being taken at another station. Similarly, acceptable / unacceptable parameters can be defined for other variables related to the manufacturing process:

Table 1

[0040] As used herein, universal input refers to values that are not specific to a particular process station but are specific to aspects of the entire manufacturing process, such as date, time, ambient temperature, humidity, or other environmental conditions that can affect the manufacturing process, operator, operator's skill level, raw materials used in the process, raw material specifications such as color, viscosity, particle size, etc. among the characteristics specific to the raw materials, unique lot numbers and costs of the raw materials, equipment / tool holding periods for each station, production work sequence numbers, batch numbers, lot numbers, finished product numbers, and finished product serial numbers, etc.

[0041] Note that the examples provided for each of the functional prior information, empirical prior information, and universal input represent one way to classify these examples, and it should be noted that other suitable classifications can be used. For example, another way to classify the inputs provided to the deep learning controller 218 is pre-process inputs (e.g., empirical prior information, functional prior information, material properties, scheduling requirements); in-process inputs (e.g., universal input, management values, station values, intermediate values, final output values, process values); post-process inputs (e.g., manufacturing performance metrics and other analyses).

[0042] Each process station can be managed by one or more associated station controllers (e.g., station controller 220 manages process station 222 and station controller 240 manages process station 242). In other embodiments, a single station controller can manage multiple process stations. The deep learning controller 218 can provide management inputs (represented by 226 and 246) based on predictive process management to each process station controller. In response to the received management inputs (e.g., 226 and 246), each station controller can provide one or more management signals (e.g., 221 and 241) that provide commands for adjusting the management values of the station (e.g., management values 225 and 245). Each station outputs an intermediate output (e.g., 224 and 244), and the intermediate output has an intermediate output value (234a and 244a, respectively). All intermediate output values and final output values from the processing stations are provided to the deep learning controller 218. Each station also outputs station values (e.g., 228 and 248) to the deep learning controller 218. FIG. 2 also shows that the intermediate output 224 is transmitted to one to n subsequent stations (step 250). The subsequent stations can represent a single station or multiple stations of n. The station 242 shown in FIG. 2 can receive an intermediate input from one to n previous stations (step 260).

[0043] As will be appreciated, the communication between the deep learning controller 218, the station controller, and the process station can use any suitable communication technology for communicating with one or more other devices and / or for exchanging data with a computer network. By way of example, the communication technology implemented can include, but is not limited to, analog technology (e.g., relay logic), digital technology (e.g., RS232, Ethernet, or wireless), network technology such as local area network (LAN), wide area network (WAN), Internet, Bluetooth technology, near field communication technology, secure RF technology, and / or any other suitable communication technology.

[0044] In some embodiments, operator input can be communicated to the deep learning controller 218 and / or to either the station controller or the process station using any suitable input device (e.g., keyboard, mouse, joystick, touch, touch screen, etc.).

[0045] In some embodiments, one or more process stations can be manually operated, for example, by a human operator who executes certain instructions. Instead of an electronic station controller, the operator follows a set of instructions, which can be provided manually or via electronic means (e.g., via a video or computer display). For example, at a manual station, the operator can perform functions such as cutting a wire to a specific length and measuring the length of the cut wire. Manual feedback such as the length of the cut wire can be provided to the deep learning controller 218. Using the predictive process management described herein, the deep learning controller 218 can determine whether the wire has been cut to the desired length specification and provide improvements to the cutting process, which can be provided, for example, in the form of a set of instructions to the operator of the manual station.

[0046] Figure 3 provides a process 300 for conditioning (training) the deep learning controller 218 according to some embodiments of the disclosed subject matter.

[0047] In step 310, set points, algorithms, and other administrative inputs for each station controller within the manufacturing process can be initialized using conventional methods. Further, an administrative algorithm / operator can provide initial administrative / station values. The administrative algorithm, initial set point values, and initial administrative / station values can be provided to the deep learning controller 218 (step 315).

[0048] Note that the administrative values, administrative algorithms, set points, and any other information provided to the station controller (e.g., process timing, equipment commands, alarm alerts, emergency stops) are collectively referred to as "station controller inputs" or "administrative inputs".

[0049] Further, other inputs such as functional prior information 238, empirical prior information 239, and universal input 236 can be provided to the deep learning controller 218.

[0050] In step 325, the manufacturing process is repeated through all process stations using conventional administrative methods. As described above, the process stations described herein can operate in series or in parallel. Further, a single station can execute a single process step multiple times (sequentially or non-sequentially), or different process steps (sequentially or non-sequentially) for a single iteration of the manufacturing process. The process stations generate intermediate outputs or, if a final station, a final output. The intermediate outputs are communicated to subsequent (downstream) stations within the manufacturing process until the final output is generated.

[0051] As the process is repeated through each station, all values associated with an individual station (e.g., control values); all values associated with the output of an individual station (e.g., station values, intermediate / final output values), or multiple stations (e.g., process values) are measured or calculated and provided to condition the machine learning algorithm of the deep learning controller 318 (steps 327 and 328).

[0052] In some embodiments, manufacturing performance metrics for a manufacturing process under conventional control (e.g., production volume over a specified period, production downtime over a specified period, resources used for a specified period or specified number of final outputs, percentage of products out of specification over a specified period, production volume for a particular operator, material costs associated with a specified number of final outputs) may be calculated and provided to the deep learning controller 218 (step 329).

[0053] Although not shown, any actions taken by the station controller in response to control values or other control inputs received from the process station may be provided to the deep learning controller 218. Such actions may include adjusting temperature, speed, etc. Additionally, deviations from acceptable set points, acceptable intermediate / final output values, acceptable control / station / process values may also be calculated and provided to the deep learning controller 218.

[0054] Note that all inputs to the deep learning controller 218 may be input electronically or via manual means by an operator.

[0055] The conditioning of the machine learning model of the deep learning controller 218 (step 235) can be achieved through unsupervised learning methods. In addition to the functional prior information 238, the empirical prior information 239, and the universal input 236 input to the deep learning controller 218, the deep learning controller 218 makes inferences by simply analyzing the received data collected during the repetition of the manufacturing process (e.g., steps 328 and 329). In other embodiments, the deep learning controller 218 can be conditioned via supervised learning methods, or a combination of supervised and unsupervised methods or similar machine learning methods. Further, the training of the deep learning controller 218 can be enhanced by providing the deep learning controller 218 with simulated data or data from similar manufacturing processes. In one embodiment, the deep learning controller 218 can be conditioned by implementing the deep learning controller 218 within a similar manufacturing process and fine-tuning the deep learning controller during implementation in the target manufacturing process. That is, the training of the deep learning controller 218 can be performed using a training process that is executed before the deep learning controller 218 is deployed in the target manufacturing process.

[0056] Based on the conditioning of the machine learning model, the deep learning controller 218 can predict a value (the "expected value" or "EV") for the characteristics of the final output that determines whether the final output value is acceptable (i.e., whether the final output is "within specifications") (step 342). The deep learning controller 218 can provide a level of confidence for its prediction at a point in time or over a specific period, for example, to provide a measure of the statistical reliability in the prediction. In some aspects, the level of confidence may be expressed as a numerical probability of the accuracy of the prediction, and in other aspects, the level of confidence may be expressed as an interval or a probability range. In step 343, the deep learning controller 218 can compare the expected value with the actual measured value (the "actual value" or "AV") of the specified characteristic of the final output.

[0057] In some embodiments, the deep learning controller 218 can be configured to perform EV predictions regarding output characteristics for each station. That is, the deep learning controller 218 can make EV predictions regarding the output at a particular station and then compare those predictions to the actual output observed at that station. Alternatively, the EV predictions can be made with respect to the output resulting from a combined process performed by two or more stations, depending on the desired implementation.

[0058] As the manufacturing process progresses through each station and the deep learning controller 218 receives additional information, the deep learning controller 218 can modify its expected value along with the confidence level. If the predictions of the deep learning controller 218 are correct at a predetermined threshold confidence level over a specified period, the deep learning controller 218 can provide a signal indicating that the deep learning controller 218 is ready to manage the operation of the process station.

[0059] In some embodiments, the deep learning controller 218 can also predict whether any control input will cause unsatisfactory station performance or affect process performance (i.e., cause unacceptable process performance) at the beginning of a repetition of the manufacturing process (i.e., as it progresses through all stations within the manufacturing process) after initialization of the station controller and throughout the course of the manufacturing process. The deep learning controller 218 can provide a confidence level for its prediction. The deep learning controller 218 can determine whether its prediction was correct. In further embodiments, if the predictions of the deep learning controller 218 are at a threshold confidence level defined by the operator over a specified period with respect to both the expected final output value and the predicted station / process performance, the deep learning controller 218 can provide a signal indicating that the deep learning controller 218 is ready to manage the operation of the process station.

[0060] Figure 4 shows an exemplary process for managing a manufacturing process using predictive process management.

[0061] The deep learning controller 218 uses its conditioned machine learning algorithm (described in connection with FIG. 3) to calculate a management input to the station controller associated with the process station of the manufacturing process. Based on the calculated management input, the deep learning controller 218 can predict the expected value (EV) of the final output for the manufacturing process, along with the reliability level for that prediction (step 405). If the deep learning controller 218 determines, at a threshold reliability level, that the expected value is within specifications (step 415), the deep learning controller 218 can output the calculated management input to the station controller associated with the process station of the manufacturing process (step 420). The deep learning controller 218 can calculate the management input at the beginning of, or continuously throughout, the manufacturing process and provide the calculated management input to one or more station controllers. The management input need not be provided to the station controllers in serial order and can be provided to one or more station controllers in parallel or in any order suitable for producing a final output within specifications. Each station controller that receives the management input from the deep learning controller 218 can transmit a management signal for managing the control value for its associated station (e.g., controlling the control value so that it matches the received set point). The machine learning algorithm continues to improve throughout the implementation of the PPC (step 335). Additionally, functional and empirical prior information can be dynamically updated throughout the PPC.

[0062] In step 430, the manufacturing process proceeds serially or in parallel through all process stations. As the process is repeated through each station, all values associated with an individual station (e.g., control values); all values associated with the output of an individual station (e.g., station values, intermediate / final output values), or multiple stations (e.g., process values) can be measured or calculated and provided to condition the machine learning algorithm of the deep learning controller 218 (step 432). Further, a manufacturing performance metric for the manufacturing process under predictive process control can be calculated and provided to the deep learning controller 218 (step 432). The process values and manufacturing performance metrics calculated under PPC can be compared with the process values and manufacturing performance metrics calculated under conventional control to determine the improvements provided by predictive process control.

[0063] Throughout the process shown in FIG. 4, the deep learning controller 218 can predict an expected value (EV) for the final output, determine whether the expected value for the final output is within specifications, determine a confidence level for that prediction, and then provide feedback for that prediction by comparing the expected final value to the actual final value (step 445). Further, if the deep learning controller 218 determines that the final output is not within specifications, it can calculate an adjustment to the control input such that the predicted expected value for the final output is within specifications.

[0064] In some aspects, the deep learning controller 218 may be configured to perform EV predictions for intermediate outputs for each station. That is, the deep learning controller 218 can perform EV predictions for the output at a particular station, determine whether the EV for the intermediate output is within specifications, determine a confidence level for the prediction, and then compare those predictions to the actual output observed at that station. Alternatively, the EV prediction can be made for an output that results as a combined process performed by two or more stations, depending on the desired implementation. Further, if the deep learning controller 218 determines that the intermediate output is not within specifications, it can calculate an adjustment to the control input such that the predicted expected value for the intermediate output is within specifications.

[0065] Note that if the confidence level determined by the deep learning controller 218 falls below a predetermined threshold, the control of the manufacturing process can return to the conventional control described in connection with FIG. 3.

[0066] In some embodiments, the deep learning controller 218 may also monitor whether any of the station / control / process or intermediate output values are unacceptable and further adjust the station controller input or generate an alert if the problem cannot be fixed by adjusting the station controller input.

[0067] Based on the data received while the station operates through the manufacturing process, the deep learning controller 218 can adjust the management input to one or more station controllers. In a further embodiment, the deep learning controller 218 can not only initialize the station controller input before the start of iterations through the process stations of the manufacturing process, but also adjust the station controller input during the period of the process itself ("feedforward management"). In particular, based on the information received from the previous station within the manufacturing process, the deep learning controller 218 can change the management input related to the subsequent station within the process. For example, if the deep learning controller 218 determines that there is a defect in the intermediate output of a particular station, the deep learning controller 218 can determine whether there is any corrective action that can be taken at a subsequent station such that the final output is within the specifications. The deep learning controller 218 can also make changes to the current and previous processing stations in real time based on feedback regarding management / station / process values and / or intermediate / final output values ("feedback management"). This ability to dynamically manage each station in real time and make adjustments to the downstream station controllers to compensate for errors, miscalculations, undesirable conditions, or unexpected results that occur upstream increases the likelihood of producing a final output within the specifications. Further, while a wide range of final outputs may be considered within the specifications, it may be desirable to produce final outputs that are of the same or similar quality and have similar final output values within a narrower range of acceptable final output values. Feedback and feedforward management, along with the predictive capabilities of the deep learning controller, enable the deep learning controller 218 to adjust the station controllers to produce consistent quality final output values and similar final output values.

[0068] It is useful to identify which parameters of the manufacturing process have the most impact on the final output value or process performance (the "key influencers"). The deep learning controller 218 can consider all parameters of the manufacturing process (e.g., one or more control values, one or more station values, one or more process values, one or more stations, one or more intermediate outputs, or any combination thereof) and can use one or more of its machine learning algorithms to identify the key influencers. In some aspects, the deep learning controller 218 can discover one or more key influencers using unsupervised machine learning techniques, e.g., each key influencer is associated with one or more parameters (or combinations of parameters) that affect various station outputs, final outputs, and / or characteristics of process performance. As will be appreciated, the discovery of key influencers and the parameters associated therewith can be performed through the operation and training of the deep learning controller 218 without the need to explicitly label, identify, or otherwise output the key influencers or parameters to a human operator.

[0069] In some approaches, the deep learning controller 218 can rank the impact of each parameter of the manufacturing process on the final output value or process performance in order of importance. Key influencers can be identified based on a cutoff ranking (e.g., the top 5 aspects of the manufacturing process that affect the final output value), a minimum level of influence (e.g., all aspects of the manufacturing process that contribute at least 25% to the final output value); or any other suitable criterion. In some aspects, the key influence characteristics may be associated with a quantitative score that correlates, e.g., with the weight of the influence on the corresponding characteristic.

[0070] The deep learning controller 218 can continuously calculate the key influencers throughout the manufacturing process (step 446).

[0071] In some embodiments, as described in connection with FIG. 5, a key influencer can be used to help construct a more robust dataset for training the deep learning controller 218.

[0072] In conventional manufacturing, the goal is to produce intermediate output values within a specified standard range of variation from the average, so the data generated from the manufacturing process is limited. As a result, the range of control inputs, control / station / process values is also limited because they are all designed to produce intermediate output values within the specified range from the average. In contrast, under the present disclosure, as long as the final output is within the specifications, the intermediate output values do not need to be within a specific range from the average. In some embodiments, as the predictions of the deep learning controller become more accurate, the deep learning controller 218 can intentionally make changes to the control inputs and create conditions for generating intermediate output values that can exceed the normal fluctuations of the in-control processes under conventional manufacturing process control (e.g., SPC), but still produce a final output within the specifications. This allows the deep learning controller 218 to detect patterns and create a more robust data training set for determining how specific stations, station / control / process values, and intermediate output values affect the final output value (e.g., whether the final output is within the specifications). The creation of the robust dataset can be done both during and prior to the implementation of the deep learning controller 218 in a production environment.

[0073] FIG. 5 shows an exemplary process for creating a more robust dataset. In some embodiments, the deep learning controller 218 can adjust known controller inputs (e.g., administrative set points) to one or more station controllers to generate intermediate output values that can exceed a specified range from the average. For example, after conditioning (step 335), the deep learning controller 218 knows at least some of the administrative inputs for each station controller that result in a final output value within the specification. In some embodiments, the deep learning controller 218 can select one or more station controllers and change the known administrative inputs (e.g., set points) to the selected station controllers by a predetermined threshold (e.g., new set point = original set point + 1% of the original set point) (step 510). The deep controller 218 can use the newly calculated administrative inputs, along with the confidence level for its prediction, to predict the expected value (EV) of the final output for the manufacturing process (step 515). If the deep learning controller 218 determines at a threshold confidence level that the expected value is within the specification (step 517), the deep learning controller 218 can provide the adjusted administrative inputs to the selected station controllers (step 520). The deep learning controller 218 can also compare the predicted expected value with the actual final output value (AV) to provide feedback on its prediction and further adjust the administrative inputs (step 525).

[0074] In one exemplary embodiment, the control inputs related to material tolerances can be intentionally changed, and the deep learning controller 218 can be used to determine what adjustments should be made to other control inputs to produce a final product within specifications. By training the deep learning controller 218 in this way, when a new material is unexpectedly introduced into the manufacturing process, the deep learning controller 218 can adapt the control inputs independently without requiring operator input. Similarly, adjustments (e.g., adjustments that simulate a virus or other attack) can be intentionally made to the control algorithm, and the deep learning controller 218 can be used to determine what adjustments should be made to other control inputs to produce a final product within specifications. By intentionally introducing these changes, when adjustments are unexpectedly made to the control algorithm during the manufacturing process, the deep learning controller 218 can adapt the control inputs independently without requiring operator input.

[0075] In some embodiments, the deep learning controller 218 can first determine the key influencers (step 446 described in connection with FIG. 4) and can change the management inputs associated with the key influencers. In other embodiments, the deep learning controller 218 can change the management inputs for all of the station controllers or can select the station controllers according to a predetermined formula. To create a robust dataset, the deep learning controller 218 can continue to adjust the management inputs associated with a particular station controller each time it iterates through the manufacturing process (step 528). For example, the deep learning controller 218 can adjust the set point associated with a key influencer by 1% and can iterate through the manufacturing process one or more times. In subsequent iterations, the deep learning controller 218 can further adjust the set point associated with the key influencer by 1% and can iterate through the manufacturing process one or more times. The deep learning controller 218 can continue to make adjustments as long as the expected value for the manufacturing process using the adjusted set point results in a final output value that is within specifications at a threshold confidence level. During each iteration, as described in connection with FIG. 4, station / management values, intermediate / final output values, process values, and manufacturing performance metrics are generated (step 430) and used to condition the machine learning algorithm of the deep learning controller 218 (step 335) and to dynamically update functional and empirical prior information.

[0076] The following example further illustrates creating a robust dataset by changing a control input (e.g., a temperature set point for a particular station). In this example, the set point temperature for a particular station is 95°, and the actual temperature of the station fluctuates between 92° and 98° (i.e., ±3° above and below the set point temperature). All of the corresponding ±3° fluctuations of the set point temperature at 95° and the actual station temperature result in final output values that are within specifications. And the deep learning controller 218 can predict whether final output values that are still within specifications will result even when the temperature set point is adjusted by a negligible amount (e.g., ±0.5°). If the deep learning controller 218 predicts a final output value within specifications at a threshold confidence level for the manufacturing process using the adjusted temperature set point, the deep learning controller 218 adjusts the temperature set point by ±0.5°. Assuming the same ±3° station temperature fluctuations from the set point, the actual station temperature ranges from 92.5° to 98.5° when the set point is 95.5°, and from 91.5° to 97.5° when the set point is 94.5°. The deep learning controller 218 can compare the final output value to the expected output values over the temperature range from 91.5° to 98.5° and determine whether it correctly predicted that the final output value is within specifications. Since the temperature set point was adjusted by ±0.5°, the resulting dataset covers a wider temperature range of 91.5° to 98.5° than the original temperature range of 92° to 98°.

[0077] Instead of changing the station controller set point, the station's administrative value can be changed by modifying other administrative inputs (e.g., administrative values) to the station controller that achieve the same goal of changing the set point. For example, assume the following. That is, assume that the set point for the station controller is 100 degrees, the actual station temperature value (i.e., the administrative value) is 100 degrees, and the goal is to raise the actual temperature value of the station by 2 degrees. Instead of raising the temperature set point to 102 degrees to achieve that change, the deep learning controller 218 changes the administrative value provided to the station controller by 2 degrees below the actual temperature value (e.g., changes the administrative value from 100 degrees to 98 degrees), and by causing the station controller to raise the station temperature by 2 degrees, can achieve the desired 2-degree increase in station temperature (i.e., 102 degrees). When existing station management does not allow changing the set point, it may be necessary to change the administrative value instead of the set point.

[0078] In some embodiments, as described in connection with FIG. 6, key influencers can be used to optimize the final output or process value.

[0079] After the most important stations, station / management / process values, final output values, or intermediate output values that affect process performance are identified, resource allocation and process optimization can target key influencers. For example, instead of collecting large amounts of data that have a minor impact on the process, data resources (e.g., collection, processing, and storage) can be mostly allocated to data related to key influencers ("curated data"). Further, the curated data (a subset of all data available from the manufacturing process) can be provided to machine learning algorithms for optimization with respect to key influencers, reducing the amount of training examples and increasing the resources available for processing the curated data. Further, the machine learning algorithms are directed at optimizing key influencers instead of the entire process, reducing the possible states and actions that the machine learning algorithms have to consider and allowing for more efficient and sophisticated resource allocation. For example, in reinforcement learning, the state and action space defines the range of possible states that an agent can recognize and the actions available to the agent. By using reinforcement learning only with respect to key influencers, the state and action space is reduced and the algorithm becomes more tractable.

[0080] In some embodiments, an operator can specify one or more characteristics of a final output or process value that the operator wishes to optimize (e.g., using a minimum amount of power or resources, highest throughput, minimum number of defects, maximum tensile strength). The deep learning controller 218 can execute a machine learning model, such as reinforcement learning, that targets key influencers and optimize for the specified characteristics of the final output ("optimal design values") and / or the specified process values ("optimal process values").

[0081] FIG. 6 shows a process for specifying optimal design / process values and using a deep learning controller to manage and optimize key influencers to achieve a desired optimal design or process value, according to some embodiments of the present disclosure.

[0082] At 600, the desired optimal design / process values are provided to the deep learning controller 218. For example, using a minimum amount of power or resources, the highest speed throughput, the minimum number of defects, the maximum tensile strength, etc., to generate the final output within the specifications.

[0083] As shown at 605, the deep learning controller 218 can determine the key influencers that drive the product within the specifications (see also FIG. 4) and predict the management inputs (“optimal management inputs”) for managing each key influencer to achieve the desired optimal design or process values. The deep learning controller 218 uses the optimal management inputs to determine the expected value of the final output for the manufacturing process and predicts whether the expected value is within the specifications and achieves the desired optimal design or process values (step 615). The deep learning controller 218 can also calculate the level of confidence that the optimal management inputs will result in the optimal design or process values (step 615). The deep learning controller 218 can provide the optimal management inputs to the relevant station controllers at the beginning of the manufacturing process or continuously throughout (step 620). The optimal management inputs need not be provided to the station controllers in serial order and can be provided to one or more station controllers in parallel or in any order suitable for producing the final output that achieves the desired optimal design or process values.

[0084] In other embodiments, the parameters for the final output within the specification are updated to match the desired optimal design / process values. For example, if the tensile strength within the specification for the final output is 40 to 90 megapascals (MPa) and the optimal tensile strength for the final output is determined to be 70 to 90 MPa, the parameters within the specification can be updated to 70 to 90 MPa. The deep learning controller 218 can predict and determine the reliability level at which the calculated control input to the station controller associated with the key influencer achieves the updated specification parameters (i.e., 70 to 90 MPa). When the deep learning controller 218 predicts that the calculated control input achieves the updated specification parameters at a reliability level above a predetermined threshold, it can update the associated station controller.

[0085] Note that the deep learning controller 218 can continue to provide control inputs to other station controllers as described in connection with FIG. 3, but it should be noted that the optimization is only targeted at the station controller associated with the key influencer. In some embodiments, the deep learning controller 218 can optimize not only the key influencer but also the station controller associated with any station, station / control / process value, or intermediate output value.

[0086] The deep learning controller 218 can also compare the expected value with the actual value, provide feedback on its prediction, and further adjust the control input (step 545). In some embodiments, when the process starts and passes through all stations, only the measured values related to the key influencer and the measured values for the final output are collected and provided to the deep learning controller 218. In other embodiments, the deep learning controller 218 can continue to collect data from all process stations. By continuously improving the control input to the key influencer using that data, the optimal design / process values are steadily achieved at a high reliability level.

[0087] Note that it should be noted that the desired optimal design / process values can be changed at any time. Further, by changing the criteria for what is considered a qualified key influencer, new key influencers can be calculated (for example, the first criterion may classify the top 5 stations driving the final output within the specification as key influencers, while the updated criterion may classify only the top 3 stations driving the final output within the specification as key influencers). Further, after the deep learning controller 218 predicts the management inputs for achieving the optimal design / process values at a certain reliability level over a period of time, the deep learning controller 218 identifies which of the key influencers are the key influencers for driving the desired optimization, and by targeting only a subset of the key influencers, further reduces the possible actions / states that the machine learning algorithm has to consider, and can allocate resources more efficiently to that subset.

[0088] In a further embodiment, the deep learning controller 218 not only initializes the station controller inputs for the key influencers before starting the iteration through the process stations of the manufacturing process, but also adjusts the management inputs during the course of the process itself. In particular, based on the information received from the previous station within the process, the deep learning controller 218 can make changes to the management inputs related to the subsequent stations within the process to ensure that the optimal design / process values are achieved. The deep learning controller 218 can also adjust the previous stations within the process while progressing through the process and receiving data from subsequent stations.

[0089] In some embodiments, manufacturing performance metrics for a manufacturing process attempting to achieve an optimal design or process value can be calculated and provided to the deep learning controller 218 (step 632). These manufacturing performance metrics can be compared to manufacturing performance metrics for the PPC and / or manufacturing performance metrics for conventional control to determine the improvements provided by optimizing the design and / or process.

[0090] An example of manufacturing system optimization that reduces possible actions / states that can be adapted and implemented by the deep learning controller 218 is described in U.S. Provisional Patent Application No. 62 / 836,199, entitled "Adaptive Methods and Real-Time Decision Making for Manufacturing Control", which is hereby incorporated by reference in its entirety. The disclosed method is merely an example and is not intended to be limiting.

[0091] Another example of manufacturing system optimization that reduces possible actions / states that must be considered by a machine learning algorithm that can be adapted and implemented by the deep learning controller 218 of the present disclosure is described in U.S. Provisional Patent Application No. 62 / 836,213, entitled "TRANSFER LEARNING APPROACH TO MULTI-COMPONENT MANUFACTURING CONTROL", which is hereby incorporated by reference in its entirety. The disclosed method is merely an example and is not intended to be limiting.

[0092] FIG. 7 shows an exemplary manufacturing system to which predictive process management can be applied to optimize the design / process of the final output. Specifically, a 3D manufacturing system can have several process stations (e.g., stations 700-750), and each process station can perform a process step (e.g., depositing an extruded layer of material that results in the final output). Using predictive process management as described in FIG. 6, an operator can specify what to optimize (e.g., the tensile strength of the final output) (step 600). Further, the deep learning controller 218 can determine key influencers that affect the tensile strength of the final output (e.g., changes to the amount of material deposited per layer and changes to the extrusion machine speed control) (step 605). The deep learning controller 218 can reduce the possible actions / states that the machine learning algorithm has to consider by executing a machine learning algorithm (e.g., reinforcement learning), thereby optimizing only the key influencers and not other control attributes (e.g., optimizing the extrusion machine nozzle temperature, the printing pattern of the print head, etc.). Further, the deep learning controller 218 can identify which extruded layers (e.g., layers 4 and 5) have the most impact on the tensile strength and optimize the key influencers for the stations (e.g., stations 4 and 5) that deposit those layers.

[0093] Note that in some embodiments, in relation to the predictive process management described in FIGS. 4-7, the conventional station controller can be turned off and the deep learning controller 218 can directly manage the process stations.

[0094] Furthermore, in some embodiments of the prediction process management, as shown in FIG. 8, the data logging module 810 may be configured to receive data from the deep learning controller 218, analyze the data, and generate a report, an email, an alert, a log file, or other data output (step 815). For example, the data logging module 810 may be programmed to search for predetermined trigger events in the received data and generate a report, an email, an alert, a log file, or other data output indicating the relevant data associated with those trigger events (step 815). For example, it may be possible to define adjusting a set point or other administrative input as a trigger event, and the following data may be reported: namely, the name of the adjusted set point or other administrative input, the affected station, the date and time, the reliability level associated with the adjusted administrative input, whether the adjusted administrative input achieved the final output within the specification or the optimal design / process value, the reason for the adjustment. In another example, it may be possible to define whether the optimal design / process value was achieved as a trigger event, and the following data may be reported: namely, the administrative input for the key influencers, the date and time, any reported stops, the reliability level for the administrative input, the throughput time, the consumed resources, the associated operator. Other suitable triggers may be defined, and other suitable data may be reported. The data logging module may also compare process values, final output values, and manufacturing performance metrics collected during the manufacturing process using conventional management with process values, final output values, and manufacturing performance metrics collected during the manufacturing process using predictive performance management. In some embodiments, the data logging module 810 may be included within the deep learning controller 218.

[0095] FIG. 9 shows a general configuration of an embodiment of a deep learning controller 218 that can implement prediction process management in accordance with some embodiments of the disclosed subject matter. The deep learning controller 218 is shown as a localized computing system in which various components are coupled via a bus 905, but it is understood that the various components and functional computing units (modules) can be implemented as separate physical or virtual systems. For example, one or more components and / or modules can be implemented within physically separate remote devices, such as by using virtual processes (e.g., virtual machines or containers) instantiated within a cloud environment.

[0096] The deep learning controller 218 can include a processing unit (e.g., a CPU and / or a processor) 910 and a bus 905 that couples the processing unit 910 to various system components including a system memory 915 such as a read only memory (ROM) 920 and a random access memory (RAM) 925. The processing unit 910 can include one or more processors such as a processor from a Motorola-based microprocessor or a MIPS-based microprocessor. In an alternative embodiment, the processing unit 910 can be specially designed hardware for controlling the operation of the deep learning controller 218 and performing prediction process management. When operating under the control of appropriate software or firmware, the processing module 910 can execute various machine learning algorithms and computations of the PPC described herein.

[0097] Memory 915 may include various memory forms with different performance characteristics, such as memory cache 912. Processor 910 can be coupled to storage device 930, which can be configured to store software and instructions necessary to implement one or more functional modules and / or database systems. Each of these modules and / or database systems can be configured such that the processor 910 and software instructions control a special-purpose processor incorporated into an actual processor design.

[0098] To enable interaction of the operator with the deep learning controller 218, the input device 945 can represent any number of input mechanisms such as a microphone for voice, a touch-sensitive screen for gesture or graphical input, a keyboard, a mouse, a motion input, etc. The output device 935 can also be one or more of several output mechanisms known to those skilled in the art (e.g., a printer, a monitor). In some examples, a multimodal system enables the operator to provide multiple types of input to communicate with the deep learning controller 218. The communication interface 940 can generally control and manage all electronic inputs received from and transmitted to the operator input and system output, as well as other components constituting the manufacturing process such as a station controller, a process station, a data logging module, and all related sensors and image capture devices. Since there is no limitation to operating on a specific hardware configuration, the basic functions here can be easily replaced with improved hardware or firmware configurations when they are deployed. The data output from the deep controller 218 can be visually displayed, printed, or generated in a file format, stored in the storage device 930, or transmitted to other components for further processing.

[0099] The communication interface 940 can be provided as an interface card (sometimes referred to as a "line card"). Generally, an interface card controls the transmission and reception of data packets over a network and supports other peripheral devices that are sometimes used with a router. Among the interfaces that can be provided are Ethernet interfaces, frame relay interfaces, cable interfaces, DSL interfaces, token ring interfaces, etc. Further, various ultra-high speed interfaces such as high speed token ring interfaces, wireless interfaces, Ethernet interfaces, gigabit Ethernet interfaces, ATM interfaces, HSSI interfaces, POS interfaces, FDDI interfaces, etc. can be provided. Generally, these interfaces can include ports suitable for communication with an appropriate medium. In some cases, the interface may also include an independent processor and, in some examples, may include volatile RAM. The independent processor may control communication-intensive tasks such as packet switching, media control, and management. By providing a separate processor for communication-intensive tasks, these interfaces enable the processing unit 910 to efficiently execute the machine learning and other computations necessary for implementing predictive process management. The communication interface 940 can be configured to communicate with other components that make up a manufacturing process such as a station controller, process station, data logging module, and all associated sensors and image capture devices.

[0100] Sensors related to the manufacturing process can include sensors that existed prior to the implementation of the PPC and any new sensors added to perform any additional measurements used by the PPC. One or more sensors can be included within each station or coupled to each station. The sensors can be used to measure values generated by the manufacturing process, such as station values, control values, intermediate and final output values. Further, the information provided by the sensors can be used by the deep learning controller 218 or by a module external to the deep learning controller 218 to calculate process values and performance manufacturing metrics. Exemplary sensors can include, but are not limited to, rotary encoders that detect position and speed; sensors that detect proximity, pressure, temperature, level, flow, current, and voltage; limit switches that detect states such as presence or movement end limits. A sensor, as used herein, includes both a sensing device and signal conditioning. For example, the sensing device responds to a station or control value, and the signal conditioner converts that response into a signal that can be used and interpreted by the deep learning controller. Examples of sensors that respond to temperature are RTDs, thermocouples, and platinum resistance probes. Strain gauge sensors respond particularly to pressure, vacuum, weight, and changes in distance. Proximity sensors respond when an object is within a certain distance from each other or within a specified tart range. In all of these examples, the response must be converted into a signal that can be used by the deep learning controller 218. In many cases, the signal conditioning function of the sensor generates a digital signal that is interpreted by the deep learning controller 218. The signal conditioner can also generate, in particular, analog signals or TTL signals.

[0101] In some embodiments, the deep learning controller 218 may include an image processing device 970 coupled to one or more processing stations and configured to process images received by various image capture devices, such as a video camera, capable of monitoring and capturing intermediate and final output images. These images are transmitted to the deep learning controller 218 via the communication interface 940 and can be processed by the image processing device 970. The images can be processed to provide data such as the number and type of defects, output dimensions, throughput, etc., that can be used by the deep learning controller 218 to calculate intermediate and final output values. In some embodiments, the image processing device is external to the deep learning controller 218 and can provide information to the deep learning controller 218 via the communication interface 940.

[0102] The memory device 930 is a non-transitory memory and can be a hard disk or other type of computer-readable medium capable of storing data accessible by a computer, such as a magnetic cassette, flash memory card, solid state memory device, digital versatile disk, cartridge, random access memory (RAM) 825, read-only memory (ROM) 820, and hybrids thereof.

[0103] In practice, the memory device 930 can be configured to receive, store, and update input data to the deep learning controller 218 and output data from the deep learning controller 218, such as functional prior information, empirical prior information, universal input; pre-process input; in-process input and post-process input.

[0104] In some embodiments, any suitable computer-readable medium may be used to store instructions for performing the functions and / or processes described herein. For example, in some embodiments, the computer-readable medium may be transient or non-transient. For example, non-transient computer-readable media may include non-transient magnetic media (such as hard disks, floppy disks, etc.), non-transient optical media (such as compact disks, digital video disks, Blu-ray disks, etc.), non-transient semiconductor media (such as flash memory, electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.), any suitable medium that is neither transient nor lacking in permanence during transmission, and / or any suitable tangible media such as any suitable physical object medium. As another example, transient computer-readable media may include signals on a network, on a conductor, in a semiconductor, in an optical fiber, in a circuit, and any suitable medium that is transient and lacking in any aspect of permanence during transmission, and / or any suitable intangible media.

[0105] The various systems, methods, and computer-readable media described herein can be implemented as part of a cloud network environment. As used herein, a cloud-based computing system is a system that provides virtualized computing resources, software, and / or information to client devices. The computing resources, software, and / or information can be virtualized by maintaining centralized services and resources that are accessible by edge devices through a communication interface such as a network. The cloud can provide software as a service (SaaS) (e.g., collaboration services, email services, enterprise resource planning services, content services, communication services, etc.), infrastructure as a service (IaaS) (e.g., security services, networking services, system management services, etc.), platform as a service (PaaS) (e.g., web services, streaming services, application development services, etc.), and other types of services such as desktop as a service (DaaS), information technology management as a service (ITaaS), managed software as a service (MSaaS), mobile backend as a service (MBaaS), etc.

[0106] The provision of examples described herein (and clauses expressed as "such as", "for example", "including", etc.) should not be construed as limiting the subject matter recited in the claims to those specific examples; rather, those examples are intended to illustrate only some of the many possible aspects. As will be understood by those skilled in the art, the term mechanism can encompass hardware, software, firmware, or any suitable combination thereof.

[0107] As is apparent from the above description, unless otherwise specified, throughout the specification, descriptions using terms such as "determine", "provide", "identify", "compare", etc. refer to actions and processes of a computer system or similar electronic computing device that manipulate and transform data represented as physical (electronic) quantities within a computer system memory or register or other such information storage, transmission, or display device. Some aspects of the present disclosure include process steps and instructions described herein in the form of algorithms. It should be noted that the process steps and instructions of the present disclosure can be embodied in software, firmware, or hardware, and when embodied in software, it can be downloaded and operated from different platforms located on different platforms used by a real-time network operating system.

[0108] The present disclosure also relates to an apparatus for performing the operations herein. The apparatus may be specially constructed for the required purposes, or may comprise a general-purpose computer selectively activated or reconfigured by a computer program stored on a computer-readable medium accessible by the computer. Such a computer program may be stored on a computer-readable storage medium, including but not limited to any type of disk, such as a floppy disk, optical disk, CD-ROM, magneto-optical disk, read-only memory (ROM), random access memory (RAM), EPROM, EEPROM, magnetic or optical card, application specific integrated circuit (ASIC), or any type of non-transitory computer-readable storage medium suitable for storing electronic instructions. Further, the computers referred to herein may include a single processor or may be an architecture that uses a multi-processor design for improved computing capabilities.

[0109] The algorithms and operations presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may also be used with programs in accordance with the teachings herein, or it may be convenient to construct more specialized apparatus to perform the required method steps and system-related actions. The required structure for various of these systems will be apparent to those skilled in the art along with equivalent variations. Further, the present disclosure has not been described with reference to any particular programming language. As will be understood, a variety of programming languages may be used to implement the teachings of the present disclosure described herein, and any reference to a specific language is provided for disclosure of the enablement and best mode of the present disclosure.

[0110] The logical operations of the various embodiments are implemented as (1) computer-implemented steps, operations, or sequences of procedures that operate on programmable circuits within a general purpose computer, (2) computer-implemented steps, operations, or sequences of procedures that operate on application-specific programmable circuits, and / or (3) interconnected machine modules or program engines within a programmable circuit. System 300 can perform all or part of the described methods, can be part of the described systems, and / or can operate in accordance with instructions within the described non-transitory computer-readable storage medium. Such logical operations can be implemented as modules configured to control processor 363 to perform specific functions according to the programming of the modules.

[0111] Any particular order or hierarchy of steps within the disclosed processes is to be understood as an example of an exemplary approach. Based on design preferences, the particular order or hierarchy of steps within a process may be rearranged, or only some of the steps illustrated may be performed. Some of the steps may be performed simultaneously. For example, in some situations, multitasking and parallel processing may be advantageous. Further, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and the described program components and systems can generally be integrated together within a single software product or packaged into multiple software products.

[0112] The apparatus, method, and system for prediction process management have been specifically described with reference to these exemplary embodiments. However, as will be apparent, various modifications and changes may be made within the spirit and scope of the disclosure described in the above specification, and such modifications and changes should be considered to be within the scope of equivalents and part of this disclosure.

[0113] Statement of Disclosure Statement 1: A computer-implemented method comprising receiving, by a deep learning controller, a plurality of management values from two or more stations, wherein the management values are generated at the two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an expected value for an intermediate or final output of a manufactured article based on the management values; determining, by the deep learning controller, whether the predicted expected value for the manufactured article is within specifications; and generating, by the deep learning controller, one or more management inputs when the predicted expected value for the manufactured article is not within specifications, wherein the one or more management inputs are configured such that an intermediate or final output for the manufactured article is within specifications.

[0114] Statement 2: The computer-implemented method of Statement 1, further comprising outputting the one or more management inputs to one or more station controllers when a reliability level associated with the one or more management inputs exceeds a predetermined threshold.

[0115] Statement 3: Outputting the one or more management inputs to the one or more station controllers further comprises providing a first management input to a first station controller associated with a first station and providing a second management input to a second station controller associated with a second station, wherein the second station is downstream of the first station within the manufacturing process, the computer-implemented method according to any one of Statements 1 to 2.

[0116] Statement 4: Predicting the intermediate or final output of the manufactured article further comprises identifying one or more key influencers associated with the manufacturing process and adjusting one or more management inputs associated with at least one of the one or more key influencers, the computer-implemented method according to any one of Statements 1 to 3.

[0117] Statement 5: The computer-implemented method according to any one of Statements 1 to 4, further comprising permitting management of the manufacturing process by the one or more station controllers when the predicted expected value for the manufactured article is within specifications.

[0118] Statement 6: The computer-implemented method according to any one of Statements 1 to 5, further comprising receiving, by the deep learning controller, a plurality of station values, the station values being associated with the one or more stations, and predicting the expected value of the manufactured article being further based on the station values.

[0119] Statement 7: The computer-implemented method according to any one of Statements 1 to 6, wherein the plurality of control values includes one or more of speed, temperature, pressure, vacuum, rotation, current, voltage, resistance, or power.

[0120] Statement 8: A system comprising one or more processors and a non-transitory memory storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations including: receiving, by a deep learning controller, a plurality of control values from two or more stations, the control values being generated at the two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an expected value for an intermediate or final output of a manufactured article based on the control values; determining, by the deep learning controller, whether the predicted expected value for the manufactured article is within specifications; and generating, by the deep learning controller, one or more control inputs when the predicted expected value for the manufactured article is not within specifications, the one or more control inputs being configured such that an intermediate or final output for the manufactured article is within specifications.

[0121] Statement 9: The system according to statement 8, wherein the processor is further configured to perform an operation including outputting the one or more management inputs to the one or more station controllers when a reliability level associated with the one or more management inputs exceeds a predetermined threshold.

[0122] Statement 10: The system according to any one of statements 8 to 9, wherein outputting the one or more management inputs to the one or more station controllers further includes providing a first management input to a first station controller associated with a first station and providing a second management input to a second station controller associated with a second station, and the second station is downstream of the first station in the manufacturing process.

[0123] Statement 11: The system according to any one of statements 8 to 10, wherein predicting the expected value of the manufactured article further includes identifying one or more key influencers associated with the manufacturing process and adjusting one or more management inputs associated with at least one of the one or more key influencers.

[0124] Statement 12: The system according to any one of statements 8 to 11, further including permitting management of the manufacturing process by the one or more station controllers when the predicted expected value for the manufactured article is within specifications.

[0125] Statement 13: The system according to any one of statements 8 to 12, wherein the processor is further configured to perform an operation including receiving a plurality of station values at the deep learning controller, the station values being associated with the one or more stations, and predicting the expected value of the manufactured article is further based on the station values.

[0126] Statement 14: The system according to any one of Statements 8 to 13, wherein the plurality of management values includes one or more of speed, temperature, pressure, degree of vacuum, rotation, current, voltage, resistance, or power.

[0127] Statement 15: A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations, the operations including: receiving, by a deep learning controller, a plurality of management values from two or more stations, the management values being generated at the two or more stations deployed within a manufacturing process; predicting, by the deep learning controller, an expected value for an intermediate or final output of a manufactured article based on the management values; determining, by the deep learning controller, whether the predicted expected value for the manufactured article is within specifications; and generating, by the deep learning controller, one or more management inputs when the predicted expected value for the manufactured article is not within specifications, the one or more management inputs being configured such that the intermediate or final output for the manufactured article is within specifications.

[0128] Statement 16: The non-transitory computer-readable storage medium according to Statement 15, further including instructions configured to cause the one or more processors to perform an operation including outputting the one or more management inputs to one or more station controllers when a reliability level associated with the one or more management inputs exceeds a predetermined threshold.

[0129] Statement 17: Outputting the one or more management inputs to the one or more station controllers further includes providing a first management input to a first station controller associated with a first station and providing a second management input to a second station controller associated with a second station, wherein the second station is downstream of the first station in the manufacturing process, the non-transitory computer-readable storage medium according to any one of Statements 15 to 16.

[0130] Statement 18: Predicting the expected value of the manufactured article further includes identifying one or more key influencers associated with the manufacturing process and adjusting one or more management inputs associated with at least one of the one or more key influencers, the non-transitory computer-readable storage medium according to any one of Statements 15 to 17.

[0131] Statement 19: Further including instructions configured to cause the one or more processors to perform an operation including permitting management of the manufacturing process by the one or more station controllers when the predicted expected value for the manufactured article is within specifications, the non-transitory computer-readable storage medium according to any one of Statements 15 to 18.

[0132] Statement 20: Further including instructions configured to cause the one or more processors to perform an operation including receiving a plurality of station values by the deep learning controller, wherein the station values are associated with the one or more stations, and predicting the expected value of the manufactured article is further based on the station values, the non-transitory computer-readable storage medium according to any one of Statements 15 to 19.

Claims

1. A computer-implemented method, comprising: receiving, by a deep learning controller, a first set of in-process inputs that are generated at a first processing station among a plurality of stations in a manufacturing process and are attributes of the first processing station, from the first processing station; predicting, by the deep learning controller, an expected value for an intermediate value of a manufactured article or a final output of the manufactured article at a downstream processing station based on the first set of in-process inputs; determining, by the deep learning controller, that the expected value for the manufactured article is out of specification; generating, by the deep learning controller based on the determination, one or more management inputs configured to optimize the intermediate value or the final output of the manufactured article to be within specification; after generating the one or more management inputs, updating, by the deep learning controller, processing parameters of a downstream station controller associated with the downstream processing station based on the one or more management inputs; moving, according to the one or more management inputs, the manufactured article from the first processing station to the downstream processing station for a further step in the manufacturing process; A computer-implemented method as described above.

2. The computer-implemented method according to claim 1, further comprising updating the one or more management inputs when a reliability level associated with the one or more management inputs of the downstream station controller exceeds a predetermined threshold.

3. Updating the one or more management inputs of the downstream station controller comprises: providing a first management input to a first station controller associated with a first downstream processing station; and further providing a second management input to a second station controller associated with a second downstream processing station, wherein the second downstream processing station is downstream of the first downstream processing station in the manufacturing process. The computer-implemented method according to claim 2.

4. Predicting the expected value for the intermediate value of the manufactured article or the final output of the manufactured article at the downstream processing station comprises: Identifying one or more key influencers associated with the manufacturing process; Adjusting the one or more management inputs associated with at least one of the one or more key influencers; The computer-implemented method according to claim 1, further comprising.

5. Receiving, by the deep learning controller, a second set of in-process inputs generated at the downstream processing station; Predicting, by the deep learning controller, a second expected value for a second intermediate value of the manufactured article or an updated final output of the manufactured article at a second downstream processing station based on the first set of in-process inputs and the second set of in-process inputs; The computer-implemented method according to claim 1, further comprising.

6. When it is determined that the second expected value of the manufactured article is out of specification, generating, by the deep learning controller, a second set of management inputs configured such that the second intermediate value or the updated final output for the manufactured article is within specification; The computer-implemented method according to claim 5, further comprising.

7. The computer-implemented method according to claim 1, wherein the first set of in-process inputs includes one or more of speed, temperature, pressure, vacuum, rotation, current, voltage, resistance, or power.

8. A system comprising: One or more processors; A non-transitory memory storing instructions that, when executed by the one or more processors, cause the processors to perform operations; A system comprising, wherein the operations are: Receiving, by a deep learning controller, a first set of in-process inputs generated at a first processing station among a plurality of stations in a manufacturing process and being an attribute of the first processing station, from the first processing station; Predicting, by the deep learning controller, an expected value for an intermediate value of a manufactured article or a final output of the manufactured article at a downstream processing station based on the first set of in-process inputs; Determining, by the deep learning controller, that the expected value for the manufactured article is out of specification; Based on the determination, one or more management inputs are generated by the deep learning controller to optimize the intermediate value or the final output for the manufactured article to be within specifications. After generating the one or more management inputs, the deep learning controller updates the processing parameters of the downstream station controller associated with the downstream processing station based on the one or more management inputs. The manufactured article is moved from the first processing station to the downstream processing station for subsequent steps within the manufacturing process in accordance with the one or more management inputs. A system comprising.

9. The operation is updating the one or more management inputs when a reliability level associated with the one or more management inputs of the downstream station controller exceeds a predetermined threshold. The system according to claim 8, further comprising.

10. Updating the one or more management inputs of the downstream station controller is providing a first management input to a first station controller associated with a first downstream processing station and further comprising providing a second management input to a second station controller associated with a second downstream processing station, The system according to claim 9, wherein the second downstream processing station is downstream of the first downstream processing station within the manufacturing process.

11. Predicting an expected value for an intermediate value of a manufactured article or a final output of a manufactured article at the downstream processing station is identifying one or more key influencers associated with the manufacturing process and adjusting the one or more management inputs associated with at least one of the one or more key influencers. The system according to claim 8, further comprising.

12. The operation is receiving, by the deep learning controller, a second set of in-process inputs generated at the downstream processing station. The system according to claim 8, further comprising, in the deep learning controller, predicting a second expected value for a second intermediate value of the manufactured article at a second downstream processing station or an updated final output of the manufactured article based on a first set of the in-process inputs and a second set of the in-process inputs.

13. When the operation determines that the second expected value of the manufactured article is out of specification, the deep learning controller generates a second set of management inputs configured such that the second intermediate value or the updated final output for the manufactured article is within specification. The system according to claim 12, further comprising.

14. The system according to claim 8, wherein the first set of the in-process inputs includes one or more of speed, temperature, pressure, degree of vacuum, rotation, current, voltage, resistance, or power.

15. A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the processor to perform operations, the operations including: receiving, by a deep learning controller, a first set of in-process inputs generated at a first processing station among a plurality of stations in a manufacturing process and being an attribute of the first processing station, from the first processing station; predicting, by the deep learning controller, an expected value for an intermediate value of a manufactured article or a final output of the manufactured article at a downstream processing station based on the first set of the in-process inputs; determining, by the deep learning controller, that the expected value for the manufactured article is out of specification; generating, by the deep learning controller based on the determination, one or more management inputs configured to optimize such that the intermediate value or the final output for the manufactured article is within specification; after generating the one or more management inputs, updating, by the deep learning controller, processing parameters of a downstream station controller associated with the downstream processing station based on the one or more management inputs; moving, according to the one or more management inputs, the manufactured article from the first processing station to the downstream processing station for a further step in the manufacturing process. A non-transitory computer-readable storage medium including

16. wherein the operation updates the one or more management inputs when a reliability level associated with the one or more management inputs of the downstream station controller exceeds a predetermined threshold; The non-transitory computer-readable storage medium according to claim 15, further comprising:

17. Updating the one or more management inputs of the downstream station controller providing a first management input to a first station controller associated with a first downstream processing station; further comprising providing a second management input to a second station controller associated with a second downstream processing station, The non-transitory computer-readable storage medium according to claim 16, wherein the second downstream processing station is downstream of the first downstream processing station in the manufacturing process.

18. Predicting an expected value for an intermediate value of a manufactured article or a final output of the manufactured article at the downstream processing station identifying one or more key influencers associated with the manufacturing process; adjusting the one or more management inputs associated with at least one of the one or more key influencers; The non-transitory computer-readable storage medium according to claim 15, further comprising:

19. wherein the operation receiving, by the deep learning controller, a second set of in-process inputs generated at the downstream processing station; The non-transitory computer-readable storage medium according to claim 15, further comprising predicting, by the deep learning controller, a second expected value for the second intermediate value of the manufactured article or an updated final output of the manufactured article at a second downstream processing station based on the first set of in-process inputs and the second set of in-process inputs.

20. wherein the operation generating, by the deep learning controller, a second set of management inputs configured such that the second intermediate value or the updated final output for the manufactured article is within specifications when it is determined that the second expected value of the manufactured article is not within specifications; The non-transitory computer-readable storage medium according to claim 19, further comprising:

Images