Real-time opportunity discovery method and system for productivity enhancement

Abstract

In a method for real-time opportunity discovery for productivity enhancement in a production process, a processor extracts a set of features from time series data using autoencoding of a neural network based on uncontrolled variables in the time series data. The processor identifies one or more operating patterns based on the extracted features, wherein the features include a dimensionality reduction of a representation learned from the time series data. The processor identifies a neighborhood of a current operating state based on the extracted features. The processor compares the current operating state with historical operating states based on the time series data in the same operating pattern. The processor uses the neighborhood to discover operating opportunities based on the comparison of the current operating state with the historical operating states. The processor identifies controlled variables in the same pattern that are correlated with the current operating state.

Description

Background Art

[0001] The present disclosure relates generally to the field of machine learning, and more particularly to real-time opportunity discovery for productivity enhancement of a production process encoded using historical data.

[0002] Many production methods can be quite complex. For example, in oil sands production, the mined ore can go through several stages of extraction, upgrading, and refining. Similar methods can be found in food and steel production processes. The outflow from the upstream process can become the inflow into the downstream process. Each stage can involve multiple components and processes, and the system is a dynamic system. For the oil sands process, in a typical configuration, when enough raw materials are available from mining and all components are running correctly, the operation is carried out at full capacity. The upgrading process operation can be performed without a vacuum process. When the quality of the bitumen from the processed oil sands is low (for example, with a high concentration of chlorides), a low production mode can be generated to avoid the degradation of the coking unit. When the raw material flow line is undergoing maintenance, partial capacity operation can occur. When the quality of the bitumen from the processed oil sands is low (for example, with a high concentration of chlorides), a low production mode can be generated to avoid the degradation of the coking unit. When the raw material flow line is undergoing maintenance, partial capacity operation can occur.

[0003] Traditionally, before being sold to the refinery on the market, the asphalt of most production is upgraded to synthetic crude oil. However, some asphalts are good enough to directly send to the high conversion refinery with the ability of processing heavy / sour crude oil. This diluted asphalt example sold directly to the refinery includes products from field facilities and other places. Petroleum products can be produced from oil sands by three basic steps: i) extracting asphalt from oil sands, wherein removing solids and water, ii) heavy asphalt is upgraded to lighter intermediate crude oil products, and iii) crude oil is refined into final product, such as gasoline, lubricant and thinner. All these methods relate to a plurality of sequential steps of physical or chemical conversion to be converted into another material from a kind of material. The optimal balance of technology is needed to reach a plurality of targets in this production system. Need factory operators to seek opportunities to increase productivity, for example, less raw materials, cheaper additives, higher final products. Further need to concentrate on the specific area with high commercial value to provide incremental value in the local step of manufacturing process. Yet another need is to identify cost, raw material, and energy saving model opportunities and help obtain incremental additional profits that are confined to local steps of the overall factory operation, where these opportunities are within a relatively short time window in the manufacturing process. Summary of the Invention

[0004] Certain shortcomings of the prior art are overcome, and additional advantages are provided by providing a method for real-time opportunity discovery for productivity enhancement of a production process. Advantageously, a processor extracts a set of features from the time series data based on non-controlled variables of the time series data by using an auto-encoder of a neural network. The processor identifies one or more operating modes based on the extracted features (including size reduction), wherein the representation is learned from the time series data. The processor identifies a neighborhood of a current operating state based on the extracted features. The processor compares the current operating state with historical operating states based on the time series data in the same operating mode. The processor uses the neighborhood to discover operating opportunities based on the comparison of the current operating state with the historical operating states. The processor identifies controlled variables in the same mode that are related to the current operating state. The processor recommends an action strategy based on one or more controlled variables, one or more non-controlled variables, and a target productivity.

[0005] In one or more embodiments, a computer-implemented method is provided for monitoring time-series data generated from one or more sensors. For example, the time-series data may be data from oil sands operations and production processes. Petroleum products may be produced from oil sands through several stages, such as extraction, upgrading, and refining. Advantageously, an optimal balance of processes is provided to achieve multiple objectives within such a production system.

[0006] In one or more embodiments, a computer-implemented method is provided for extracting a set of features from time series data based on one or more uncontrolled variables of the time series data by autoencoding using a neural network (e.g., an LSTM autoencoder). Advantageously, the LSTM autoencoder not only uses the set of features to learn, but also learns the set of features through the LSTM autoencoder. A feature set is information related to the time series data. A feature set can be an individual measurable attribute or characteristic of a phenomenon observed from the time series data. An opportunity discovery module can select a subset of relevant features for creating a predictive model based on uncontrolled variables that define a user's lack of control over the time series data. Advantageously, the opportunity discovery module can reduce the number of resources required to describe the time series data. The opportunity discovery module can construct a combination of uncontrolled variables that describe the time series data with sufficient accuracy through the LSTM autoencoder. Advantageously, the LSTM autoencoder can invoke an autoencoding process based on the time series data to reduce the dimensionality of the sensor tag space to a finite embedding space.

[0007] In one or more embodiments, a computer-implemented method is provided for identifying one or more operating modes based on extracted features, the features comprising dimensionality reduction accompanied by learning from representations of the time series data. Advantageously, dimensionality reduction is a transformation of the time series data from a high-dimensional space to a low-dimensional space such that the low-dimensional representation retains some meaningful properties of the original data, ideally close to the intrinsic dimensionality. Advantageously, a neighborhood can be identified for the current operating state. The neighborhood can be a dynamic pattern within the same operating mode and can be found by the Euclidean distance between historical operating states and the current operating state. Advantageously, automatic processing is provided rather than relying on rule-based pattern detection that requires a large amount of prior knowledge and storage principles. The opportunity discovery module can implement opportunity realization through analysis using unsupervised machine learning. For example, the opportunity discovery module can identify operating opportunities by comparing the current state with historical similar operations within a pattern or neighborhood.

[0008] In one or more embodiments, a computer-implemented method is provided for comparing a current operating state with historical operating states based on time series data in the same operating mode in one or more operating modes using an opportunity discovery module. Advantageously, the opportunity discovery module can identify the specific mode in which the current operating state is located. The opportunity discovery module can project clusters using t-distributed stochastic neighbor embedding (t-SNE) compression to generate a graph. The opportunity discovery module can embed high-dimensional points in a low-dimensional manner in a manner that respects similarities between points with t-SNE compression. In an instance, the opportunity discovery module can achieve high asphalt extraction by comparing the current operating state with other operations in the same mode. The opportunity discovery module can analyze events of poor performance and can discover operational opportunities for improvement.

[0009] In one or more embodiments, a computer-implemented method is provided for discovering operational opportunities based on a comparison of a current operating state to a historical operating state. Advantageously, operational opportunities can be identified by comparing the current state to historical similar operations within a pattern or neighborhood. In an example, an operational opportunity can be a set of operational changes derived from the historical operating state to increase the current operating state to a higher output in a defined short-term future period (e.g., a two-hour window). In another example, an operational opportunity can be a set of operational changes derived from the historical operating state to reduce the current operating state to a low usage of an additive or raw material in a defined short-term future period. Other suitable opportunities can be found.

[0010] In one or more embodiments, a computer-implemented method is provided for identifying control variables that are in a common pattern and are associated with a current operating state. Advantageously, the control variables can be used to calculate rewards (or opportunities) from episodes of time series data. For example, the control variables can be production rates and raw material variables that can be optimized by the user based on the best neighboring events found. The control variables can be identified from established neighborhoods of similar historical non-control variables to create a possible action strategy associated with the current state based on the time series data.

[0011] In one or more embodiments, a computer-implemented method is provided for recommending an action strategy based on controlled variables, uncontrolled variables, and target productivity. Advantageously, a similarity measure can be defined to identify historical episodes with similar operating states from time series data based on a comparison of the current state to historical operating states. Episodes can be created from historical episodes that exhibited higher productivity or throughput. Scores can be generated based on alternative action strategies and used to recommend an action strategy based on the scores for each alternative action strategy.

[0012] In one or more embodiments, a computer-implemented method is provided for outputting an action strategy for a user. Advantageously, a neighborhood scenario using a time-stamped graph can be presented in a user interface. An estimated gain of the action strategy can be presented.

[0013] In another aspect, a computer program product is provided, comprising one or more computer-readable storage media and program instructions stored collectively on the one or more computer-readable storage media. Advantageously, the program instructions extract a set of features from the time series data using an autoencoder of a neural network based on non-controlled variables of the time series data. The program instructions identify one or more operating modes based on the extracted features, wherein the features include dimensionality reduction using a representation learned from the time series data. The program instructions identify a neighborhood of a current operating state based on the extracted features. The program instructions compare the current operating state with historical operating states based on the time series data within the same operating mode. The program instructions identify operating opportunities based on a comparison of the current operating state with the historical operating states using the neighborhood. The program instructions identify controlled variables within the same mode that are correlated with the current operating state. The program instructions recommend an action strategy based on the one or more controlled variables, the one or more non-controlled variables, and a target productivity.

[0014] In a further aspect, a computer system is provided, comprising one or more computer processors, one or more computer-readable storage media, and program instructions stored on the one or more computer-readable storage media for execution by at least one of the one or more computer processors. Advantageously, the program instructions extract a set of features from the time series data using an autoencoder using a neural network based on non-controlled variables in the time series data. The program instructions identify one or more operating modes based on the extracted features, wherein the features include dimensionality reduction using a representation learned from the time series data. The program instructions identify a neighborhood of a current operating state based on the extracted features. The program instructions compare the current operating state with historical operating states based on the time series data within the same operating mode. The program instructions identify operating opportunities based on a comparison of the current operating state with the historical operating states using the neighborhood. The program instructions identify controlled variables within the same mode that are correlated with the current operating state. The program instructions recommend an action strategy based on the one or more controlled variables, the one or more non-controlled variables, and a target productivity.

[0015] Additional features and advantages are realized through the techniques of the present invention.Other embodiments and aspects of the invention are described in detail herein and are considered a part of the claimed invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a functional block diagram illustrating an operating opportunity discovery environment according to an embodiment of the present disclosure.

[0017] Figure 2 The present invention is described in accordance with an embodiment of the present invention. Figure 1 A flowchart of the operating steps of an opportunity discovery module within a computing device.

[0018] Figure 3 The embodiment according to the present disclosure is shown Figure 1 An exemplary functional diagram of an opportunity discovery module within a computing device.

[0019] Figure 4 The embodiment according to the present disclosure is shown Figure 1 An exemplary architectural diagram of an opportunity discovery module within a computing device.

[0020] Figure 5 According to the embodiment of the present disclosure Figure 1 A block diagram of the components of a computing device. DETAILED DESCRIPTION

[0021] The present disclosure relates to systems and methods for real-time opportunity discovery using historical data encoding to improve the productivity of a production process.

[0022] Embodiments of the present disclosure recognize the need for plant operators to seek opportunities to improve productivity, such as less raw materials, cheaper additives, and higher final products. Embodiments of the present disclosure can focus on specific areas with high commercial value to provide incremental value in local steps of the manufacturing process. Embodiments of the present disclosure can discover opportunities for cost, raw material, and energy saving models. These opportunities can have a relatively short time window. Embodiments of the present disclosure can help obtain additional profit increases limited to local steps of the entire plant operation. Embodiments of the present disclosure can select a complete set of time series from sensors (e.g., Internet of Things (IoT) sensors) so that the time series has a complete picture of the plant status. Embodiments of the present disclosure can separate the time series into controlled and uncontrolled variables. During the production process, embodiments of the present disclosure can dynamically provide recommendations and suggestions for improving productivity through less raw material consumption, cheaper additive use, less energy consumption, and higher product output. Embodiments of the present disclosure can obtain more timely and accurate opportunities in a window of only a few hours. Embodiments of the present invention can use extracted features as an embedding space to utilize automatic encoding technology to find operating patterns and neighborhoods. The autoencoding technique enables time series dimensionality reduction and generates production recommendations from historical similarity analysis among neighborhoods in the pattern or embedded space.

[0023] Embodiments of the present disclosure may apply a long short-term memory (LSTM) autoencoder to extract features or embedded spaces for non-controlled variables. Embodiments of the present disclosure may define neighborhoods or use Gaussian mixture clustering applied to embedded spaces to identify static operating modes. Embodiments of the present disclosure disclose identifying neighborhoods for the current operating state. Neighborhoods may be dynamic patterns within the same operating mode and may be found by the Euclidean distance between historical operating states and the current operating state. Embodiments of the present disclosure may identify opportunities by looking at differences in control variables to identify suggestions for improvement. Embodiments of the present disclosure may compare the current operating state with other operations in the same mode. Poor performance episodes present opportunities for improvement. Embodiments of the present disclosure may limit historical episodes by selecting neighborhoods for the current operating state. Embodiments of the present disclosure may perform complete validation of time series data for the performance accuracy of a prediction model developed using all control variables and non-controlled variables.

[0024] The present disclosure will now be described in detail with reference to the accompanying drawings. Figure 1 is a functional block diagram illustrating an operational opportunity discovery environment, generally designated 100 , in accordance with an embodiment of the present disclosure.

[0025] In the depicted embodiment, the operational opportunity discovery environment 100 includes a computing device 102, time series data 104, and a network 108. In one embodiment, the time series data 104 can be directly accessed by the computing device 102. In another embodiment, the time series data 104 can be accessed via a communication network (such as network 108). In one or more embodiments, the time series data 104 can be data captured by one or more sensors. For example, the time series data 104 can be data from oil sands operations and production processes. In the example of oil sands operations and production processes, some asphalt can be upgraded to synthetic crude oil before being sold to refineries. Some asphalt may be high enough to be transported to high-conversion refineries that can process heavy crude oil. Petroleum products can be produced from oil sands through several stages, such as extraction, upgrading, and refining. For example, solids and water can be removed during the extraction stage, which can extract asphalt from the oil sands. During the upgrading stage, the asphalt can be upgraded to lighter intermediate crude oil products. During the refining stage, the crude oil can be refined into final products such as gasoline, lubricants, and diluents. The processes during these stages may involve multiple sequential steps of physical or chemical transformations to convert from one material to another.An optimal balance of processes is required to achieve multiple goals within such a production system.

[0026] In another example, time series data 104 may be data from a steelmaking process that produces steel from iron ore and / or scrap. Impurities such as nitrogen, silicon, phosphorus, sulfur, and excess carbon may be removed from the source iron, and alloying elements such as manganese, nickel, chromium, carbon, and vanadium may be added to produce different grades of steel. In another example, time series data 104 may be data from a production process that processes soybeans into soy sauce, where additives are added to the soy sauce. In yet another example, time series data 104 may be data from any other suitable operation or production process.

[0027] In one or more embodiments, the time series data 104 may include, for example, non-control variables 122 and control variables 124. The non-control variables 122 and control variables 124 may be separated to ensure the ability to take action to capture production enhancement opportunities. For example, the non-control variables 122 may be time-stamped variables defined as a set of variables from sensors over which the user has little control. The non-control variables 122 may be parameters used to define how similar operating and production conditions are. The non-control variables 122 may be retrieved from the time series data 104 to identify episodes. The control variables 124 may be time-stamped variables defined as a set of variables from sensors for user-controllable actions. In one example, the control variables 124 may be used to calculate rewards (or opportunities) from episodes in the time series data 104. The control variables 124 may be production rate and raw material variables that the user can optimize based on the best neighboring events found. The control variables 124 may be identified from an established neighborhood of similar historical non-control variables 122 to create a possible action strategy relevant to the current state based on the time series data 104.

[0028] In the foam production example, non-controlled variables 122 can be environmental variables that the user has little or no control over, such as ambient temperature, flash, turbidity, viscosity, hydrogen utilization, coking rate, and raw production. Control variables 124 can be variables of actions that can be controlled by the user, such as productivity and raw material variables that the user can optimize or change. Control variables 124 can include, for example, diesel hydrotreating or catalytic hydrogen treatment, with the feed plate loaded with diesel hydrotreating having low vacuum gas oil, low vacuum gas oil, side-cut kerosene, heavy naphtha, and coking kerosene. Diesel hydrotreating or catalytic hydrogen treatment can primarily reduce undesirable substances from the straight-run diesel fraction by selectively reacting these substances with hydrogen in a reactor at elevated temperatures and moderate pressures. In order to successfully produce ultra-low sulfur diesel, it is necessary to remove organic sulfur species, including substituted dibenzothiophenes and other refractory sulfur species. Multiple reactions can occur in parallel on the diesel hydrotreating catalyst surface, including hydrodesulfurization, hydrodenitrogenation, and aromatic saturation / hydrogenation. The feedstock for a diesel hydrotreating unit may have a nominal distillation range of 300-700° F. Different process designs and flows may be used for diesel hydrotreating, depending on the process goals and the characteristics of the feedstock being processed.

[0029] In various embodiments of the present disclosure, the computing device 102 may be a laptop computer, a tablet computer, a netbook computer, a personal computer (PC), a desktop computer, a mobile phone, a smart phone, a smart watch, a wearable computing device, a personal digital assistant (PDA), or a server. In another embodiment, the computing device 102 represents a computing system that utilizes clustered computers and components to act as a single seamless resource pool. In other embodiments, the computing device 102 may represent a server computing system that utilizes multiple computers as a server system, such as in a cloud computing environment. In general, according to embodiments of the present disclosure, the computing device 102 may be any computing device or combination of devices that has access to the opportunity discovery module 110 and the network 108 and is capable of processing program instructions and executing the opportunity discovery module 110. The computing device 102 may include internal and external hardware components, such as those described with respect to Figure 5 More detailed and described.

[0030] Further, in the depicted embodiment, computing device 102 includes opportunity discovery module 110. In the depicted embodiment, opportunity discovery module 110 is located on computing device 102. However, in other embodiments, opportunity discovery module 110 may be located externally and accessed through a communication network, such as network 108. The communication network may be, for example, a local area network (LAN), a wide area network (WAN) such as the Internet, or a combination of both, and may include wired, wireless, fiber optic, or any other connection known in the art. In general, the communication network may be any combination of connections and protocols that will support communication between computing device 102 and opportunity discovery module 110, in accordance with desired embodiments of the present disclosure.

[0031] In the illustrated embodiment, the opportunity discovery module 110 includes an LSTM autoencoder 112, a prediction model 114, a variable identification module 116, a policy recommendation module 118, and an output module 120. In the illustrated embodiment, the LSTM autoencoder 112, the prediction model 114, the variable identification module 116, the policy recommendation module 118, and the output module 120 are located on the computing device 102. However, in other embodiments, the LSTM autoencoder 112, the prediction model 114, the variable identification module 116, the policy recommendation module 118, and the output module 120 may be located externally and accessed via a communication network, such as the network 108.

[0032] In one or more embodiments, the opportunity discovery module 110 is configured to monitor time series data 104 generated from one or more sensors. For example, the time series data 104 may be data from oil sands operations and production processes. Petroleum products can be produced from oil sands through several stages, such as extraction, upgrading, and refining. For example, solids and water can be removed during the extraction stage, which can extract asphalt from the oil sands. During the upgrading stage, asphalt can be upgraded to a lighter intermediate crude oil product. During the refining stage, crude oil can be refined into final products such as gasoline, lubricants, and diluents. The processes during these stages can involve multiple sequential steps of physical or chemical transformation to convert one material into another. An optimal balance of processes is required to achieve multiple goals within such a production system. In another example, the time series data 104 may be data from a steelmaking process that produces steel from iron ore and / or scrap. Impurities such as nitrogen, silicon, phosphorus, sulfur, and excess carbon can be removed from the source iron, and alloying elements such as manganese, nickel, chromium, carbon, and vanadium can be added to produce different grades of steel. In another example, the time series data 104 may be data from a production process that processes soybeans into a soy sauce to which additives are added. In yet another example, the time series data 104 may be data from any other suitable operation and production process.

[0033] In one or more embodiments, the opportunity discovery module 110 is configured to extract a set of features from the time series data 104 based on one or more uncontrolled variables 122 for the time series data 104 using a recurrent neural network, such as an LSTM autoencoder 112. The opportunity discovery module 110 may not only learn using the set of features, but also learn the set of features through the LSTM autoencoder 112. The set of features is information related to the time series data 104. A set of features may be individual, measurable attributes or characteristics of a phenomenon observed in the time series data 104. The opportunity discovery module 110 may select a subset of relevant features for creating a predictive model 114 based on the uncontrolled variables 122 that do not define a user's control over the time series data 104. The opportunity discovery module 110 may reduce the number of resources required to describe the time series data 104. The opportunity discovery module 110 may construct a combination of uncontrolled variables 122 that describes the time series data 104 with sufficient accuracy through the LSTM autoencoder 112.

[0034] The LSTM autoencoder 112 may be an artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may be a recurrent neural network that implements an autoencoder for time series data 104 using an encoder-decoder LSTM architecture. The LSTM autoencoder 112 may learn a representation (encoding) of the time series data 104, for example, for dimensionality reduction. The LSTM autoencoder 112 may generate a representation that is as close as possible to the original input of the time series data 104 from the simplified encoding. The LSTM autoencoder 112 may include an encoder and a decoder. The encoder may use the original data (e.g., time series data 104) as input and generate features or representations as output, while the decoder may use features extracted from the encoder as input and reconstruct the original data of the original input as output. Training may be repeated until some stopping criteria is met. The LSTM autoencoder 112 may invoke an autoencoding process based on the time series data 104 to reduce the dimensionality of the sensor label space to a finite embedding space. The LSTM autoencoder 112 may invoke a clustering method to generate latent patterns. The LSTM autoencoder 112 may invoke a type of artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may obtain a fixed-size vector from the time series data 104 .

[0035] In one or more embodiments, the opportunity discovery module 110 is configured to identify one or more operating modes based on extracted features, including dimensionality reduction using representation learning from the time series data 104. Dimensionality reduction can be a transformation of the time series data 104 from a high-dimensional space to a low-dimensional space, such that the low-dimensional representation retains some meaningful properties of the original data, ideally close to the intrinsic dimensionality. The opportunity discovery module 110 can identify neighborhoods of the current operating state. Neighbors can be dynamic patterns within the same operating mode and can be found using the Euclidean distance between historical operating states and the current operating state. The opportunity discovery module 110 can define neighborhoods or use Gaussian mixture clustering applied to embedding to identify static operating modes. A pattern can include the state of plant operation. The hard category type of the pattern can consider the specific operating configuration of the entire production process, such as the manufacturing pipeline configuration or the unit operation status. The soft category type of the pattern can include the production level of the local or global operation. In the oil sands industry, there is a complex process for converting oil sands into synthetic crude oil. To complete the production of synthetic crude oil, multiple stages are involved, including primary extraction, secondary extraction, and upgrading. Each phase involves multiple components and processes, and the system is dynamic. Some of the patterns involved may be explicitly known to field engineers. Other patterns may be hidden and can be identified through advanced analytical models. The opportunity discovery module 110 can automatically detect operating patterns by learning from historical sensor and other production data, and can implement compact feedback in the form of patterns. The opportunity discovery module 110 can determine operating pattern algorithms and models. If similar production conditions have existed in the past, or if a new pattern can be identified, the opportunity discovery module 110 can use the operating pattern as a benchmark. The operating pattern can be used to recommend better control actions or automatically change control parameters. If a new pattern is identified, the new pattern can be saved to expand the system's memory and knowledge. When the plant is in the same mode, the opportunity discovery module 110 can calculate a historical benchmark of the detected pattern against historical plant data. The opportunity discovery module 110 can use the detected pattern as a basis for identifying similar time periods from history. For example, time periods from history can be identified when the same mode was active. The opportunity discovery module 110 can also use factor analysis on sets of process variables that collectively define detected patterns to identify key differences in current versus historical variable values. The opportunity discovery module 110 can calculate possible root causes of low-level operations / production and display them to field engineers. The opportunity discovery module 110 can provide analytical methods for identifying or classifying patterns across complex manufacturing processes. The opportunity discovery module 110 can provide detailed or general multivariate pattern identification and, as a byproduct, can also enable segmentation of unsegmented data into subsets.Rather than relying on rule-based pattern detection that requires extensive prior knowledge and stored principles, the opportunity discovery module 110 can provide an automated process, thereby providing a less human-involved approach to pattern generation. The opportunity discovery module 110 can implement opportunity discovery through analysis using unsupervised machine learning. For example, the opportunity discovery module 110 can identify operational opportunities by comparing the current state to a pattern or historical similar operations within a neighborhood.

[0036] In one or more embodiments, the opportunity discovery module 110 is configured to compare the current operating state with the historical operating state based on time series data in the same operating mode in one or more operating modes. The opportunity discovery module 110 can identify the specific mode in which the current operating state is located. The opportunity discovery module 110 can use t-distributed stochastic neighbor embedding (t-SNE) compression to project the clusters to generate a graph. t-SNE is a dimensionality reduction algorithm suitable for visualizing high-dimensional data (e.g., time series data 104). The opportunity discovery module 110 can embed high-dimensional points in low dimensions in a manner that respects the similarity between points with t-SNE compression. In an instance, the opportunity discovery module 110 can achieve high asphalt extraction by comparing the current operating state with other operations in the same mode. The opportunity discovery module 110 can analyze events of poor performance and can discover operational opportunities for improvement. The opportunity discovery module 110 can focus on these operational events that have historically reached high foam generation and can extract key operational actions from these events to help current operations.

[0037] In one or more embodiments, opportunity discovery module 110 is configured to discover operational opportunities based on a comparison of the current operating state with historical operating states. Operational opportunities can be identified by comparing the current state with historical similar operations within a pattern or neighborhood. In an example, an operational opportunity can be a set of operational changes derived from the historical operating state to increase the current operating state to a higher output in a defined short-term future period (e.g., a two-hour window). In another example, an operational opportunity can be a set of operational changes derived from the historical operating state to reduce the current operating state to a low usage rate of an additive or raw material in a defined short-term future period. Other suitable opportunities may be found.

[0038] In one or more embodiments, the opportunity discovery module 110 is configured to identify control variables 124 that are in the same pattern and are related to the current operating state. The control variables 124 can define actions that can be controlled by the user. The control variables 124 can be time-stamped variables and can be defined as a set of variables from sensors for actions that can be controlled by the user. In an example, the control variables 124 can be used to calculate rewards (or opportunities) from the episodes of the time series data 104. For example, the control variables 124 can be production rate and raw material variables that the user can optimize based on the best neighboring events found. The control variables 124 can be identified from the established neighborhood of similar historical non-control variables 122 to create possible action strategies related to the current state based on the time series data 104. In the foam production example, the control variables 124 can include, for example, diesel hydrotreating or catalytic hydrotreating, with the feed plate being loaded with diesel hydrotreating having low vacuum gas oil, low vacuum gas oil, side cut kerosene, heavy naphtha, and coker kerosene. Diesel hydrotreating, or catalytic hydrogenation, can primarily reduce undesirable species from the straight-run diesel fraction by selectively reacting these species with hydrogen in a reactor at elevated temperatures and moderate pressures. To successfully produce ultra-low sulfur diesel, organic sulfur species, including substituted dibenzothiophenes and other refractory sulfur species, need to be removed. Multiple reactions can occur in parallel on the surface of a diesel hydrotreating catalyst, including hydrodesulfurization, hydrodenitrogenation, and aromatic saturation / hydrogenation.

[0039] In one or more embodiments, the opportunity discovery module 110 is configured to recommend an action strategy based on the controlled variables 124, the uncontrolled variables 122, and the target productivity. The opportunity discovery module 110 may define a similarity measure to identify historical episodes from the time series data 104 with similar operating states based on a comparison of the current state with the historical operating states. The opportunity discovery module 110 may create episodes from the historical episodes that exhibit higher productivity or throughput. The opportunity discovery module 110 may generate scores based on the alternative action strategies and may recommend an action strategy based on the score of each alternative action strategy.

[0040] In one or more embodiments, the opportunity discovery module 110 is configured to output an action strategy to a user. The opportunity discovery module 110 may provide a user interface for interacting with the user. The opportunity discovery module 110 may provide other suitable output methods to the user. The opportunity discovery module 110 may provide an indicator or alert to the user regarding the discovered action opportunities. The opportunity discovery module 110 may use the t-SNE method to display the action pattern as a graph. The opportunity discovery module 110 may use a time-stamped chart to display the neighborhood plot. The opportunity discovery module 110 may display the estimated gain of the action strategy.

[0041] In one or more embodiments, the LSTM autoencoder 112 is configured to extract a set of features from the time series data 104 based on the uncontrolled variables 122 of the time series data 104. In an example, the LSTM autoencoder 112 may be an artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may be capable of automatically extracting the effects of past events. The LSTM autoencoder 112 may be a recurrent neural network that implements an autoencoder for the time series data 104 using an encoder-decoder LSTM architecture. The LSTM autoencoder 112 may learn a representation (encoding) of the time series data 104, for example, for dimensionality reduction. The LSTM autoencoder 112 may generate a representation that is as close as possible to the original input of the time series data 104 from the simplified encoding. The LSTM autoencoder 112 may include an encoder and a decoder. The encoder may use the original data (e.g., the time series data 104) as input and may generate features or representations as output, while the decoder may use the features extracted from the encoder as input and may reconstruct the original data of the original input as output. The LSTM autoencoder 112 may be repeatedly trained until some stopping criteria are met. The LSTM autoencoder 112 may utilize an autoencoding process to reduce the dimensionality of the sensor tag space to a finite embedding space. The LSTM autoencoder 112 may utilize a clustering method to generate latent patterns. The LSTM autoencoder 112 may utilize a type of artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may obtain a fixed-size vector from the time series data 104. The LSTM autoencoder 112 may not only learn from a set of features, but also learn the set of features themselves. The set of features may be information related to the time series data 104. The feature set may be individual measurable attributes or characteristics of a phenomenon observed in the time series data 104. The LSTM autoencoder 112 may select a subset of relevant features for creating a predictive model 114 based on uncontrolled variables 122 that do not define user control over the time series data 104. The LSTM autoencoder 112 may reduce the number of resources required to describe the time series data 104. The LSTM autoencoder 112 may construct a combination of uncontrolled variables 122 that describes the time series data 104 with sufficient accuracy.

[0042] In one or more embodiments, the forecasting model 114 is configured to discover operational opportunities based on the time series data 104. Operational opportunities can be identified by comparing the current state to similar historical operations within a pattern or neighborhood. In one example, an operational opportunity can be a set of operational changes derived from historical operational states to increase the current state to a higher production output within a defined near-future time period. In another example, an operational opportunity can be a set of operational changes derived from historical operational states to reduce the current state to a lower usage rate of additives or raw materials within a defined near-future time period. Other suitable opportunities can be identified.

[0043] In one or more embodiments, the prediction model 114 is configured to identify one or more operating modes based on extracted features, including dimensionality reduction using representations learned from the time series data 104. The prediction model 114 may identify neighborhoods of the current operating state. Neighbors may be dynamic patterns within the same operating mode and can be found using the Euclidean distance between historical operating states and the current operating state. The prediction model 114 may define neighborhoods or use Gaussian mixture clustering applied to embedding to identify static operating modes. Modes may include plant operating states. In the oil sands industry, converting oil sands into synthetic crude oil involves a complex process. Synthetic crude oil production involves multiple stages, including primary extraction, secondary extraction, and upgrading. Each stage involves multiple components and processes. Some of the involved modes may be explicitly known to field engineers. Others may be hidden and can be identified through advanced analytical models. The prediction model 114 can automatically detect operating modes by learning from historical sensor and other production data, and can implement compact feedback in the form of patterns. If similar production conditions have existed in the past, or if new patterns can be identified, the prediction model 114 can use operating modes as benchmarks. The predictive model 114 can use operating patterns to recommend better control actions or automatically change control parameters. The predictive model 114 can calculate historical benchmarks of detected patterns against historical plant data when the plant was operating in the same pattern. The predictive model 114 can use the detected patterns as a basis for identifying similar time periods from history. For example, time periods from history when the same pattern was active can be identified. The predictive model 114 can further use factor analysis on the set of process variables that jointly define the detected pattern to identify key differences between current and historical variable values. The predictive model 114 can calculate possible root causes of low levels of operation / production and display them to field engineers. The predictive model 114 can provide an automated process, thereby simplifying human interaction in generating patterns. The predictive model 114 can implement opportunity realization through analysis using unsupervised machine learning. For example, the predictive model 114 can identify operational opportunities by comparing the current state with historical similar operations within a pattern or neighborhood. The predictive model 114 can identify specific patterns within which the current operating state resides. The predictive model 114 can project clusters using t-SNE compression to generate a graph. T-SNE can be a dimensionality reduction algorithm suitable for visualizing high-dimensional data (e.g., time series data 104). Prediction model 114 can embed high-dimensional points into low dimensions in a manner that respects the similarity between points through t-SNE compression. In one example, prediction model 114 compares the current operating state with other operations in the same mode to achieve high bitumen extraction. Prediction model 114 can analyze instances of poor performance and identify opportunities for improved operations.The predictive model 114 may focus on those operational episodes in history that reached high foam generation, and may extract key operational actions from these episodes to aid current operations.

[0044] In one or more embodiments, the variable identification module 116 is configured to identify non-control variables 122 and control variables 124 from the time series data 104. The non-control variables 122 and control variables 124 can be separated to ensure the ability to take actions to obtain production enhancement opportunities based on the time series data 104. For example, the non-control variables 122 can be time-stamped variables and can be defined as a set of variables from sensors over which the user has little or no control. The non-control variables 122 can be parameters used to define how similar operating and production conditions are. The non-control variables 122 can retrieve episodes from the time series data 104. The control variables 124 can be time-stamped variables and can be defined as a set of variables from sensors for actions that the user can control. In an example, the control variables 124 can be used to calculate rewards (or opportunities) from episodes in the time series data. The control variables 124 can be production rate and raw material variables that the user can optimize based on the best neighboring events found. The control variables 124 can be identified from an established neighborhood of similar historical non-control variables 122 to create a possible action strategy related to the current state based on the time series data 104. In one or more embodiments, the variable identification module 116 is configured to identify one or more control variables 124 in the same mode that are relevant to the current operating state.

[0045] In one or more embodiments, the strategy recommendation module 118 is configured to recommend an action strategy based on the controlled variables 124, the uncontrolled variables 122, and the target productivity. The strategy recommendation module 118 can define a similarity measure based on a comparison of the current state with the historical operating state to identify historical episodes from the time series data 104 with similar operating states. The strategy recommendation module 118 can create episodes that demonstrate higher productivity or throughput from the historical episodes. The strategy recommendation module 118 can generate scores based on the alternative action strategies and can recommend an action strategy based on the scores for each alternative action strategy.

[0046] In one or more embodiments, the output module 120 is configured to output the user's action strategy. The output module 120 may provide a user interface for user interaction. The output module 120 may also provide other suitable output methods to the user. The output module 120 may provide the user with alerts regarding discovered action opportunities. The output module 120 may use the t-SNE method to display action patterns. The output module 120 may use a time-stamped chart to display neighborhood plots. The output module 120 may display the estimated gain of the action strategy.

[0047] Figure 2 is a flowchart depicting the operating steps of the opportunity discovery module 110 according to an embodiment of the present disclosure.

[0048] The opportunity discovery module 110 operates to monitor time series data 104 generated from one or more sensors. The opportunity discovery module 110 also operates to extract a set of features from the time series data 104 based on the non-controlled variables 122 of the time series data 104 by auto-encoding using a neural network (e.g., an LSTM autoencoder 112). The opportunity discovery module 110 operates to identify one or more operating patterns based on the extracted features, including dimensionality reduction using representations learned from the time series data 104. The opportunity discovery module 110 operates to compare the current operating state with historical operating states within the same operating mode based on the time series data 104. The opportunity discovery module 110 is configured to discover operating opportunities based on the comparison of the current operating state with the historical operating states. The opportunity discovery module 110 operates to identify controlled variables 124 that are in the same pattern as the current operating state, which are correlated with the current operating state. The opportunity discovery module 110 operates to recommend an action strategy based on the controlled variables 124, the non-controlled variables 122, and the target productivity. The opportunity discovery module 110 operates to output the action strategy to a user.

[0049] In step 202, the opportunity discovery module 110 monitors the time series data 104 generated from one or more sensors. For example, the time series data 104 may be data from oil sands operations and production processes. Petroleum products can be produced from oil sands through several stages, such as extraction, upgrading, and refining. For example, solids and water can be removed during the extraction stage, which can extract asphalt from the oil sands. During the upgrading stage, asphalt can be upgraded to a lighter intermediate crude oil product. During the refining stage, crude oil can be refined into final products such as gasoline, lubricants, and diluents. The processes during these stages can involve multiple sequential steps of physical or chemical transformation to convert one material into another. An optimal balance of processes is required to achieve multiple goals within such a production system. In another example, the time series data 104 may be data from a steelmaking process that produces steel from iron ore and / or scrap. Impurities such as nitrogen, silicon, phosphorus, sulfur, and excess carbon can be removed from the source iron, and alloying elements such as manganese, nickel, chromium, carbon, and vanadium can be added to produce different grades of steel. In another example, the time series data may be data from a production process that processes soybeans into a soy sauce to which additives are added. In yet another example, the time series data may be data from any other suitable operation and production process.

[0050] In step 204, the opportunity discovery module 110 extracts a set of features from the time series data 104 based on one or more uncontrolled variables 122 of the time series data 104 by autoencoding using a neural network (e.g., an LSTM autoencoder 112). The opportunity discovery module 110 may not only learn using the set of features, but also learn the set of features through the LSTM autoencoder 112. The set of features may be information related to the time series data 104. A feature set may be individual measurable attributes or characteristics of a phenomenon observed from the time series data 104. The opportunity discovery module 110 may select a subset of relevant features for creating a predictive model 114 based on the uncontrolled variables 122, which define a user's lack of control over the time series data 104. The opportunity discovery module 110 may reduce the number of resources required to describe the time series data 104. The opportunity discovery module 110 may construct a combination of uncontrolled variables 122 that describes the time series data 104 with sufficient accuracy through the LSTM autoencoder 112. The LSTM autoencoder 112 may be an artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may be a recurrent neural network that implements an autoencoder for time series data 104 using an encoder-decoder LSTM architecture. The LSTM autoencoder 112 may learn a representation (encoding) of the time series data 104, for example, for dimensionality reduction. The LSTM autoencoder 112 may generate a representation that is as close as possible to the original input of the time series data 104 from the simplified encoding. The LSTM autoencoder 112 may include an encoder and a decoder. The encoder may use the original data (e.g., the time series data 104) as input and may generate features or representations as output, while the decoder may use features extracted from the encoder as input and may reconstruct the original data of the original input as output. Training may be repeated until some stopping criteria is met. The LSTM autoencoder 112 may use an autoencoding process to train the time series data 104 to reduce the dimensionality of the sensor tag space to a finite embedding space. The LSTM autoencoder 112 may employ clustering methods to generate latent patterns. The LSTM autoencoder 112 may invoke a type of artificial neural network for learning efficient data encoding in an unsupervised manner. The LSTM autoencoder 112 may obtain a fixed-size vector from the time series data 104 .

[0051] In step 206, the opportunity discovery module 110 identifies one or more operating patterns based on the extracted features, including dimensionality reduction using representations learned from the time series data 104. The opportunity discovery module 110 may identify neighborhoods of the current operating state. Neighbors can be dynamic patterns within the same operating mode and can be found using the Euclidean distance between historical operating states and the current operating state. The opportunity discovery module 110 may define neighborhoods or use Gaussian mixture clustering applied to embedding to identify static operating patterns. Patterns may include plant operating states. Hard category types for patterns may consider the specific operating configuration of the entire production process, such as manufacturing pipeline configuration or unit operation status. Soft category types for patterns may include local or global production levels. In the oil sands industry, the conversion of oil sands into synthetic crude oil is a complex process. Synthetic crude oil production involves multiple stages, including primary extraction, secondary extraction, and upgrading. Each stage involves multiple components and processes, and the system is dynamic. Some of the patterns involved are explicitly known to field engineers. Other patterns may be hidden and can be identified through advanced analytical models. The opportunity discovery module 110 can automatically detect operating patterns by learning from historical sensor and other production data, and can implement compact feedback in the form of patterns. The opportunity discovery module 110 can determine operating pattern algorithms and models. If similar production conditions have existed in the past, or if a new pattern can be identified, the opportunity discovery module 110 can use the operating pattern as a benchmark. The operating pattern can be used to recommend better control actions or automatically change control parameters. If a new pattern is identified, the new pattern can be saved to expand the system's memory and knowledge. When the plant is in the same pattern, the opportunity discovery module 110 can calculate a historical benchmark of the detected pattern against historical plant data. The opportunity discovery module 110 can use the detected pattern as a basis for identifying similar time periods from history. For example, time periods from history when the same pattern was active can be identified. The opportunity discovery module 110 can also use factor analysis on the set of process variables that collectively define the detected pattern to identify key differences in current variable values ​​compared to historical variable values. The opportunity discovery module 110 can calculate the possible root causes of low-level operations / production and display them to field engineers. Opportunity discovery module 110 provides analytical methods for identifying or classifying patterns in complex manufacturing processes. Opportunity discovery module 110 can provide detailed or general multivariate pattern identification and, as a byproduct, can also segment unsegmented data into subsets. Rather than relying on rule-based pattern detection that requires prior knowledge and stored principles, opportunity discovery module 110 can provide an automated process, thereby providing a less human-involved approach to pattern generation. Opportunity discovery module 110 can achieve opportunity realization through analysis using unsupervised machine learning.For example, the opportunity discovery module 110 may identify an operation opportunity by comparing the current state to a pattern or historical similar operations within a neighborhood.

[0052] In step 208, the opportunity discovery module 110 compares the current operating state with the historical operating state based on the time series data in the same operating mode in one or more operating modes. The opportunity discovery module 110 can identify the specific mode in which the current operating state is located. The opportunity discovery module 110 can use t-SNE compression to project the clusters to generate a graph. In an example, T-SNE can be an algorithm for dimensionality reduction suitable for visualizing high-dimensional data (e.g., time series data 104). The opportunity discovery module 110 can embed high-dimensional points in low dimensions in a manner that respects the similarity between points using t-SNE compression. In an instance, the opportunity discovery module 110 can achieve high asphalt extraction by comparing the current operating state with other operations in the same mode. The opportunity discovery module 110 can analyze events of poor performance and can discover operational opportunities for improvement. The opportunity discovery module 110 can focus on these operational events that have historically reached high foam generation and can extract key operational actions from these events to help current operations.

[0053] At step 210, opportunity discovery module 110 discovers operational opportunities based on a comparison of the current operating state with historical operating states. Operational opportunities can be identified by comparing the current state with similar historical operations within a pattern or neighborhood. In one example, an operational opportunity can be a set of operational changes derived from historical operating states to increase the current operating state to a higher output within a defined near-future time period. In another example, an operational opportunity can be a set of operational changes derived from historical operating states to reduce the current operating state to a lower usage rate of additives or raw materials within a defined near-future time period. Other suitable opportunities can be identified.

[0054] In step 212, the opportunity discovery module 110 identifies one or more control variables 124 in the same pattern that are relevant to the current operating state. The control variables 124 can define actions that can be controlled by the user. The control variables 124 can be time-stamped variables and can be defined as a set of variables from sensors for actions that can be controlled by the user. In an example, the control variables 124 can be used to calculate rewards (or opportunities) from the episodes of the time series data 104. For example, the control variables 124 can be production rate and raw material variables that the user can optimize based on the best neighboring events found. The control variables 124 can be identified from the established neighborhood of similar historical non-control variables 122 to create a possible action strategy related to the current state based on the time series data 104. In the foam production example, the control variables 124 can include, for example, loading diesel hydrotreating or catalytic hydrotreating, and the feed plate is loaded with diesel hydrotreating with low vacuum gas oil, low vacuum gas oil, side cut kerosene, heavy naphtha and coker kerosene.

[0055] In step 214, the opportunity discovery module 110 recommends an action strategy based on the controlled variables 124, the uncontrolled variables 122, and the target productivity. The opportunity discovery module 110 may define a similarity measure to identify historical episodes from the time series data 104 with similar operating states based on a comparison of the current state with the historical operating states. The opportunity discovery module 110 may create episodes from the historical episodes that exhibit higher productivity or throughput. The opportunity discovery module 110 may generate scores based on the alternative action strategies and may recommend an action strategy based on the score of each alternative action strategy.

[0056] At step 216, the opportunity discovery module 110 outputs the action strategy to the user. The opportunity discovery module 110 may provide a user interface for interacting with the user. The opportunity discovery module 110 may provide other suitable output methods to the user. The opportunity discovery module 110 may indicate a signal indicating the discovered action opportunity. The opportunity discovery module 110 may use the t-SNE method to display the action pattern. The opportunity discovery module 110 may use a time-stamped chart to display the neighborhood plot. The opportunity discovery module 110 may display the estimated gain of the action strategy.

[0057] Figure 3 An exemplary functional diagram of the opportunity discovery module 110 according to one or more embodiments of the present disclosure is shown.

[0058] exist Figure 3In the example of , the opportunity discovery module 110 can implement opportunity realization through analysis using unsupervised machine learning. Operational opportunities can be identified by comparing the current state (or scenario) with historical similar episodes within a pattern or neighborhood. The opportunity discovery module 110 can monitor operations as time series data 104 from sensors. The opportunity discovery module 110 can apply an LSTM autoencoder 112 to extract features of the time series data 104 into an embedded space. The LSTM autoencoder 112 can implement time series representation learning with dimensionality reduction. The LSTM autoencoder 112 can learn a representation (encoding) of the time series data 104, for example, for dimensionality reduction. The LSTM autoencoder 112 can generate a representation that is as close as possible to the original input of the time series data 104 from the simplified encoding. In the illustrated embodiment, the LSTM autoencoder 112 includes an encoder 320 and a decoder 322. The encoder 320 can take the original input 324 (e.g., time series data 104) as input and can generate features or representations as output. Decoder 322 may use the features extracted from encoder 320 as input and may reconstruct original input 324 into reconstructed input 326 as output. Opportunity discovery module 110 may use Gaussian mixture clustering applied to the embedded space to identify one or more static operating patterns 302 (e.g., pattern 304). Opportunity discovery module 110 may project the clusters using t-SNE compression to generate a graph of patterns 302. Patterns 302 may serve as a benchmark against which the system can determine whether similar production conditions have existed in the past. Patterns 302 may be used to recommend better control actions or automatically change control parameters. Opportunity discovery module 110 may use the embedded space to identify dynamic patterns (e.g., neighborhood 308) of current operating state 306. Opportunity discovery module 110 may discover operating opportunities based on, for example, operating patterns 304 or neighborhood 308. In an example, opportunity discovery module 110 may limit historical scenarios by selecting neighborhood 308 of current operating state 306 using the embedded space. Neighbors 308 may be found using the Euclidean distance between historical scenarios and current state 306 defined in the feature space. Block 310 demonstrates opportunity realization by comparing the current operational state to other operations in the same mode, such as mode 304. Operational opportunities demonstrated as poor performance in opportunity windows 312 may indicate opportunities for improvement.

[0059] Figure 4 FIG. 1 shows an exemplary architecture diagram of the opportunity discovery module 110 according to one or more embodiments of the present disclosure.

[0060] In block 402, a set of features may be selected based on historical data 404 (e.g., time series data 104) to create a predictive model 114. Predictive model 114 may identify potential manufacturing gains from reduced use of raw materials or additives without reducing production, or increased production of intermediate or final products with the same use of raw materials or additives. Predictive model 114 may predict the potential for lower productive use of raw materials or additives, or increased productivity with the same use of raw materials or additives. Predictive model 114 may invoke an opportunity identification algorithm to identify alternative operations to capture opportunities. Historical data 404 may be data captured by one or more sensors. In one example, historical data 404 may be data from oil sands operations and production processes. In another example, historical data 404 may be data from a steelmaking process that produces steel from iron ore and / or scrap. In yet another example, historical data 404 may be data from any other suitable operations and production processes. Predictive model 114 may be validated using control variables 124 and uncontrolled variables 122. The prediction model 114 may use non-controlled variables 122 (e.g., environmental variables over which the user has little or no control) to retrieve episodes from the historical data 404 via neighborhood selection 406. The prediction model 114 may use controlled variables 124 (e.g., action variables that the user may control or change during the process) to calculate rewards (or opportunities) from the episodes of the historical data 404 with clustering 408. The clustering 408 may group a set of objects from the historical data 404 such that objects in the same group (referred to as a cluster) are more similar (in some sense) to each other than objects in other groups (clusters).

[0061] Alternative control strategy 410 may be an action strategy for selecting appropriate actions and changing the environment to achieve the ultimate opportunity.

[0062] Alternative control strategies 410 may allow a user or project manager to obtain real-time support for decision-making to increase productivity. For example, the predictive model 114 may provide the user with a score 412 and a ranking 414 of a recommended list 416 for execution. In an example, a user interface may be provided to the user on a portable device. The recommended list 416 may include, for example, a list of potential target quantities for raw materials, intermediate and final products, or expensive additives.

[0063] Figure 5 A block diagram 500 of components of a computing device 102 is depicted according to an illustrative embodiment of the present disclosure. It should be understood that Figure 5 This merely provides an illustration of one implementation and does not imply any limitations with regard to the environments in which different embodiments may be implemented. Many modifications may be made to the depicted environments.

[0064] The computing device 102 may include a communications fabric 502 that provides communications between a cache 516, memory 506, persistent storage 508, a communications unit 510, and an input / output (I / O) interface 512. The communications fabric 502 may be implemented using any architecture designed to transfer data and / or control information between processors (such as microprocessors, communications and network processors, etc.), system memory, peripheral devices, and any other hardware components within the system. For example, the communications fabric 502 may be implemented using one or more buses or crossbar switches.

[0065] Memory 506 and persistent storage 508 are computer-readable storage media. In this embodiment, memory 506 includes random access memory (RAM). In general, memory 506 may include any suitable volatile or non-volatile computer-readable storage media. Cache 516 is a fast memory that enhances the performance of computer processor 504 by storing recently accessed data and data that will be accessed soon from memory 506.

[0066] The opportunity discovery module 110 can be stored in persistent storage 508 and memory 506 for execution by one or more of the corresponding computer processors 504 via cache 516. In an embodiment, persistent storage 508 includes a magnetic hard drive. Alternatively, or in addition to the magnetic hard drive, persistent storage 508 can include a solid-state drive, a semiconductor memory device, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), flash memory, or any other computer-readable storage medium capable of storing program instructions or digital information.

[0067] The media used by persistent storage 508 also may be removable. For example, a removable hard drive may be used for persistent storage 508. Other examples include optical and magnetic disks, thumb drives, and smart cards, which are inserted into a drive for transfer to another computer-readable storage medium that is also part of persistent storage 508.

[0068] In these examples, communication unit 510 provides for communication with other data processing systems or devices. In these examples, communication unit 510 includes one or more network interface cards. Communication unit 510 may provide for communication using either or both physical and wireless communication links. Opportunity discovery module 110 may be downloaded to persistent storage 508 via communication unit 510.

[0069] (One or more) I / O interface 512 allows input and output of data with other devices that may be connected to computing device 102. For example, I / O interface 512 can provide connection to external devices 518 (such as a keyboard, keypad, touch screen and / or some other suitable input device). External devices 518 can also include portable computer-readable storage media, such as, for example, a thumb drive, a portable optical or magnetic disk, and a memory card. Software and data for practicing embodiments of the present invention (e.g., opportunity discovery module 110) can be stored on such portable computer-readable storage media and can be loaded into permanent storage 508 via I / O interface 512. I / O interface 512 is also connected to display 520.

[0070] Display 520 provides a mechanism for displaying data to a user and may be, for example, a computer monitor.

[0071] The programs described herein are identified based on the applications implemented in the specific embodiments of the present invention. However, it should be understood that any specific program terminology herein is used for convenience only, and thus the present invention should not be limited to use only in any specific application identified and / or implied by such terminology.

[0072] The present invention may be a system, method and / or computer program product of any possible degree of technical detail integration. The computer program product may include a computer-readable storage medium (or multiple media) having computer-readable program instructions thereon for causing a processor to execute various aspects of the present invention.

[0073] Computer-readable storage media can be a tangible device that can retain and store the instructions used by the instruction execution device.Computer-readable storage media can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the above. A non-exhaustive list of more specific examples of computer-readable storage media includes the following: portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), static random access memories (SRAM), portable compact disc read-only memories (CD-ROM), digital versatile discs (DVD), memory sticks, floppy disks, mechanical encoding devices such as punch cards, or projection structures in the grooves with instructions recorded thereon, and any suitable combination of the above. Computer-readable storage media as used herein should not be interpreted as temporary signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated by waveguides or other transmission media (for example, light pulses passing through fiber optic cables), or electrical signals emitted by wires.

[0074] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to a corresponding computing / processing device via a network (e.g., the Internet, a local area network, a wide area network, and / or a wireless network), or downloaded to an external computer or external storage device. The network can include copper transmission cables, optical transmission fibers, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in a computer-readable storage medium within the corresponding computing / processing device.

[0075] The computer-readable program instructions for performing the operation of the present invention can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, configuration data for integrated circuits, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​(such as Python, C++, etc.) and procedural programming languages ​​(such as " C " programming language or similar programming languages). The computer-readable program instructions can be executed completely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer, partially on a remote computer, or completely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any type of network (including a local area network (LAN) or a wide area network (WAN)), or can be connected to an external computer (for example, using an internet service provider through the internet). In certain embodiments, the electronic circuit comprising, for example, a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA) can be personalized to perform the computer-readable program instructions by utilizing the state information of the computer-readable program instructions to perform various aspects of the present invention.

[0076] The present invention will be described below with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0077] These computer-readable program instructions can be provided to a processor of a computer or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device create a device for implementing the functions / actions specified in the flowchart and / or block diagram or multiple blocks. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to operate in a specific manner. Thus, the computer-readable storage medium having the instructions stored therein includes an article of manufacture containing instructions that implement aspects of the functions / actions specified in the flowchart and / or block diagram or multiple blocks.

[0078] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, so that a series of operational steps are performed on the computer, other programmable apparatus, or other device to produce computer-implemented processing, so that the instructions executed on the computer, other programmable apparatus, or other device implement the functions / actions specified in the flowchart and / or block diagram or multiple boxes.

[0079] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functions and operations of possible implementations of the systems, methods and computer program products according to different embodiments of the present invention. To this end, each box in the flowchart or block diagram may represent a module, segment or portion of an instruction, which includes one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the box may not occur in the order marked in the figure. For example, the two boxes shown in succession can actually be completed as a step, performed simultaneously, substantially simultaneously, in a partially or completely time-overlapping manner, or the boxes can sometimes be performed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or action or performs a combination of dedicated hardware and computer instructions.

[0080] The description of various embodiments of the present invention has been presented for illustrative purposes, but is not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the invention. The terminology used herein is selected to best explain the principles of the embodiments, practical applications, or technical improvements over technologies found in the marketplace, or to enable those skilled in the art to understand the embodiments disclosed herein.

[0081] Although specific embodiments of the present invention have been described, those skilled in the art will appreciate that there are other embodiments that are equivalent to the described embodiments. Therefore, it should be understood that the present invention is not limited to the specific embodiments shown, but only by the scope of the appended claims.

Claims

1. A computer-implemented method for productivity-enhancing operational opportunity discovery for a production process, comprising: extracting, by one or more processors, a set of features from the time series data generated from one or more sensors based on one or more uncontrolled variables of the time series data by autoencoding using a neural network, the one or more uncontrolled variables defining a lack of control by a user over the time series data; monitoring, by one or more processors, the time series data generated from the one or more sensors; identifying, by one or more processors, one or more operating modes based on the extracted features, the features comprising dimensionality reduction using a representation learned from the time series data; identifying, by one or more processors, a neighborhood of the current operating state based on the extracted features, the neighborhood being a dynamic mode within the same operating mode; In the same operating mode of the one or more operating modes, comparing, by the one or more processors, a current operating state with one or more historical operating states based on the time series data; using the neighborhood, by one or more processors, to discover operating opportunities based on the comparison of the current operating state with the one or more historical operating states; identifying, by one or more processors, one or more control variables in a same mode, the variables being associated with the current operating state, the one or more control variables defining actions controllable by the user; as well as An action strategy is recommended by one or more processors based on the one or more controlled variables, the one or more non-controlled variables, and a target productivity.

2. The computer-implemented method of claim 1 , wherein: The neural network is a long short-term memory autoencoder.

3. The computer-implemented method of claim 1 , wherein: Discovering the operating opportunity is based on the comparison of the current operating state using the pattern with the one or more historical operating states.

4. The computer-implemented method of claim 1 , wherein: The operating opportunity is selected from the group consisting of: a set of operating changes inferred from the one or more historical operating states to increase the current operating state to a higher output within a defined near-term future time period, and a set of operating changes inferred from the one or more historical operating states to reduce the current operating state to a low usage rate of additives or raw materials within a defined near-term future time period.

5. The computer-implemented method of claim 1 , further comprising: defining, by one or more processors, a similarity measure based on a comparison of the current operating state with the one or more historical operating states to identify historical episodes having similar operating states from the time series data; as well as A scenario selected from the group consisting of: increased productivity and throughput is created by one or more processors from the historical scenarios.

6. The computer-implemented method of claim 1 , further comprising: Output the action strategy, the output including: Providing an alert to the user regarding the action opportunity, revealing the one or more modes of operation using a t-distributed random neighbor embedding method, Display neighborhood plots using time-stamped graphs, and Shows the estimated gain of the action policy.

7. A computer program product comprising: One or more computer-readable storage media, and program instructions stored together on the one or more computer-readable storage media, the program instructions comprising: program instructions for extracting a set of features from time series data generated from one or more sensors by auto-encoding the time series data using a neural network based on one or more uncontrolled variables of the time series data, the one or more uncontrolled variables defining a lack of control by a user over the time series data; program instructions for monitoring said time series data generated from said one or more sensors; program instructions for identifying one or more operating modes based on the extracted features, the features comprising dimensionality reduction using representations learned from the time series data; program instructions for identifying a neighborhood of the current operating state based on the extracted features, the neighborhood being a dynamic mode within the same operating mode; program instructions for comparing the current operating state with one or more historical operating states based on the time series data for the same one of the one or more operating modes; program instructions for using the neighborhood to discover operating opportunities based on the comparison of the current operating state to the one or more historical operating states; Program instructions for identifying one or more control variables in the same mode, the variables being associated with the current operating state, the one or more control variables defining actions that can be controlled by a user; and Program instructions are provided for recommending an action strategy based on the one or more controlled variables, the one or more uncontrolled variables, and a target productivity.

8. The computer program product of claim 7, wherein: The neural network is a long short-term memory autoencoder.

9. The computer program product of claim 7, wherein: Discovering the operating opportunity is based on the comparison of the current operating state using the pattern with the one or more historical operating states.

10. The computer program product of claim 7, wherein: The operating opportunity is selected from the group consisting of: a set of operating changes derived from the one or more historical operating states to increase the current operating state to a higher production rate within a defined near-term future time period, and a set of operating changes derived from the one or more historical operating states to reduce the current operating state to a low usage rate of additives or raw materials within a defined near-term future time period.

11. The computer program product of claim 7, further comprising: program instructions stored on the one or more computer-readable storage media for defining a similarity measure based on a comparison of the current operating state to the one or more historical operating states to identify historical episodes from the time series data having similar operating states; as well as Program instructions stored on the one or more computer-readable storage media are for creating from the historical stories a story selected from the group consisting of: increased productivity and throughput.

12. The computer program product of claim 7, further comprising: Program instructions for outputting the action strategy stored on the one or more computer-readable storage media, wherein the program instructions for outputting include program instructions: Providing an alert to the user regarding the action opportunity, revealing the one or more modes of operation using a t-distributed random neighbor embedding method, Display neighborhood plots using time-stamped graphs, and Displays the estimated gain of the action policy.

13. A computer system for discovering operational opportunities for productivity enhancement in a production process, comprising: One or more computer processors, one or more computer-readable storage media, and program instructions stored on the one or more computer-readable storage media for execution by at least one of the one or more computer processors, the program instructions comprising: program instructions for extracting a set of features from time series data generated from one or more sensors by auto-encoding the time series data using a neural network based on one or more uncontrolled variables of the time series data, the one or more uncontrolled variables defining a lack of control by a user over the time series data; program instructions for monitoring said time series data generated from said one or more sensors; program instructions for identifying one or more operating modes based on the extracted features, the features comprising dimensionality reduction using representations learned from the time series data; program instructions for identifying a neighborhood of the current operating state based on the extracted features, the neighborhood being a dynamic mode within the same operating mode; program instructions for comparing the current operating state with one or more historical operating states based on the time series data for the same one of the one or more operating modes; program instructions for using the neighborhood to discover operating opportunities based on the comparison of the current operating state to the one or more historical operating states; Program instructions for identifying one or more control variables in the same mode, the variables being associated with the current operating state, the one or more control variables defining actions that can be controlled by a user; and Program instructions are provided for recommending an action strategy based on the one or more controlled variables, the one or more uncontrolled variables, and a target productivity.

14. The computer system according to claim 13, wherein: The neural network is a long short-term memory autoencoder.

15. The computer system according to claim 13, wherein: Discovering the operating opportunity is based on the comparison of the current operating state using the pattern with the one or more historical operating states.

16. The computer system according to claim 13, wherein: The operating opportunity is selected from the group consisting of: a set of operating changes inferred from the one or more historical operating states to increase the current operating state to a higher output within a defined near-term future time period, and a set of operating changes inferred from the one or more historical operating states to reduce the current operating state to a low usage rate of additives or raw materials within a defined near-term future time period.

17. The computer system of claim 13, further comprising: Program instructions for outputting the action strategy stored on the one or more computer-readable storage media, wherein the program instructions for outputting include program instructions: Providing an alert to the user regarding the action opportunity, revealing the one or more modes of operation using a t-distributed random neighbor embedding method, Display neighborhood plots using time-stamped graphs, and Displays the estimated gain of the action policy.

Images