Abstract: A data pipeline is used as a fundamental processing element for implementing techniques that automatically or autonomously perform signal processing-based learning in a process plant or monitoring system. Each data pipeline includes a set of communicatively interconnected data processing blocks that perform processing on one or more sources of data in a predetermined order to, for example, clean the data, filter the data, select data for further processing, perform supervised or unsupervised learning on the data, etc. The individual processing blocks or modules within a data pipeline may be stored and executed at different devices in a plant network to perform distributed data processing. Moreover, each data pipeline can be integrated into one or more higher level analytic modules that perform higher level analytics, such as quality prediction, fault detection, etc. on the processed data.

Abstract: Techniques for automatically or autonomously performing signal processing-based learning in a process plant are disclosed. Generally, said techniques automatically or autonomously perform signal processing on a real-time signal that is generated based on the process plant controlling a process. Typically, the signal corresponds to a parameter value that varies over time, and the signal is processed as it is generated in real-time during on-line plant operations. Results of the signal processing may indicate characteristics of the signal, and one or more analytics functions may determine the sources of the characteristics, which may include a process element or device, a piece of equipment, and/or an asset of the process plant that is upstream, within the process, of the source of the signal. An autonomous signal processor may be integrated with or included in a process control device and/or a big data node of the process plant.

Abstract: A regional big data node oversees or services, during real-time operations of a process plant or process control system, a respective region of a plurality of regions of the plant/system, where at least some of the regions each includes one or more process control devices that operate to control a process executed in the plant/system. The regional big data node is configured to receive and store, as big data, streamed data and learned knowledge that is generated, received, or observed by its respective region, and to perform one or more learning analyses on at least some of the stored data. As a result of the learning analyses, the regional big data node creates new learned knowledge which the regional big data node may use to modify operations in its respective region, and/or which the regional big data node may transmit to other big data nodes of the plant/system.

Abstract: A multiple-input/multiple-output control routine in the form of a model predictive control (MPC) routine operates with wireless or other sensors that provide non-periodic, intermittent or otherwise delayed process variable measurement signals at an effective rate that is slower than the MPC controller scan or execution rate. The wireless MPC routine operates normally even when the measurement scan period for the controlled process variables is significantly larger than the operational scan period of the MPC controller routine, while providing control signals that enable control of the process in a robust and acceptable manner. During operation, the MPC routine uses an internal process model to simulate one or more measured process parameter values without performing model bias correction during the scan periods at which no new process parameter measurements are transmitted to the controller.

Abstract: A process modeling technique uses a single statistical model, such as a PLS, PRC, MLR, etc. model, developed from historical data for a typical process and uses this model to perform quality prediction or fault detection for various different process states of a process. The modeling technique determines means (and possibly standard deviations) of process parameters for each of a set of product grades, throughputs, etc., compares on-line process parameter measurements to these means and uses these comparisons in a single process model to perform quality prediction or fault detection across the various states of the process. Because only the means and standard deviations of the process parameters of the process model are updated, a single process model can be used to perform quality prediction or fault detection while the process is operating in any of the defined process stages or states.

Abstract: In a batch process control system employing storage tanks without mixers, properties of the storage tank pump out feedstock may be modeled to more accurately control the quality of a process. This model may not require the measurement of input or pump out flow or assume perfect blending. Rather, the developed model may assume that feedstock input into a storage tank may remain layered with some mixing due to continuous convection, turbulence during loading, or other factors. The model may include a projection of the properties describing a storage tank layer of input material into the model. For each new load of storage tank input feedstock, model zones may be shifted and the zone from which the feedstock is drawn may be updated with the properties from the new load.

Abstract: A process is modeled by resolving the process into a plurality of process stages, including at least a first process stage and a second process stage, and developing a plurality of models, each model corresponding to a respective one of the plurality of process stages, wherein the model corresponding to each process stage is developed using data from one or more runs of that process stage and output quality data relating to the one or more runs of that process stage and wherein the model corresponding to each process stage is adapted to produce an output quality prediction associated with that process stage, and wherein the output quality prediction produced by the model of a first one of the process stages is used to develop the model of a second one of the process stages.

Abstract: A simulation system that includes interconnected simulation blocks which use process models to perform simulation activities for a process plant is integrated into a process control environment for the process plant in a manner that makes the simulation system easy to use and easily updated for on-line process simulation. The disclosed simulation system enables future predicted values as well as the current predicted values of process parameters produced by the simulation system to be made available for performance evaluation as well as to guide plant operations. Additionally, the simulation system is connected to the operating process plant to receive various on-line process plant measurements, and uses these measurements to automatically update the process models used in the simulation system, to thereby keep the simulation system coordinated with the actual operating conditions of the process plant.

Abstract: Techniques for automatically determining, without user input, one or more sources of a variation in the behavior of a target process element operating to control a process in a process plant include using a process element alignment map to determine process elements upstream of the target process element in the process; performing a data analysis on data corresponding to the upstream elements with respect to the target element to determine behavior time offsets, strengths of impact, and impact delays; and determining the source(s) based on the data analysis outputs. Techniques may include automatically defining the process element alignment map by obtaining and processing data from a plurality of diagrams or data sources of the process and/or plant. Furthermore, the techniques may be performed during plant run-time by any high-volume, high density device such as centralized or embedded big data appliances, controllers, field or I/O devices, and/or by an unsupervised application.

Abstract: A process modeling technique uses a single statistical model developed from historical data for a typical process and uses this model to perform quality prediction or fault detection for various different process states of a process. The modeling technique determines means (and possibly standard deviations) of process parameters for each of a set of product grades, throughputs, etc., compares on-line process parameter measurements to these means and uses these comparisons in a single process model to perform quality prediction or fault detection across the various states of the process. In this manner, a single process model can be used to perform quality prediction or fault detection while the process is operating in any of the defined process stages or states.

Abstract: A simulation system that includes interconnected simulation blocks which use process models to perform simulation activities for a process plant is integrated into a process control environment for the process plant in a manner that makes the simulation system easy to use and easily updated for on-line process simulation. The disclosed simulation system enables future predicted values as well as the current predicted values of process parameters produced by the simulation system to be made available for performance evaluation as well as to guide plant operations. Additionally, the simulation system is connected to the operating process plant to receive various on-line process plant measurements, and uses these measurements to automatically update the process models used in the simulation system, to thereby keep the simulation system coordinated with the actual operating conditions of the process plant.

Abstract: A method of diagnosing an adaptive process control loop includes measuring process control loop signal data, generating a plurality of process control loop parameters from the process loop signal data and evaluating a condition of the adaptive process control loop from one or more of the plurality of process control loop parameters. The process control loop data is generated as a result of a normal operation of one or more process control devices within the adaptive process control loop when the adaptive process control loop is connected on-line within a process control environment. A self-diagnostic process control loop includes a diagnostic tool adapted to receive a diagnostic index pertaining to a process control loop parameter for a plurality of components of the process control loop and for the complete process control loop. Each diagnostic index is generated from signal data by a corresponding index computation tool.

Abstract: A batch modeling and analysis system uses a simple and computationally inexpensive technique to align data collected from an on-going, currently running or on-line batch process with a batch model formed for the batch process so as to enable the reliable determination of the current operational state of the on-line batch process with respect to the batch model. This data alignment technique enables further statistical processing techniques, such as projection to latent sources (PLS) and principle component analysis (PCA) techniques, to be applied to the on-line batch data to perform analyses on the quality of the currently running batch. These analyses, in turn, provide useful information to a user, such as a batch operator, that enables the user to determine the quality of the batch at the present time, based on the batch model, and the likelihood that the desired batch output quality metrics will be reached at the end of the batch run.

Abstract: A distributed big data device in a process plant includes an embedded big data appliance configured to locally stream and store, as big data, data that is generated, received, or observed by the device, and to perform one or more learning analyses on at least a portion of the stored data. The embedded big data appliance generates or creates learned knowledge based on a result of the learning analysis, which the device may use to modify its operation to control a process in real-time in the process plant, and/or which the device may transmit to other devices in the process plant. The distributed big data device may be a field device, a controller, an input/output device, or other process plant device, and may utilize learned knowledge created by other devices when performing its learning analysis.

Abstract: A control technique that enables the use of received process variable values in a Kalman filter based control scheme without the need to change the control algorithm includes a controller, such as a PID controller, and a Kalman filter, coupled to receive feedback in the form of, for example, process variable measurement signals from a process. The Kalman filter is configured to produce an estimate of the process variable value from slow or intermittent process feedback signals while providing a new process variable estimate to the controller during each of the controller execution cycles to enable the controller to produce a control signal used to control the process. The Kalman filter is also configured to compensate the process variable estimate for process noise with non-zero mean value that may be present in the process. The Kalman filter may apply this compensation to both continuously and intermittently received process variable values.

Abstract: A control technique that enables the use of slow or intermittently received process variable values in a predictor based control scheme without the need to change the control algorithm includes a controller, such as a PID controller, and a predictor, such as a model based predictor, coupled to receive intermittent feedback in the form of, for example, process variable measurement signals from a process. The predictor, which may be an observer like a Kalman filter, or which may be a Smith predictor, is configured to produce an estimate of the process variable value from the intermittent or slow process feedback signals while providing a new process variable estimate to the controller during each of the controller execution cycles to enable the controller to produce a control signal used to control the process.

Abstract: A method of controlling and managing a process control system having a plurality of control loops includes implementing a plurality of control routines to control operation of the plurality of control loops, respectively, wherein the control routines may include at least one non-adaptive control routine. The method then collects operating condition data in connection with the operation of each control loop, and identifies a respective process model for each control loop from the respective operating condition data collected for each control loop. The identification of the respective process models may be automatic as a result of a detected process change or may be on-demand as a result of an injected parameter change. The process models are then analyzed to measure or determine the operation of the process control loops.

Abstract: A method of controlling a process control system having a plurality of control loops includes collecting operating condition data in connection with the operation of each control loop of the plurality of control loops, identifying a respective process model for each control loop of the plurality of control loops from the respective operating condition data collected for each control loop of the plurality of control loops and developing a process model history by storing data indicative of the multiple identified process models for each control loop of the plurality of control loops.

Abstract: Example methods and apparatus to predict process quality in a process control system are disclosed. A disclosed example method includes receiving process control information relating to a process at a first time including a first value associated with a first measured variable and a second value associated with a second measured variable, determining if a variation based on the received process control information associated with the process exceeds a threshold, if the variation exceeds the threshold, calculating a first contribution value based on a contribution of the first measured variable to the variation and a second contribution value based on a contribution of the second measured variable to the variation, determining at least one corrective action based on the first contribution value, the second contribution value, the first value, or the second value, and calculating a predicted process quality based on the at least one corrective action at a time after the first time.

Abstract: A method of diagnosing an adaptive process control loop includes measuring process control loop signal data, generating a plurality of process control loop parameters from the process loop signal data and evaluating a condition of the adaptive process control loop from one or more of the plurality of process control loop parameters. The process control loop data is generated as a result of a normal operation of one or more process control devices within the adaptive process control loop when the adaptive process control loop is connected on-line within a process control environment. A self-diagnostic process control loop includes a diagnostic tool adapted to receive a diagnostic index pertaining to a process control loop parameter for each component of the process control loop and for the complete process control loop. Each diagnostic index is generated from signal data by a corresponding index computation tool.