System for determining operating points in a chemical plant
Patent Information
- Application Number
- JP2023527039
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2020-11-06
- Filing Date
- 2021-11-08
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2041-11-08
AI Technical Summary
The chemical manufacturing industry faces challenges in making vast amounts of signal time series data from numerous measurement points in chemical plants contextually interpretable, which is crucial for effective monitoring and control, as manual interpretation is inefficient and limited to small amounts of data.
A computer-implemented method and system that determines operating states of chemical plants by analyzing signal time series data through scoring based on criteria like volatility, activity, anomaly, and constancy, clustering these scores, and attributing operating states to provide contextually interpretable data for monitoring and control.
Enhances the reliability and efficiency of determining and monitoring chemical plant operating states, reducing computational burden and improving safety by providing contextually interpretable data for stable and unstable conditions, enabling early warning systems and reducing hazardous failures.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] Field of the Invention The present disclosure relates to a system, a computer-implemented method, and a computer program element for determining operating conditions in a chemical plant. The present disclosure further relates to a system, a computer-implemented method, and a computer program element for controlling and / or monitoring a chemical plant. The present disclosure further relates to a method, a system, and a computer program element for annotating a signal time series in a chemical plant.
Background Art
[0002] Background In the chemical manufacturing industry, a vast number of measurement points are dispersed throughout the plant. Many of these measurement points generate signal time series data. These signal time series data often contain useful information about the operating state of the plant. Therefore, these signal time series are often stored as historical data. In principle, this enables retroactive analysis of the signal time series data. This retroactive analysis of the signal time series data is very long and tedious. Since data without context conveys little information, the signal time series data needs to be made interpretable in context for further evaluation. This can be done manually. However, when dealing with a vast amount of data, typically only a small amount can be made interpretable in context manually. Therefore, in the chemical industry, it is necessary to make signal time series data interpretable in context.
Summary of the Invention
Means for Solving the Problems
[0003] Description In a first aspect, a computer-implemented method for determining the operating state of a chemical plant, particularly the operating state within a process group, the method comprising: The method comprises: - A step of providing signal time-series data related to a first measurement point within a chemical plant, - A step of determining each signal time series score related to the degree of change over time of the signal time series data, based on the signal time series data related to a first measurement point in the chemical plant, - A step of determining the operating state based on each signal time series score related to the degree of change over time of the signal time series data, - A step of providing data related to the operating status of a chemical plant, suitable for monitoring and / or controlling the chemical plant, Methods including the following are disclosed.
[0004] In another embodiment, a system for determining the operating state of a chemical plant is disclosed, and the system is - A time-series acquisition unit configured to receive signal time-series data related to a first measurement point within a chemical plant, - An on / off score unit configured to determine each signal time series score related to the degree of change over time of the signal time series data, based on signal time series data associated with a first measurement point within the chemical plant, - An operating state determination unit configured to determine the operating state based on each signal time series score related to the degree of change over time of the signal time series data, and an output unit configured to provide data related to the operating state of a chemical plant, suitable for monitoring and / or controlling the chemical plant. It is equipped with.
[0005] In another embodiment, a computer program product for determining the operating state of a chemical plant is disclosed, which includes instructions configured to perform steps of a method for determining the operating state of a chemical plant when the computer program product is executed on a computing device in a computing environment.
[0006] In another embodiment, a non-temporary computer-readable storage medium is disclosed, which includes instructions causing a computer to perform steps of a method for determining the operating state of a chemical plant when executed by a computer.
[0007] In another embodiment, a method for monitoring and / or controlling a chemical plant is proposed, the method comprising the steps of receiving data relating to the operating state of a chemical plant, which is suitable for monitoring and / or controlling the chemical plant, and in particular, providing a control signal based on the received data when the received data indicates an unstable operating state.
[0008] In another embodiment, a system for monitoring and / or controlling a chemical plant is disclosed, the system including an input interface for receiving data relating to the operating state of the chemical plant, and an output interface for providing control signals based on the received data, in particular when the received data indicates an unstable operating state.
[0009] In another embodiment, a computer program product for monitoring and / or controlling a chemical plant is disclosed, the computer program product includes instructions configured to perform steps of a method for monitoring and / or controlling a chemical plant when executed on a computing device in a computing environment.
[0010] In another embodiment, a non-temporary computer-readable storage medium is disclosed, which, when executed by a computer, includes instructions causing the computer to perform steps of a method for monitoring and / or controlling a chemical plant.
[0011] In another embodiment, a computer implementation method for annotating signal time-series data relating to a first measurement point in a chemical plant is disclosed, comprising the steps of: receiving data relating to the operating state of a chemical plant, which is generated according to a method for determining the operating state of the chemical plant and is suitable for monitoring and / or controlling the chemical plant; and storing the data relating to the operating state of the plant together with signal time-series data.
[0012] In another embodiment, a system for annotating a signal time series is disclosed, the system including a receiving unit configured to receive signal time series data relating to a first measurement point in a chemical plant and data relating to the operating state of the chemical plant, suitable for monitoring and / or controlling the chemical plant, and a storage unit configured to store the signal time series data relating to the first measurement point in the chemical plant together with data relating to the operating state of the chemical plant, suitable for monitoring and / or controlling the chemical plant.
[0013] In another embodiment, a computer program product for annotating signal time series is disclosed, the computer program product includes instructions configured to perform the step of annotating signal time series data when executed on a computing device in a computing environment.
[0014] In another embodiment, a non-temporary computer-readable storage medium is disclosed, which includes instructions causing a computer to perform steps of a method for annotating signal time-series data when executed by a computer.
[0015] In one embodiment, a computer implementation method for making a signal time series in a chemical manufacturing plant interpretable in context, Using a processing device, - Optionally, the step of receiving a signal time series for at least one measurement point in the chemical manufacturing plant, - A step of providing plant measurement point information for each minimum single measurement point, - A step of determining at least one signal time series score for each signal time series at at least one measurement point based on the probabilistic properties of the signal time series, - A step of generating an annotated signal time series, 〇Time series of signals from at least one measurement point within a chemical manufacturing plant, ○ For each signal time series at least at one measurement point, the determined score for each signal time series, ○ Plant measurement point information for at least one measurement point, The steps include generating an annotated signal time series, 〇 - A step of clustering the annotated signal time series into clusters based on at least one determined signal time series score and each cluster classifier, - A step of assigning the operating state to a signal time series based on clustering, - A step of providing a time series of signals that can be interpreted by context, One signal time series for each of one or more measurement points, The minimum time-series score of each signal determined for each minimum measurement point, 〇Information on each plant measurement point for each minimum measurement point, 〇Assigned operating state, A step of providing a signal time series that is interpretable by the context including, A method including the following is disclosed.
[0016] Any disclosures and embodiments described herein relate to the methods, systems, treatment devices, and computer program products outlined above, and vice versa. Conveniently, any benefits derived from any one embodiment and example are equally applicable to all other embodiments and examples, and vice versa.
[0017] As used herein, "determining" includes "starting to determine or causing to determine", "generating" includes "starting to generate or causing to generate", and "providing" includes "starting to determine, generate, select, transmit, or receive, or causing to determine, generate, select, transmit, or receive". "Starting to perform or causing to perform an action" includes any processing signal that triggers a computing device to perform each action.
[0018] In the context of the present disclosure, a "chemical plant" may refer to any manufacturing facility based on a chemical process, for example, converting raw materials into products using a chemical process. In contrast to discrete manufacturing, chemical manufacturing is based on a continuous process or a batch process. Thus, monitoring and / or controlling a chemical plant is time-dependent and thus based on a large-scale signal time series dataset. A chemical plant may include more than 1,000 measurement points that create measurement data points every 2 seconds. At such a scale, multiple terabytes of data are generated that are handled by a system for controlling and / or monitoring the chemical plant. A small-scale chemical plant may include 2,000 measurement points that create data points every 1 to 10 seconds. By way of comparison, a large-scale chemical plant may include 20,000, for example, 10,000 to 30,000 measurement points that create data points every 1 to 10 seconds. Making such data interpretable in context results in handling from hundreds of gigabytes to several terabytes.
[0019] A chemical plant can manufacture products by one or more chemical processes that convert raw materials into products through one or more intermediate products. Preferably, the chemical plant provides an encapsulated facility that manufactures products that can be used as raw materials for the next step in the value chain. The chemical plant may be a large-scale plant such as an oil and gas facility, a gas purification plant, a carbon dioxide recovery facility, a liquefied natural gas (LNG) plant, an oil refinery, a petrochemical facility, or a chemical facility. For example, a chemical plant upstream of petrochemical process manufacturing includes a steam cracker that begins with naphtha being processed into ethylene and propylene. These upstream products can then be further provided to a chemical plant to derive downstream products such as polyethylene or polypropylene. Those downstream products can be further supplied as raw materials for a chemical plant that derives further downstream products. A chemical plant can be used to manufacture discrete products. In one example, one chemical plant can be used to manufacture a precursor for a polyurethane foam. Such a precursor can be provided to a second chemical plant for the manufacture of discrete products such as an insulating plate that includes the polyurethane foam.
[0020] Value chain manufacturing to a final product through various intermediate products may be dispersed in various locations or integrated in a fairbunt site or a chemical park. Such a fairbunt site or chemical park includes a network of interconnected chemical plants where products manufactured in one plant can be supplied as raw materials for another plant.
[0021] A chemical plant may include multiple assets, some of which may include heat exchangers, chemical reactors, pumps, pipes, distillation columns, or absorption columns. In a chemical plant, some assets may be critical. Critical assets are those that, when interrupted, have a significant impact on plant operation. This can lead to the manufacturing process being jeopardized. This can result in a decline in product quality or even a complete shutdown. In the worst-case scenario, fire, explosion, or release of toxic gases may be the consequence of such an interruption. Therefore, such critical assets may need to be monitored and / or controlled more rigorously than other assets, depending on the chemical processes and chemicals involved. Multiple measurement points are needed to monitor and / or control chemical processes and assets. Such measurement points may include actors and sensors. Such actors or sensors may provide plant-specific data. Plant-specific data may include, for example, process or asset-specific data relating to the operating status of individual assets, individual actors, chemical components, or the operating status of chemical processes. In particular, process or asset-specific data may include one or more of the following data categories: - Process operation data such as the components of raw materials or intermediate products - Process monitoring data such as flow rate, material temperature, and pressure - Asset operation data such as current and voltage, and - Asset monitoring data such as asset temperature, asset pressure, and vibration.
[0022] A chemical plant may include one or more process groups. A process group may be, for example, a group of unit operations within the plant that perform macroscopic steps in a process. A process group may include unit operations. A process group within a plant may be a section within the chemical plant that shares the same operating state. In particular, a process group may include a process line.
[0023] As used herein, the "aggregation" of measurement points can be understood as a grouping of measurement points according to their relationships. This may be a technical relationship. A technical relationship may include information, process groups to which the measurement points may belong.
[0024] Plant-specific signal time-series data from at least one measurement point within a chemical manufacturing plant may be received from a storage medium in a processing device. The storage medium may be, for example, volatile memory, a hard disk, a cloud service, or a database. The plant-specific signal time-series data may include measured values from at least one of at least two measurement points. The measurement points may have certain types, depending on whether they are sensors or actors. Measurement points may be distinguished by type. Measurement points may also be distinguished by type. The type of sensor may refer to the type of signal time-series data provided, for example, temperature measurements.
[0025] The operating state refers to the operating state of the plant and may include the operating states of one or more process groups within the plant. These may be states that the plant can adopt and / or states that a process group can adopt. Several notable and appropriate embodiments of operating states are described below.
[0026] An appropriate operating state may be an ON state. An ON state may refer to an operating state in which a chemical plant is, for example, performing manufacturing. At a more detailed level, an ON state may refer to an operating state of a running process group. An ON operating state may refer to a manufacturing process.
[0027] A suitable operating state may be a stable state. A stable state may also be an ON state where the plant is operating under stable conditions. At a more detailed level, a stable state may refer to a process group operating under stable conditions. Stability conditions may be conditions under which the process operates within the scope of predefined specifications.
[0028] An appropriate operating state may be the off state. The off state may refer to a state where the plant is not operating. More specifically, the off state may refer to a state where a process group is not operating. The off operating state may refer to an operating state where there is no manufacturing process. This may be due to maintenance.
[0029] A suitable operating state may be a startup state, which may be a state in which the plant is running and already ON, but not stable. More specifically, a process group is running to execute processes and is already ON, but not stable.
[0030] This could be a situation where manufacturing is already underway, but the product's performance and quality do not yet meet specifications, potentially leading to waste.
[0031] An appropriate operating state may be a stopped state, which may be a state in which the plant is stopped but still in the ON state. More specifically, a stopped state may refer to a state in which the process group is stopped but still in the ON state. This may be a state in which manufacturing is still in progress, but the performance quality of the product is not yet according to specifications, leading to waste.
[0032] An appropriate operating state may be an event state, which may be a state in which the plant is ON but not stable and therefore receiving an event. During an event, the plant may be in an unstable state. This may be due to errors / failures during the process or during the startup or shutdown state.
[0033] Different event types may exist, such as startup, shutdown, and failure events. The stopped state and the started state may be subgroups of event states. The started state may follow the off state, and the stopped state may precede the off state. A failure may also be a subgroup of event states, with the stable state preceding and following it.
[0034] The processing device may be a processor, a network of processors, a computer network including one or more computers containing one or more processors, or a dedicated processor for each step. Some or all of the steps may be performed on one processing device or on a group of processing devices.
[0035] The objective of the present invention is to provide a reliable and robust determination of the operating status of a chemical plant.
[0036] By using a score, it becomes possible to determine the operating state independently of the measured values of the measurement point data. For illustrative purposes, the following embodiment is given. A threshold for a temperature sensor may be used to determine whether it is on or off. In the off state, the temperature sensor may measure the ambient temperature. However, in hot summers, the ambient temperature may exceed the threshold used to distinguish between the off and on states. In that case, an incorrect operating state will be determined. This can be avoided by determining an on / off score independent of the measured values.
[0037] An abstraction layer is provided, which involves determining at least one signal time series score for each of the specific signal time series data. This is particularly useful when signal time series data is provided from multiple measurement points. The abstraction layer has the advantage that signal time series scores from different measurement points can be compared, or that mathematical operations can be performed on the signal time series scores. For example, signal time series scores from signal time series generated from one type of measurement point (e.g., temperature) can be compared with each other, or mathematical operations can be performed on them.
[0038] In one embodiment, each signal time series score related to the degree of change over time in the signal time series data may also be called an on / off score.
[0039] Determining at least one respective signal time series score for at least one signal time series at at least one measurement point can be understood as determining the respective signal time series score related to the degree of change over time of the signal time series data, based on the signal time series data associated with a first measurement point in the chemical plant.
[0040] In one embodiment, determining each signal time series score related to a criterion for the degree of change over time of the signal time series data, based on signal time series data associated with a first measurement point in a chemical plant, may include determining a volatility score. The volatility score may be based on volatility, which describes the degree of change over time of the signal time series. Determining volatility for a signal time series is described, for example, in Somarajan, Siddarth & Shankar, Monica & Sharma, Tanmay & Ramasamy, Jeyanthi. (2019). Modelling and Analysis of Volatility in Signal time series Data: Proceedings of ICSCSP 2018, Volume 2.10.1007 / 978-981-13-3393-4_62. Volatility has the advantage of providing a dimensionless and scale-invariant score.
[0041] In one embodiment, determining each signal time series score related to a criterion for the degree of change over time of the signal time series data, based on signal time series data associated with a first measurement point within a chemical plant, may include determining an activity score.
[0042] In one embodiment, at least one of each signal time series scores may be an activity. Activity is associated with the likelihood of having an eigenvalue within the signal time series over a certain period. The activity score is based on the concept of a near-zero variance predictor.
[0043] Volatility and activity are interpretable scores, which allows for the verification of those scores.
[0044] In one embodiment, the method further includes the step of determining a volatility score and an activity score, and determining the operating state based on the respective signal time series scores related to a criterion for the degree of change over time of the signal time series data includes determining based on the volatility score and the activity score. Since two independent scores are considered, this improves the reliability of the determination.
[0045] In one embodiment, determining at least one respective signal time series score for each signal time series at at least one measurement point may include determining a signal time series score that explains an outlier in the signal time series.
[0046] In one embodiment, determining at least one respective signal time series score for each signal time series at least at one measurement point may include determining an activity score.
[0047] In one embodiment, the method is - To provide time-series signal data related to a second measurement point within a chemical plant, - Based on the signal time series data associated with the second measurement point, determine the on / off score related to the criterion for the degree of change over time of the second signal time series data, It further includes, - Determining the operating state includes determining the operating state based on the on / off scores of the first time series data and the second time series data.
[0048] Determining the operating state based on time-series data provided from one or more measurement points improves the reliability of the operating state determination. This is important when plant monitoring and / or control require high reliability. In particular, in chemical plants, failures in process control can lead to hazardous conditions. While the advantages associated with two measurement points have been disclosed, it is clear that more than two measurement points may be included.
[0049] In one embodiment, measurement point information is provided to a first measurement point and a second measurement point, and the measurement point information includes technical markers.
[0050] In one embodiment, determining the operating state is further based in particular on technical indicators. The step of determining the operating state includes determining the operating state based on on / off scores of first signal time series data and second time series data only if the technical indicators indicate that the measurement point is technically related and in particular belongs to one process group.
[0051] Technically relevant signal time-series data may share the same operating state, such as ON / OFF. This reduces ambiguity when determining the operating state of a chemical plant, particularly the operating state of a process group. In one embodiment, the operating states of various technically relevant measurement points may be averaged, and the operating state may be determined based on the average. In another embodiment, a majority vote among measurement points may be used additionally to determine the operating state.
[0052] In one embodiment, the method includes the step of aggregating sensors based on their technical relationships.
[0053] In one embodiment, determining the operating state is based on the on / off score and the respective cluster classifiers.
[0054] Each cluster identifier allows for a clear determination of whether each signal time series value belongs to the ON state or the OFF state.
[0055] Each cluster classifier used to determine the operating state based on the on / off score and its respective classifier is an ON / OFF cluster classifier.
[0056] In one embodiment, signal time series data may be clustered based on signal time series scores fluctuating around two distinct signal time series score values. These two distinct signal time series score values may then be attributed to the on or off operating state of their respective process groups. This allows for easy labeling of plant-specific signal time series data based on the on and off operating states of process groups. Alternatively, the cluster classifier may be a threshold, and the signal time series data at measurement points may be clustered based on whether they are above or below the threshold.
[0057] In one embodiment, determining the operating state includes determining whether the state is ON or OFF.
[0058] In one embodiment, the method may further include the step of segmenting time-series data based on the determined operating state.
[0059] In one embodiment, the signal time series data is segmented based on the ON state and / or OFF state.
[0060] This causes the on and / or off states to be temporally ordered. This provides a reduced set of information, and the binary state representation requires less storage for the sake of saving.
[0061] In one embodiment, the method further includes selecting signal time series data related to the ON state and determining a stability score for the selected signal time series.
[0062] This reduces the amount of data that needs to be used to assess stability scores. This, in turn, reduces the required storage capacity, memory capacity, and computing power. Furthermore, it makes it possible to determine stable operating states related to plant operation, particularly process group operation. Measurement points from process groups or the plant in the OFF state will be considered as stable operating conditions. However, this only applies to stable OFF states. This can be particularly problematic when measurement points related to the OFF state are grouped together with measurement points related to the ON state.
[0063] In one embodiment, the stability score may be at least one time-series score for each signal.
[0064] The stability score may be based on criteria related to the degree to which the signal time series changes over time. In other words, the signal values of the signal time series data are independent of the time over which the signal series is observed.
[0065] In one embodiment, the method further includes determining stable operation based on the stability score and the respective cluster classifiers.
[0066] Each cluster classifier used to determine the operating state based on the stability score is called a stability cluster classifier.
[0067] In one embodiment, signal time-series data may be clustered based on stability scores fluctuating around two distinct signal time-series score values. These two distinct signal time-series score values may then be assigned to the stable or unstable operating state of the respective process group. This allows for easy and reliable identification of stable and unstable operating states of a plant or process group. Alternatively, the stability cluster classifier may be a threshold, and the signal time-series data at measurement points may be clustered according to whether they are above or below the threshold.
[0068] This enables easy and reliable determination of stable operating conditions and / or events.
[0069] Discrimination of plant-specific signal time-series data based on the stable and unstable operating states of process groups.
[0070] By determining the stability score, it is possible to determine the stable operating state of the plant. This represents the desired operating state of the plant. Stable operating conditions are essential for consistent product quality and are also related to process conditions that reduce energy requirements. Unstable operating conditions may indicate events.
[0071] In one embodiment, the stability score may be associated with the anomaly score. Anomaly scores explain outliers relative to the standard signal. Various methods for determining anomaly scores are known in the art (ARIMA-GEPBox, GM Jenkins: Signal time series analysis: Forecasting and control. Holden-Day, San Francisco 1970; STL decomposition, Cleveland, RBet.al. 1990. STL: A seasonal trend decomposition procedure based on low-level data, Journal of Official Statistics 6(1):3-73). Another method suitable for determining anomaly scores is the hybrid seasonal extreme studentized deviate test (HSESD: Automatic Anomaly Detection in the CloudVia Statistical Learning, Jordan Hochenbaum et al., arXiv:1704.07706v1, Submitted on 24 Apr 2017).
[0072] Anomaly scores have the advantage of providing dimensionless and scale-invariant scores. Anomaly scores provide values between lower and higher values, which can then be normalized to any range between 0 and 1. Normalization to a range between 0 and 1 improves comparability.
[0073] In some embodiments, the stability score may be a stationarity score. Stationarity describes the fact that the statistical properties of a signal time series (or rather, the process that generates it) do not change over time. Various methods for determining the stationarity of a signal time series are known in the art (e.g., Dickey, DA und WA Fuller: Distribution of the Estimators for Autoregressive Signal time series with a Unit Root, Journal of the American Statistical Association, 1979, 74, pp. 427-431; Kwiatkowski, D.; Phillips, PCB; Schmidt, P.; Shin, Y. (1992) "Testing the null hypothesis of stationarity against the alternative of a unit root". Journal of Econometrics. 54 (1-3): 159-178). Stationarity scores have the advantage of providing dimensionless and scale-invariant scores.
[0074] The stationarity score provides a value between a lower and a higher value. This can then be normalized to any range of values between 0 and 1. Normalization to a range between 0 and 1 improves comparability.
[0075] In one embodiment, the method may further include providing data relating to the operating state of a chemical plant, suitable for monitoring and / or controlling the chemical plant, and further including data relating to the stable state of the plant, including data relating to the ON state.
[0076] As mentioned above, stable operating conditions are desired, and it is important to determine whether the plant is in a stable operating state for monitoring and control purposes. When the plant is operating stably, monitoring may be sufficient, and if unstable operating conditions occur, the plant may begin to shut down to prevent dangerous conditions from occurring. In this way, the early warning system is activated, and the safety of the chemical plant is improved.
[0077] In one embodiment, the method may include the step of determining an event based on a stability score. An event may be determined when the stability score indicates unstable behavior.
[0078] In one embodiment, determining the operating state may include determining the event type corresponding to the previous operating state. This makes it possible to distinguish between failure events and normal operating states such as startup and shutdown. Failure events are potentially dangerous.
[0079] In one embodiment, the stable operating point of a plant may be determined based on a stable operating state. The stable operating point may be defined by a time series of signals from a measurement point. Determining the stable operating point is crucial for the efficient operation of a chemical plant and may be closely related to product quality.
[0080] In one embodiment, the method may further include providing signal time-series data from a measurement point and receiving, in particular, data related to the operating state of a chemical plant, suitable for monitoring and / or controlling the chemical plant, provided by and received by a client device.
[0081] In one embodiment, signal time series scores from signal time series generated from different types of measurement points (e.g., temperature, pressure) can be compared with each other, or mathematical operations can be performed.
[0082] At least one signal time series score can be scale-invariant. Scale-invariant signal time series scores offer advantages when signal time series scores from one type of measurement point (e.g., temperature) are compared to each other, or when mathematical operations are performed.
[0083] At least one signal time series score can be dimensionless. Dimensionless signal time series scores offer advantages when signal time series scores from different types of measurement points (e.g., temperature, pressure) are compared to each other, or when mathematical operations are performed.
[0084] Each signal time series score can, in principle, be a signal time series. It is beneficial to define bins to determine each signal time-series score from specific signal time-series data from at least one measurement point within a chemical manufacturing plant. A bin may contain a sequence of signal time-series data. The sequence may be selected so that a sufficient amount of sensor data is available to determine each signal time-series score. A sufficient amount of data may include at least 25 data points.
[0085] Additionally, the sequence of signal time series data may be selected according to the plant's time constant. In particular, the sequence may be shorter than the dynamic time constant of the monitored process. A suitable time constant may be the time of the plant's startup and / or shutdown procedures. A sequence shorter than the process's time constant is beneficial. It is ensured that the information in the signal time series data is preserved and that different operating conditions within the plant are not mixed. The bins may then be moved along the signal time series as a sliding window. At least one signal time series score may vary as a function of time.
[0086] In one embodiment, determining at least one respective signal time series score for each signal time series at at least one measurement point includes determining a signal time series score that describes the degree to which the respective statistical characteristics of each signal time series at at least one measurement point over time have not changed over time.
[0087] The step of determining at least one signal time series score for specific signal time series data may include determining two or more different signal time series scores.
[0088] Determining the scores of two or more signal time series has the advantage that each of the two or more signal time series may have different sensitivities to different operating states of the process group.
[0089] When each change in the operating state is reflected in the change in the signal time series score, the signal time series score is considered to be highly sensitive to certain changes in the plant's operating state.
[0090] In one embodiment, a set of rules may be provided, which associates clusters of signal time-series scores with operating states. In one embodiment, the set of rules may include associating combinations of signal time-series scores for each time series with operating states.
[0091] For example, volatility and activity may be highly sensitive to the on / off operating state of a plant, but less sensitive to changes throughout the plant's overall operating state. The on operating state of a plant may refer to a plant or a process group of a plant that is manufacturing. The off operating state of a plant may refer to a plant or a process group of a plant that is not manufacturing. Volatility and abnormalities are, in this case, very suitable for identifying the on / off operating state of a plant or a process group of a plant.
[0092] In one embodiment, the rule set for volatile clusters may then define that a high volatility score relates to, for example, the ON state of a process group, and a low volatility relates to, for example, the OFF state of a process group.
[0093] In one embodiment, the rule set for an activity cluster may then define that a high activity score is related to the process group's operational state being on, and low activity is related to the process group's operational state being off.
[0094] In another embodiment, the abnormality score and the steady-state score may be highly sensitive to the on / off operating state of the plant, but may also be highly sensitive to changes during the plant's on-process phase. These scores, then, may be good indicators of changes during the plant's on-process phase, but not good indicators of the on / off operating state of the plant's process groups.
[0095] In one embodiment, a rule set for instability score clusters may define that high instability scores are associated with unstable operating states, and low instability scores are associated with unstable operating states. In one embodiment, high and low are defined with respect to threshold lines.
[0096] In one embodiment, the rule set for anomaly score clusters may define that high anomaly scores are associated with unstable operating states, and low anomaly scores are associated with unstable operating states. In one embodiment, high and low are defined with respect to threshold lines.
[0097] Additional information regarding the signal time-series data is provided by providing plant measurement point information for at least one measurement point.
[0098] In one embodiment, plant measurement point information may refer to the type of measurement point. The type of measurement point distinguishes which device provides the data, for example, whether it is a sensor or an actor.
[0099] In one embodiment, plant measurement point information may refer to the type of measurement point. Sensor type may refer to the type of signal time-series data provided, for example, temperature measurements may be assigned to the same type.
[0100] In one embodiment, plant measurement point information may include technical relationship indicators that show technical relationships to process groups within the plant. These technical relationship indicators may indicate that different types or kinds of measurement points are technically related to each other, such as belonging to the same process group. This is beneficial because, for example, it allows for the aggregation of sensors based on their technical relationships within a process group. This allows the plant's operating state to be given a resolution at the process group level. This adds additional information when technically related measurement points, for example, from one process group, share the same operating state. In an embodiment, This makes it possible to determine the operating state of a plant based on sensors belonging to the same process group. This, in turn, enhances the reliability of determining the operating state. Providing technical relation indicators for the measuring sensors is beneficial because it provides additional context to the signal time series. In a plant with multiple identical production lines, the technical relation indicators may refer to a specific production line. Identical production lines may each be in different operating states. In one embodiment, the technical relation indicators are based on the location of the measurement point.
[0101] Providing the type of measuring sensor is beneficial because it provides additional context to the signal time series. Temperature changes may be gradual, for example, while pressure changes may be rapid in certain scenarios.
[0102] In one embodiment, plant measurement point information refers to at least one of the following: the type of measurement point, the type of measurement point, and a list of technical indicators.
[0103] Plant measurement point information may be provided from process and instrumentation diagrams (P&IDs). These diagrams are often used to describe plants in the chemical industry. Alternatively, plant measurement point information may be provided from a graphical representation of the plant.
[0104] The step of clustering annotated signal time series into clusters based on at least one determined signal time series score and cluster classifier provides additional information about the operational state of process groups within a particular section of signal time series data. Clustering can be understood as the task of grouping sets of objects such that objects within the same group (called a cluster) are (in a sense) more similar to one another than objects in other groups (clusters). For the purposes of this application, an object is a signal time series representing its respective signal time series score.
[0105] The step of clustering annotated signal time series into clusters based on at least one determined signal time series score for each provides additional information in determining which signal time series data have similar signal time series scores.
[0106] In one embodiment, the step of clustering annotated signal time series data based on the minimum determined signal time series score may include clustering based on each cluster classifier.
[0107] In one embodiment, cluster classifiers may be assigned a priority. In another embodiment, the cluster classifier may be modified based on the respective signal time series scores.
[0108] The cluster classifier may include thresholds, and clusters are formed depending on whether the signal time series score value is above or below the threshold.
[0109] A cluster classifier may include two distinct clusters, and clustering is performed depending on the likelihood that the signal time series score value belongs to one of the two distinct clusters. An example of a suitable method for clustering may be K-means clustering.
[0110] Cluster classifiers can be modified through unsupervised learning. In some embodiments, clustering can be performed by unsupervised learning. Various mechanisms for clustering and unsupervised learning on signal time series data are described in the review article (Pattern Recognition 38 (2005) 1857-1874, Clustering of signal time series data - a survey, Warren Liao).
[0111] In one embodiment, clustering may be performed by defining thresholds, and clusters are formed depending on whether the signal time-series score value is above or below the threshold.
[0112] Unsupervised learning is particularly useful when signal time series scores fluctuate around two distinct signal time series score values. Because it eliminates the need to manually label data against the training dataset, unsupervised learning is also useful for large datasets.
[0113] In one embodiment, signal time series data can be clustered based on signal time series scores fluctuating around two distinct signal time series score values. These two distinct signal time series score values can then be attributed to the on or off operating state of their respective process groups. This allows for easy labeling of plant-specific signal time series data based on the on and off operating states of process groups.
[0114] In one embodiment, signal time-series data can be clustered based on volatility scores. The volatility scores fluctuate around two distinct volatility score values. These two distinct volatility values can then be attributed to on or off operating states. This allows for easy labeling of plant-specific signal time-series data by on and off operating states.
[0115] Unsupervised learning is particularly useful when signal time series scores represent a small number of distinct score values. Volatile scores generally exhibit binary behavior of score values.
[0116] In one embodiment, the step of clustering annotated signal time series data based on the minimum determined signal time series score may include clustering based on volatility scores.
[0117] In one embodiment, clustering of volatility scores can be performed by unsupervised learning.
[0118] In one embodiment, one of these volatility score clusters may belong to the ON state, while the other volatility score clusters may belong to the OFF state.
[0119] Unsupervised learning is particularly useful when signal time series scores represent a small number of distinct score values. Activity scores generally exhibit binary behavior of score values.
[0120] In one embodiment, signal time-series data can be clustered based on activity scores. The activity scores fluctuate around two distinct volatility score values. These two distinct volatility values can then be attributed to the on or off operating states of their respective process groups. This allows for easy labeling of plant-specific signal time-series data by on and off operating states.
[0121] In one embodiment, activity score clustering may be performed by unsupervised learning.
[0122] In one embodiment, one of these activity score clusters may belong to the ON state, while the other activity score clusters may belong to the OFF state.
[0123] In one embodiment, clustering annotated signal time series based on at least one determined signal time series score includes clustering based on volatility scores and clustering based on activity scores. A method for clustering signal time series in two-dimensional multivariate clustering may be used. A suitable model is disclosed by Zhou, Pei-Yuan et al. (2014); A Model-Based Multivariate Signal time series Clustering Algorithm.
[0124] In one embodiment, clustering annotated signal time series into clusters based on at least one determined signal time series score for each signal time series may be based on signal time series scores that explain outliers in the signal time series.
[0125] In one embodiment, the step of clustering annotated signal time series data based on the minimum determined signal time series score may include clustering based on anomaly scores, which vary between a lower and upper limit.
[0126] In one embodiment, the step of clustering annotated signal time series data based on anomaly scores may be performed by defining an anomaly threshold, and clusters are formed depending on whether the anomaly score value is above or below the anomaly threshold.
[0127] Clusters whose anomaly score exceeds the anomaly threshold may be attributed to an event state. Clusters whose anomaly score falls below the anomaly threshold may be assigned to a no-event state.
[0128] In one embodiment, clustering annotated signal time series into clusters based on at least one determined signal time series score for each signal time series may be based on a signal time series score that describes the degree to which the statistical characteristics of each signal time series do not change over time for at least one measurement point over time.
[0129] In one embodiment, the step of clustering annotated signal time series data based on the minimum determined signal time series score may include clustering based on a stationarity score, which varies between a lower and upper limit.
[0130] In one embodiment, the step of clustering annotated signal time series data based on stationarity scores may be performed by defining a threshold, and clusters are formed depending on whether the stationarity score value is above or below the threshold.
[0131] Clusters whose stationarity score exceeds the stationarity threshold can be attributed to a stable operating state. Clusters whose stationarity score falls below the stationarity threshold may be attributed to an unstable operating state.
[0132] When an off state precedes an event state, the cluster may be attributed to an active state. When an event state is followed by an off state, the cluster may be classified as inoperable.
[0133] From the definitions of abnormal and steady state, it is clear that detecting abnormalities is detrimental to the steady state. Once an event state is detected, a stable operating state cannot be reached.
[0134] Labels may be assigned based on clustering. In one embodiment, labels may be assigned to the on / off operational state of process groups.
[0135] In another embodiment, labels may be given indicating abnormality / no abnormality. In yet another embodiment, labels may be given indicating steady / absent.
[0136] The signal time series is interpretable here in context and can be used for further evaluation.
[0137] In one embodiment, at least one measurement point in a chemical manufacturing plant includes two or more measurement points, and the method is - Based on the measurement point information for each plant, select one or more of at least two measurement points, - Determine at least one signal time series score for each signal time series, for each of the one or more selected measurement points. It also includes.
[0138] Selecting one or more of at least two measurement points based on the information from each plant measurement point, and determining at least one signal time series score for each of the selected measurement points, has the advantage that not all signal time series from all measurement points need to be assessed. This greatly reduces processing time.
[0139] In one embodiment, selecting one or more of at least two measurement points based on the information of each plant measurement point may include selecting based on the technical reference markings of the measurement points.
[0140] Selecting signal time-series data from at least two measurement points based on plant measurement point information can yield more robust results for clustering, as only significant measurement points can be selected.
[0141] As an example, one might consider being interested only in the on / off operating state of a specific process group. This is achieved by selecting only the measurement points related to that specific process group. In this example, the selection is based on plant measurement point information labels. Clustering can then be applied only to the selected signal time series. This can be beneficial when it is known that each signal time series score is highly sensitive to a particular operating state of the plant. For example, a flow sensor might react rapidly to changes in reactant flow, and the reactant flows only when the process group's operating state is on.
[0142] In one embodiment, selecting one or more of at least two measurement points based on the information of each plant measurement point may include selecting based on the type of measurement point.
[0143] Selecting signal time series data from at least two measurement points based on the type of measurement point can yield more robust results for clustering, as it ensures that only significant measurement points are selected.
[0144] As an example, when we are interested only in signal time series data from one or more measurement points based on the type of measurement point, only measurement points related to a particular type of measurement point may be selected. In this example, the selection is based on the type of measurement point. Clustering can then be applied only to the selected signal time series.
[0145] In one embodiment, selecting one or more of at least two measurement points based on the information of each plant measurement point may include selecting based on the type of measurement point.
[0146] In one embodiment, the step of clustering annotated signal time series into clusters based on at least one determined signal time series score for each of the selected measurement points may include clustering annotated signal time series only for one signal time series for each of the selected measurement points.
[0147] This has the advantage that only selected signal time series are used for clustering. Clustering is a computationally expensive task. Therefore, it is beneficial to select signal time series data from one or more measurement points based on plant measurement point information. This can significantly reduce the amount of signal time series data that needs to be evaluated for clustering.
[0148] In one embodiment, the method may further include the step of aggregating at least one respective signal time series score for each signal time series for each of the selected one or more measurement points.
[0149] Aggregation can involve, for example, summing, averaging, or weighting.
[0150] By aggregating the signal time-series scores across selected measurement points, the signal-to-noise ratio of the signal time-series scores can be increased.
[0151] By aggregating the time-series scores of each signal across selected measurement points, the robustness of the subsequent clustering process can be improved.
[0152] In one embodiment, annotated signal time series can be stored. Storing annotated signal time series has the advantage that, when the data processing is further evaluated at a later point, the computationally intensive task of determining each of the at least one signal time series scores for each signal time series at least at one measurement point does not have to be repeated multiple times.
[0153] In one embodiment, - Determining at least one signal time series score for each signal time series for at least one measurement point includes determining the respective volatility score and / or the respective activity score. - Clustering annotated signal time series into clusters based on at least one determined signal time series score includes clustering based on their respective volatility scores and / or their respective activity scores. - The assigned operating state is the ON state.
[0154] In one embodiment, determining the signal time series score that explains an outlier in the signal time series requires attributing it to the ON state.
[0155] In one embodiment, determining at least one anomaly score for each signal time series at least at one measurement point is required to attribute it to the ON state.
[0156] In one embodiment, determining the degree to which the statistical characteristics do not change over time for each signal time series at at least one measurement point over time requires attribution to the ON state.
[0157] In one embodiment, determining at least one stationarity score for each signal time series at least at one measurement point requires at least one to be attributed to the ON state.
[0158] In one embodiment, each anomaly score is determined only for signal time series that belong to the ON state. The OFF state of the plant or process group is not considered of interest to data analysis. By determining anomaly scores only for signal time series that belong to the ON state, irrelevant data does not need to be assessed. This greatly reduces computation time.
[0159] In one embodiment, each stationarity score is determined only for signal time series that belong to the ON state.
[0160] In one embodiment, each stationarity score is determined only for signal time series that belong to the ON state. The OFF state of the plant or process group is not considered of interest to the data analysis. By determining the stationarity score only for signal time series that belong to the ON state, irrelevant data does not need to be assessed. This greatly reduces computation time.
[0161] In one embodiment, the method may further include the step of deriving statistics for each signal time series data from at least two sensors. The statistics may be, for example, the mean, median, variance, standard deviation, quantiles, quartiles, percentiles, or missing values. These statistics make it possible to determine the informational content of the signal time series data. This makes it possible to define restrictions on the signal time series data in accordance with these restrictions to exclude sequences of signal time series from further evaluation. Processing time can be greatly reduced by excluding signal time series data that does not contain useful information. For example, if the variance of a signal time series exceeds a certain threshold, this may indicate that the sensor is providing a noisy signal and therefore the signal does not need to be used for further evaluation. A malfunctioning sensor may, for example, provide a flat signal and therefore have low signal variance. As a result, sensor signals with low variance may be excluded from further evaluation.
[0162] In one embodiment, a computer program or computer program product or computer-readable non-volatile storage medium includes computer-readable instructions that perform the methods disclosed herein when loaded and executed by a processing device.
[0163] In one embodiment, a system is proposed which includes input devices and output devices, as well as a processing device configured to perform the methods disclosed herein. [Brief explanation of the drawing]
[0164] [Figure 1] This shows a block / flow diagram of a computing system for determining the operating status of a chemical plant. [Figure 2] This disclosure provides an exemplary workflow. [Figure 3] An exemplary example of the plant is shown. [Figure 4] A detailed diagram of an exemplary plant process group is shown. [Figure 5] This document shows an example of sensor data and the linking process. [Figure 6] The data flow described in this disclosure is illustrated in an example. [Figure 7] This shows how the event type can be determined based on the previous operating state. [Figure 8] A system is presented to enable contextual interpretation of signal time series in a chemical manufacturing plant. [Figure 9] This document presents an example of a method for making signal time series interpretable in context in a chemical manufacturing plant. [Figure 10] This shows the behavior of the score signal over time. [Modes for carrying out the invention]
[0165] Detailed explanation The present disclosure will be further explained below in relation to the enclosed drawings.
[0166] Figure 1 shows a block / flow diagram of a computing system 100 for determining the operating status of a chemical plant according to this disclosure.
[0167] A system for determining the operating state of a chemical plant is referred to as 102, and signal time-series data 114 related to a first measurement point in the chemical plant may be provided via a time-series acquisition unit 112. In this embodiment, signal time-series data from one or more measurement points may be provided. The signal time-series data may be provided from a database, measurement points, or a combination of both. The signal time-series data may be historical signal time-series data or current signal time-series data. Using historical signal time-series data may be useful when information about the historical operating state is not available for historical signal time-series data. Using current signal time-series data may be preferred when monitoring and / or control of the chemical plant is assumed. In this embodiment, technical relationship indicators 134 may be provided for each measurement point. The technical relationship data may include information about the technical relationships between measurement points, for example, if the measurement points belong to the same process group. The embodiment will be described in more detail with respect to the chemical plant shown in Figure 3. System 100 may include a workstation or system 102, processing device 104. The system preferably includes one or more processors 104 and memory 106 for storing applications, units, and other data. In one embodiment of the present disclosure, the system 102 may include one or more displays 108 for viewing. The displays 108 may enable a user to interact with the system 102 and its components and functions. This may be further facilitated by a user interface 110, which may include a mouse and / or keyboard.
[0168] The sensor aggregation unit 116 may be configured to aggregate signal time-series data related to measurement points within the chemical plant, which in this embodiment may be signal time-series data provided from three measurement points. In the sensor aggregation unit, the time-series data is aggregated according to technical reference indicators 134. In the on / off unit 120, each signal time-series score related to the degree of change over time of the signal time-series data is determined based on the signal time-series data related to a first measurement point within the chemical plant. This may be performed by the activity detection unit 122 when the activity score is determined by the volatility detection unit 138 when the volatility score is determined. Alternatively, both the volatility score and the activity score may be determined by their respective detection units. In the operating state unit 136, the operating state may be determined based on each signal time-series score related to the degree of change over time of the signal time-series data. In the stability unit 142, the stability score is determined when selecting signal time-series data related to the on state. The stability score may be determined by the steady-state detection unit 124 when the steady-state score is determined, or by the anomaly detection unit 126 when the anomaly score is determined. The operating state unit 136 determines the operating state based on the steady-state score. This may be stable operation or unstable operation, with unstable operation being associated with an event. The concatenation unit 140 may annotate the signal time series data by concatenating the operating states to their respective signal time series. In output 128, the annotated signal time series data may be provided to storage in a database (not shown).
[0169] The system may generate outputs in block 128, which may include data 130 and / or warnings 132 related to the operating status of the chemical plant, suitable for monitoring and / or controlling the chemical plant. Further monitoring / control signals may be provided in combination with a system for monitoring and / or controlling the plant system 102.
[0170] Figure 2 shows an exemplary workflow according to this disclosure. The workflow is shown as 200. At step 205, time-series signal data related to a first measurement point, a second measurement point 210, and a third measurement point 215 in the chemical plant are provided, and the time-series data may be from flow sensor 332, temperature sensor 334, and flow sensor 322, respectively. Measurement point information for each time-series signal is also provided by time-series acquisition unit 112. At step 220, sensor aggregation unit 116 evaluates a technical relation indicator, which indicates that the time-series signal relates to process group 330, while the time-series signal from flow sensor 322 relates to process group 320. At steps 225 and 230, the time-series signal data is aggregated accordingly. Hereinafter, as an embodiment, the workflow is sequential with sensor group I. At step 235, the on / off state is determined separately for each aggregated sensor. This is done by determining the volatility score, and then determining the on / off state based on the volatility score. The embodiment describes a flow relating to non-volatility. This refers to the step of determining each signal time series score related to a criterion for the degree of change over time of the signal time series data, based on signal time series data relating to a first measurement point in a chemical plant. In step 235, the operating state is determined based on each signal time series score related to a criterion for the degree of change over time of the signal time series data. In this embodiment, this relates to an on / off state. In step 240, an on / off state may be provided. In step 245, further evaluation is performed only for the determined on state. In steps 250 and 255, a stability score is determined. In step 250, an abnormal score is determined, and in step 255, a steady-state score is determined. In 260, it is determined whether the abnormal score indicates an abnormality. If it does not, a stable operating point may be determined in step 270. In a second branch following step 245, the steady-state score is determined in 255. If the steady-state score indicates stable operating conditions, step 270 follows, in which a stable operating point is provided.A stable operating point is provided only when both the abnormality score and the steady-state score indicate a stable operating state. If an abnormality is detected in step 260 and a non-steady state is detected in step 265, a signal time series may be provided in step 275, and the provided signal time series may include the stability score for the aggregated sensors. Additionally, or alternatively, an event may be determined in step 280. The event type may also be determined in accordance with the previous operating state. The event type may be startup, shutdown, or failure. In the case of a failure, an alarm signal may be provided for monitoring, or a stop signal may be provided for control, and the stop signal may shut down the plant or process group, respectively.
[0171] Figure 3 shows an example of a chemical plant including several process groups. In this embodiment, the plant includes different process groups 310, 320, 330, 3340, and 350. Process group 310 may be a pre-treatment group for the pre-treatment of reactants, and process groups 320, 330, and 340 may be production trains. Production trains may also be process groups in which reactions occur from reactants supplied to the product from pre-treatment group 310. The plant in this embodiment includes three identical production trains. All production trains may have individual operating states. In this embodiment, the plant further includes a process group for post-treatment 350. All production trains may communicate with the process groups for pre-treatment and post-treatment. In this exemplary embodiment, each production train may have various measurement points configured to provide signal time-series data. In this embodiment, the measurement points are sensors, and flow sensors 322, 332, 342, temperature sensors 324, 334, 344, and pressure sensors 326, 336, 346 are provided in each process train 320, 330, 340, which are indicated by the first digit.
[0172] The flow sensor 322, temperature sensor 324, and pressure sensor 326 may be technically related, for example, being located on the same process train 320. The flow sensor 332, temperature sensor 334, and pressure sensor 336 may be technically related, for example, being located on the same process train 330. The flow sensor 342, temperature sensor 344, and pressure sensor 346 may be technically related, for example, being located on the same process train 340.
[0173] Figure 4 shows a zoomed-in view into the manufacturing train 340. The manufacturing train may include measurement points, in this embodiment, which are a first temperature sensor 344 and a second temperature sensor 345, and a first flow sensor 342 and a second flow sensor 343. Plant measurement point information may be provided for each measurement point.
[0174] In the case of the first temperature sensor, the plant measurement point information is: Type: Sensor Type: Temperature Technical indicator: Process train 340 It is possible.
[0175] In the case of the second temperature sensor, the plant measurement point information is: Type: Sensor Type: Temperature Technical indicator: Process train 340 It is possible.
[0176] In the case of the first flow sensor, the plant measurement point information is: Type: Sensor Type: Mass flow rate Technical indicator: Process train 340 It is possible.
[0177] In the case of the second flow sensor, the plant measurement point information is: Type: Sensor Type: Volumetric flow rate Technical indicator: Process train 340 It is possible.
[0178] Figure 5 shows an example of sensor data and the concatenation process. 510 relates to the signal time series from sensor 1. Then volatility 520 is determined and shown, and volatility is a time series. In the clustering step 530, the time series data is assigned to off and on operating states 540. In a further determination step 550, a stationarity score 555 is determined only for the on state. In step 560, the stability of the operating state is determined based on the stationarity score. This is done by clustering as described above. Finally, the operating state 570 is assigned to the signal time series data. The bottom line from left to right shows an example of concatenated signal time series data.
[0179] Figure 6 illustrates the data flow according to this disclosure. For each aggregate, the aggregate includes signal data from one or more technically relevant measurement points. First, on and off are determined between the operating states. The signal time-series data associated with the on state is then further evaluated by determining a stability score, as described with respect to Figure 2, for example. Based on the stability score, it is determined whether the operating state is stable or whether an event has occurred. If an event has been determined, the type of event may be further determined. In this embodiment, only three event types are shown: startup, shutdown, and failure.
[0180] Figure 7 illustrates how event types can be determined based on previous operating states. An unstable operating state, the ON state, is followed by a stable ON state, and the OFF state may be related to the plant startup process.
[0181] An unstable ON state is followed by a stable ON state. The stable ON state can be an indicator of a failure event. The plant is operating in a stable state and recovers to a stable state after a failure event. The last embodiment shows an embodiment in which a shutdown process can be determined. A stable ON state is followed by an unstable ON state, followed by an OFF state. The embodiment corresponds to the schematic diagram in Figure 10.
[0182] Figure 8 shows a system 1000 for making signal time series contextually interpretable in a chemical manufacturing plant. System 1000 includes an output device 1300, which in this embodiment is a monitor for displaying the signal time series that has been made contextually interpretable in the chemical plant. Additionally or alternatively, the output device may be a data storage device. Additionally or alternatively, the output device may be a database. Additionally or alternatively, the output device may be connected to another processing device. The system further includes an input device 1100 for providing one signal time series for at least one measurement point in the chemical manufacturing plant. The system further includes a processing device 1200, which for the purposes of this embodiment is a multipurpose computer, but may be any other suitable processing device. The processing device is configured to perform a method for making signal time series contextually interpretable in the chemical manufacturing plant.
[0183] Figure 9 shows an example of a method for making signal time series interpretable in context in a chemical manufacturing plant.
[0184] In the first step 700 of the processing device, one signal time series is received for each measurement point in the chemical manufacturing plant. In this embodiment, a first signal time series from the first time series is received from a first temperature sensor, a second time series is received from a second temperature sensor, a third time series is received from a first flow sensor, and a fourth time series is received from a fourth flow sensor.
[0185] In the second step 720, plant measurement point information is provided for each minimum measurement point, and measurement point information is provided for each sensor.
[0186] In this embodiment, step 730 includes determining at least one respective signal time series score for each signal time series at at least one measurement point based on the probabilistic characteristics of the signal time series, which includes determining the time series activity score and the volatility score. The activity score and volatility score are time series.
[0187] In this embodiment, in step 740, - One signal time series for at least one measurement point within a chemical manufacturing plant, - For each signal time series at least at one measurement point, the determined score for each respective signal time series, - Plant measurement point information for at least one measurement point, It is possible to generate annotated signal time series that include - Annotated signal time series for the first temperature sensor, including the following: 〇Time series of temperature signal from the first temperature sensor ○ Determined volatile time series for the first temperature sensor ○Activity time series determined for the first temperature sensor Plant sensor information for the first temperature sensor - Annotated signal time series for the first temperature sensor, including the following: 〇Time series of temperature signal from the second temperature sensor ○ Determined volatile time series for the second temperature sensor 〇Activity time series determined for the second temperature sensor Plant sensor information for the second temperature sensor - Annotated signal time series for the first flow sensor, including the following: 〇Time series of temperature signal from the first flow sensor ○ Determined volatile time series for the first flow sensor ○Activity time series determined for the first flow sensor Plant sensor information for the first flow sensor - Annotated signal time series for the second flow sensor, including the following: 〇Temperature signal time series from the second flow sensor ○ Determined volatile time series for the second flow sensor ○Activity time series determined for the second flow sensor Plant sensor information for the second flow sensor This includes generating [something].
[0188] Step 760, which clusters the annotated signal time series into clusters based on at least one determined signal time series score and each cluster classifier, is performed in this embodiment for the volatility score of the first temperature sensor. The classifier for the volatility score may include two clusters, high and low, as shown in Figure 5. Clustering may be performed by assessing whether the volatility score value is closer to the high cluster or the low cluster. A suitable method is k-means clustering. Clustering may also be performed for the activity score for the first temperature sensor.
[0189] In step 780, the operating state is assigned to the signal time series based on clustering and rule sets. In the preceding clustering step, all elements of the volatility score of the first temperature sensor are assigned to either the high cluster or the low cluster. The rule set for volatility clusters defines that high volatility is related to the ON state of the process group's operating state. The rule set for volatility clusters defines that low volatility is related to the OFF state of the process group's operating state. The rule set is based on the assumption that the sensor does not produce significantly changing signals while in the OFF state of its operating state. Here, for each sequence of the signal time series of the first flow sensor, it is made contextually interpretable whether the process group is ON or OFF.
[0190] In step 790, - One signal time series for each of one or more measurement points, - The minimum time-series score of each signal determined for each minimum measurement point, - Plant measurement point information for each minimum measurement point, - Assigned operating state, A context-interpretable signal time series, including It will be provided.
[0191] Figure 10 illustrates, in an illustrative scenario, how the activity score, volatility score, anomaly score, and steady-state score can fluctuate depending on the operating state of the process group. The plant's operating state is shown along the x-axis. From left to right, the plant is in an off state, which is indicated by a low volatility score and a low activity score. During plant startup, the volatility score and activity score show steps, such as an increase. This makes these two scores viable for distinguishing between the on and off states. The steady-state score shows a more stable progression over time. However, after the startup phase, when the plant is under normal operating conditions, the steady-state score shows a relatively flat slope. A threshold can be defined to distinguish between stable and unstable operating states. Similar to steady-state during stable operation, the anomaly score shows a relatively flat slope. In this example, a failure condition occurs, indicated as an event. Both the anomaly score and steady-state score show strong fluctuations. After the event, both scores essentially recover and again indicate a stable operating state. During shutdown, the steady-state score no longer indicates a stable operating state.
Claims
1. 1. A computer-implemented method for determining an operating state of a chemical plant, in particular an operating state within a process group, the method comprising: providing signal time series data relating to a first measurement point in a chemical plant; - determining, based on said signal time series data associated with said first measurement points in a chemical plant, a respective signal time series score associated with a measure of the degree of change of said signal time series data over time; - determining said operating state based on said respective signal time series scores related to a measure of the degree of change over time of said signal time series data; providing data relating to said operating conditions of said chemical plant suitable for monitoring and / or controlling said chemical plant; A method comprising:
2. providing signal time series data relating to a second measurement point within the chemical plant; determining, based on the signal time series data relating to the second measurement points, a respective signal time series score related to a measure of the degree of change over time of the second signal time series data; Further comprising: The method of claim 1 , wherein determining the operating state comprises determining the operating state based on the respective signal time series scores of the first and second signal time series data.
3. The method of claim 2 , wherein measurement point information is provided for the first measurement point and the second measurement point, the measurement point information including a technology related indicia.
4. 4. The method according to claim 3, wherein determining the operational state is further based in particular on the technology related indicator, and the step of determining the operational state comprises determining the operational state based on the respective signal time series scores of the first signal time series data and the second signal time series data related to a criterion of the degree of change of the signal time series data over time only if the technology related indicator indicates that the measurement point is technologically related, in particular belongs to a process group.
5. 5. The method of claim 1, wherein determining the operating state is based on the respective signal time series scores and respective cluster classifiers associated with a measure of temporal change in the signal time series data.
6. The method of any one of claims 1 to 5, wherein determining the operating state comprises determining an on state or an off state.
7. The method of any one of claims 1 to 6, wherein the data relating to the operational state of the chemical plant suitable for monitoring and / or controlling the chemical plant is indicative of an on or off state.
8. The method of any one of claims 1 to 7, further comprising: selecting signal time series data associated with an ON state; and determining a stability score for the selected signal time series data.
9. The method of claim 8 , further comprising determining stable operation based on the stability score.
10. The method of claim 8 , further comprising determining an event based on the stability score.
11. A method according to any one of claims 1 to 10, comprising providing signal time series data from measurement points and receiving data related to the operating state of the chemical plant, the data being generated according to any one of claims 1 to 10 and suitable for monitoring and / or controlling the chemical plant.
12. 1. A system for determining an operating state of a chemical plant, the system comprising: a time series acquisition unit configured to receive signal time series data relating to a first measurement point in the chemical plant; an on / off score unit configured to determine, based on the signal time series data associated with the first measurement point in the chemical plant, a respective signal time series score associated with a criterion of the degree of change of the signal time series data over time; an operating state determination unit configured to determine the operating state based on a respective signal time series score related to a criterion of the degree of change over time of the signal time series data, and an output unit configured to provide data related to the operating state of the chemical plant suitable for monitoring and / or controlling the chemical plant, said data being generated according to any one of claims 1 to 10; A system comprising:
13. A computer program product for determining an operational state of a chemical plant is disclosed, said computer program product comprising instructions configured to perform the steps of the method for determining an operational state of a chemical plant according to any one of claims 1 to 10 when executed on a computing device of a computing environment.
14. 11. A computer implemented method for monitoring and / or controlling a chemical plant, the method comprising the steps of receiving data related to an operating condition of the chemical plant, the data being suitable for monitoring and / or controlling the chemical plant, generated according to any one of claims 1 to 10, and providing a control signal based on the received data, in particular providing a control signal for shutting down the plant, in particular when the received data indicates an unstable operating condition.
15. 11. A computer implemented method for annotating signal time series data associated with a first measurement point in a chemical plant, the method comprising: receiving data associated with the operational state of the chemical plant generated in accordance with any one of claims 1 to 10, the data being suitable for monitoring and / or controlling the chemical plant, the data being generated in accordance with the method for determining an operational state of the chemical plant; and storing the data associated with the operational state of the plant together with the signal time series data.
