Facility operation assisting system and facility operation assisting method

Abstract

To provide a facility operation assisting system and a facility operation assisting method which are suitable for the training of an operator, and which can be configured at a low cost.SOLUTION: A facility operation assisting system includes: an input unit into which input data is input, the input data being alarm data when an abnormality is detected from a facility subjected to monitoring and control, operation data for an operation given by an operator to an apparatus in the facility, and monitoring data for the process amount or state of each component of the facility; an arithmetic unit that processes the input data; and a display unit that displays the process result by the arithmetic unit in a visualized manner. The arithmetic unit includes: a model creating unit that creates, using the input data input in past, a model for the feature quantity of the input data; a model utilizing unit which obtains the feature quantity of the input data using the input data input at the present time point, and which evaluates such a feature quantity using the model; and a visualizing unit that makes the evaluation result at the model utilizing unit visualized for displaying such an evaluation result on the display unit.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a facility operations support system and a facility operations support method. [Background technology]

[0002] In various industrial and other facilities (for example, water facilities in this case), a monitoring and control system using a computer is configured for monitoring and control, and if an abnormality occurs in the condition of the facility, an alarm message output by the monitoring and control system is displayed line by line on the control console, or the control console displays the values ​​of data (monitoring data) monitored by various sensors installed within the facility.

[0003] In response to such an alarm, operators in the monitoring and control work of water facilities check the alarm message displayed on the control console and any deterioration in the monitored data values, and then perform manual operations to restore the condition of the water facilities. After performing the operation, the operators also monitor the value of the monitored data and confirm that the operation has restored the condition.

[0004] In the series of processes from the occurrence of such an event to its response and confirmation, operators must manually determine and implement the appropriate operation based on the situation at the time, which requires operator know-how and places a heavy burden on junior operators in particular.

[0005] In this regard, Patent Document 1 proposes, with the aim of providing an operation assistance system that can utilize the know-how of operators, that "the operation assistance system comprises: an operation proposal rule generation unit that generates past intervention history information that associates the operating state of the equipment with an intervention operation based on operation history information that indicates the history of the operating state of the equipment operating in the plant and operation history information that indicates the history of intervention operations performed on the equipment, and generates rule information for performing the intervention operation based on the past intervention history information; an operation state comparison and determination unit that acquires operating state information that indicates the operating state of the equipment, and determines the intervention operation that corresponds to the operating state information based on the acquired operating state information and rule information; and an operation instruction information output unit that outputs operation instruction information related to the intervention operation determined by the operation state comparison and determination unit." [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2017-49801 Summary of the Invention [Problem to be solved by the invention]

[0007] According to the method of Patent Document 1, the objective can be achieved by configuring an AI guidance system, statistically analyzing and learning monitoring history data acquired over many years by numerous sensors installed in waterworks facilities, generating a plant control model, and providing operational support in accordance with the model.

[0008] However, the system in Patent Document 1 requires a huge amount of data, and the installation costs increase because it is necessary to install many sensors in various locations in the waterworks facilities to acquire data. Small and medium-sized facilities, in particular, tend to have few sensors. Furthermore, in the waterworks field, due to the declining population, many local governments are cutting their budgets, making it difficult to expand facilities.

[0009] As described above, with the conventional method, it is inevitable that the monitoring and control system itself will be expensive and complicated.

[0010] In view of the above, an object of the present invention is to provide an equipment operation support system and an equipment operation support method that can be configured even with a small amount of sensor data. [Means for solving the problem]

[0011] In light of the above, the present invention is described as "an equipment operation support system comprising an input unit that receives as input data alarm data when an abnormality is detected from a facility that is subject to monitoring and control, operation data regarding operations performed by operators on equipment within the facility, and monitoring data regarding the process quantities or states of each part of the facility, a calculation unit that processes the input data, and a display unit that visualizes and displays the processing results of the calculation unit, wherein the calculation unit comprises a model creation unit that creates a model for the feature quantities of the input data using the input data that was input in the past, a model utilization unit that determines the feature quantities of the input data using input data that is input at the current time and evaluates this using the model, and a visualization unit that visualizes the evaluation results of the model utilization unit and displays them on the display unit."

[0012] Furthermore, the present invention is described as "a facility operation support method using a computer to input as input data alarm data when an abnormality is detected in a facility to be monitored and controlled, operation data on operations performed by operators on equipment within the facility, and monitoring data on the process quantities or states of each part of the facility, processing the input data, and visualizing and displaying the processing results of the input data, and in processing the input data, executing in advance a model creation process to create a model for the feature quantities from the input data, and executing the processing results obtained by determining the feature quantities from the input data and referring to the model on the input data that has been input at the current time." [Effects of the Invention]

[0013] It is possible to provide an equipment operation support system and an equipment operation support method that can be configured inexpensively even with a small amount of sensor data. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a diagram showing an example of the configuration of a processing function 10 of a facility operation support system according to an embodiment of the present invention. [Figure 2] A diagram showing the usage relationships of various programs and models. [Figure 3] FIG. 2 is a diagram showing an example of the configuration of alarm history data D1. [Figure 4] FIG. 10 is a diagram showing an example of the configuration of operation history data D2. [Figure 5] FIG. 10 is a diagram showing an example of the configuration of monitoring history data D3. [Figure 6] FIG. 10 is a diagram showing the processing contents of a monitoring data fluctuation model creation function 11. [Figure 7] FIG. 10 is a diagram showing an example of a monitoring data fluctuation model M1. [Figure 8] 10 is a diagram showing an example of an extraction process of an operation within an alarm in the processing of the alarm / operation correspondence model creation function 12. FIG. [Figure 9A] 10 is a diagram showing an example of a specific operation corresponding to an alarm in the processing of the alarm / operation correspondence model creation function 12. FIG. [Figure 9B] 10 is a diagram showing an example of a specific operation corresponding to an alarm in the processing of the alarm / operation correspondence model creation function 12. FIG. [Figure 10] A diagram showing an example of alarm and operation compatible model M2. [Figure 11] FIG. 10 is a diagram showing the processing and generated contents of the alarm number frequency model creation function 13. [Figure 12] FIG. 10 is a diagram showing the processing and generated contents of the operation amount / monitoring data corresponding model creation function 14. [Figure 13] FIG. 10 is a diagram showing an example of the manipulated variable / monitoring data compatible model M4. [Figure 14] FIG. 10 is a diagram showing the processing and generated contents of the monitoring data recovery model creation function 15. [Figure 15] FIG. 2 is a diagram showing a list of configuration examples of a model M created by a model creation unit. [Figure 16] FIG. 1 is a diagram showing an example of the configuration of a water purification plant as an example of a facility. [Figure 17]A diagram showing how alarm occurrence information is displayed in different colors on equipment in a water purification plant where an alarm has occurred. [Figure 18] FIG. 10 is a diagram showing an example of a display screen when an operation is not supported. [Figure 19] FIG. 10 is a diagram showing an example of a display screen when recovery after operation is poor. DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. [Example]

[0016] FIG. 1 is a diagram showing an example of the configuration of a processing function 10 of a facility operations support system according to an embodiment of the present invention, which is configured using a computer.

[0017] In Figure 1, the computer is configured with a main memory device, a processing unit CPU, a display interface, an input / output interface, a network interface, etc. connected to a communication line BUS, and Figure 1 illustrates various data, processing programs, and models stored in storage units such as the main memory device as processing functions (processing units) 10.

[0018] The various data in Figure 1 are alarm history data D1, operation history data D2, and monitoring history data D3, which are data obtained from the equipment being monitored. Note that these data are not always referred to as "history." These data are numerical information or status information indicated by on / off, and are imported from the facility via the input unit.

[0019] The processing program consists of a model creation section for creating model M from data D1, D2, and D3, a model utilization section for processing and judging data D1, D2, and D3 obtained from the equipment using model M, and other programs.

[0020] In FIG. 1, the model creation unit includes a monitoring data variation model creation function 11 for creating a monitoring data variation model M1 in advance, an alarm-operation association model creation function 12 for creating an alarm-operation association model M2 in advance, an alarm count frequency model creation function 13 for creating an alarm count frequency model M3 in advance, an operation quantity-monitoring data correspondence model creation function 14 for creating an operation quantity-monitoring data correspondence model M4 in advance, and a monitoring data recovery model creation function 15 for creating a monitoring data recovery model M5 in advance.

[0021] In Figure 1, the model utilization part refers to a monitoring data value sudden change detection function 21 that detects a sudden change in the monitoring data value using a monitoring data fluctuation model M1, an operation content error detection function 22A that detects an error in the operation content performed by an operator using an alarm / operation response model M2, an operation incompatibility detection function 22B that detects an operation that an operator did not respond to using the alarm / operation response model M2, an alarm number increase detection function 23 that detects an increase in the number of alarms using an alarm number frequency model M3, an operation amount error detection function 24 that detects an error in the operation amount performed by an operator using an operation amount / monitoring data response model M4, and a post-operation behavior error detection function 25 that detects a post-operation behavior error using a monitoring data recovery model M5.

[0022] In Figure 1, the other programs are a monitoring data status deterioration detection function 32 that monitors the monitoring data chronologically and detects when the status of the monitoring data is deteriorating, and a visualization function 31 that visualizes the processing results (detection results) of the various programs mentioned above and displays them on a visualization screen 90.

[0023] Figure 2 shows the usage relationships of the various programs and models shown in Figure 1, and from the bottom up, it shows the model creation functions (11, 12, 13, 14, 15), various models (M1, M2, M3, M4, M5), model utilization functions (21, 22A, 22B, 23, 24, 25, 26), and visualization function 31.

[0024] This diagram shows that the creation of models (M1, M2, M3, M4, M5) by model creation functions (11, 12, 13, 14, 15) is performed in the advance preparation stage, and that various data obtained from the equipment under normal operating conditions is detected by various detection functions (21, 22A, 22B, 23, 24, 25, 26, 32), and the status is evaluated and judged using the models (M1, M2, M3, M4, M5). Note that this configuration may be such that modeling (enhancing the model) is performed using data obtained at the current time on one hand, and monitoring processing using the model for the data obtained at the current time on the other hand is performed simultaneously.

[0025] Next, we will explain in detail the various types of data obtained from the equipment. First, Figure 3 shows an example of the configuration of alarm history data D1 when alarm devices installed in various parts of the facility detect some kind of abnormality and go off, and the alarm history data D1 is configured in chronological order and includes data acquisition time information D1a, information on the facility that acquired the data and the facility name D1b, information on the alarm content D1c, and information D1d on whether the alarm started or recovered.

[0026] FIG. 4 shows an example of the configuration of operation history data D2 as a history of operations of equipment, etc. performed by an operator using various devices within a facility. The operation history data D2 is configured in chronological order and includes data acquisition time information D2a, ID number information D2b, the facility where the data was acquired, facility name information D2c, operation name information D2d, and operation amount information D2e.

[0027] Figure 5 shows an example of the configuration of monitoring history data D3. The monitoring history data D3 is a collection of monitoring data on various process quantities in a facility. In the example of Figure 5, which is intended to monitor a waterworks facility, the data is configured in chronological order to include, in addition to data acquisition time information D3a, pump discharge pressure information D3b, raw water inflow pressure information D3c, and purified water turbidity information D3d as process quantities.

[0028] In the present invention, models (M1, M2, M3, M4, M5) are generated from these various data in model generation units (11, 12, 13, 14, 15) as a preliminary step shown in Fig. 2. The process of generating these models will be described below.

[0029] First, the processing and generated contents of the monitoring data variation model creation function 11 for pre-creating the monitoring data variation model M1 will be explained using Figures 6 and 7. In Figure 6, the monitoring data variation model creation function 11 obtains the process quantities at each location in the facility, which are the monitoring data in Figure 5, in time series as monitoring data D3. The process quantities may be detected either analog or digitally, but are handled digitally in the processing of the monitoring data variation model creation function 11. However, in Figure 6, analog images are used to make it easier to understand.

[0030] The monitoring data D3 is pre-processed to exclude data from periods when various equipment in the facility is in a fault state, and then input into the monitoring data fluctuation model creation function 11. This allows data to be acquired during periods when the equipment is in operation. This period includes monitoring data D3 from normal operating states of the equipment in the facility as well as from abnormal states.

[0031] Then, as shown in the left frame, the monitoring data fluctuation model creation function 11 uses the maximum or minimum value of the monitoring data D3, which increases or decreases over time, as the starting point and the minimum or minimum value as the ending point, or vice versa, and grasps and accumulates the rate of change between these values. This can be achieved, for example, by approximating analog data to multiple line segments using the top-down method of piecewise linear functions, but here we want to find the slope of change (rate of change) for one fluctuation wave of the monitoring data D3, so there are several methods for achieving this, and any of these methods can be used in the present invention.

[0032] However, monitoring data D3 generally fluctuates over time, and in the case of the present invention, what is desired to be grasped is to monitor large fluctuations that are unlikely to occur under normal circumstances, such as during abnormal conditions, so it is desirable that the data be designed to make it easy to grasp fluctuations in long-period waves.

[0033] In this way, the monitoring data fluctuation model creation function 11 collects all the slopes of each approximated line segment, as shown in the right-hand box in Figure 6, which shows the monitoring data values ​​(rate of change: here written as data with a positive slope) versus time. It then stores the maximum value of the slope in one wave of this fluctuation. While chemical formulas can be used for water quality, this is more practical since the above data includes disturbances.

[0034] Figure 7 shows an example of the resulting monitoring data fluctuation model M1, with its characteristics represented by the magnitude of the slope on the horizontal axis and the number of past experiences of slope fluctuations on the vertical axis. This characteristic is obtained by accumulating experience, counting up the number of occurrences of the slope corresponding to the maximum value of the slope in one wave of the aforementioned fluctuation, each time this maximum value is experienced.

[0035] According to the characteristics shown in FIG. 7, the monitoring data D3 generally undergoes small fluctuations over short periods, and large fluctuations occur in the event of an abnormality, etc. Reflecting this, the number of times small fluctuations with a small slope are experienced is high, and the number of times large fluctuations with a large slope are experienced is low.

[0036] Next, the processing and generated contents of the alarm-operation correspondence model creation function 12 for pre-creating the alarm-operation correspondence model M2 will be described. The alarm-operation correspondence creation function 12 obtains the alarm history data D3 in Fig. 3 and the operation history data D2 in Fig. 4, and focuses on the equipment operations performed by operators at the time of alarm issuance in the same facility. Specifically, for example, since the alarm history data D3 and the operation history data D2 in Fig. 4 both contain time information D1a, D2a and facility name information D1b, D2c, the equipment names and operation amounts when operators performed equipment operations within the period when alarms were issued for the same facility are associated and understood.

[0037] Fig. 8 shows an example of the process of extracting operations during alarm issuance in the process of the alarm-operation correspondence model creation function 12. In this process, operations performed during the period from the alarm issuance time to the recovery time are extracted.

[0038] Specifically, in the example in the upper left of Figure 8, an operator performed equipment operation α while alarm A was sounding, and then performed operation β within this period. In this way, if multiple operations are performed for one type of alarm within the data period to be analyzed, all operations are extracted.

[0039] In addition, in the example at the bottom left of Figure 8, alarm B was issued after alarm A, and after a period of simultaneous alarms, alarms A and then B recovered, and operations α and γ were performed sequentially during the simultaneous alarm period, and operation δ was performed during the period when only alarm B was issued. In this case, if multiple alarms are being issued in the same time period, (multiple) operations are extracted for the multiple alarms.

[0040] Events that are triggered within this alarm period are organized as a list of pairs of (multiple) alarms and (multiple) operations, as shown in Table 1 in Figure 8. Here, the types of single alarms and multiple alarms are written on the vertical axis, and the number of times each alarm type, the operation type, and the number of times each operation type is written and organized on the horizontal axis.

[0041] 9A and 9B show examples of specifying an operation corresponding to an alarm in the processing of the alarm-operation correspondence model creating function 12. In this processing, an operation corresponding to an alarm is specified.

[0042] Specifically, in the example of operations during an alarm in Fig. 9A, we focus on alarm A as a certain alarm (called the target alarm), classify it into target alarm A and non-target alarms (other than A), and count the number of times each alarm is triggered. Then, we calculate the ratio of operations during each alarm to the number of target alarms triggered, and the ratio of operations during each alarm to the number of alarms other than the target alarms triggered.

[0043] In the example of Figure 9A, the two operations α and γ are counted as target alarm A, and the two operations δ and γ are counted as non-target alarms, but operation ε is not counted as either.

[0044] Table 2 in Figure 9B shows target alarm A, non-target alarms, and the ratio of operation counts on the horizontal axis, and each operation (α, β, γ) on the vertical axis, with the numerical ratios entered inside each matrix. For example, the execution rate of operation α out of the total number of operations performed during the period of target alarm A was 7.1%, the execution rate of operation β was 0.4%, and the execution rate of operation γ was 0.1%. Similarly, for non-target alarms, the execution rate of operation α out of the total number of operations performed during the period of non-target alarms was 1.4%, the execution rate of operation β was 0.1%, and the execution rate of operation γ was 1.9%.

[0045] Table 3 in Figure 9B is a diagram showing a matrix of individual alarms and individual operations, and the matrix shows the ratio of the individual operation rate of target alarms to the individual operation rate of non-target alarms calculated for each operation, as described in the right column of Table 2. In the matrix of Table 3, combinations of alarms and operations with a setting value greater than 0.5 are determined to be "related alarms and operations." In Table 3 in Figure 9B, the shaded areas represent "related alarms and operations."

[0046] Figure 10 shows an example of the alarm-operation correspondence model M2 created by the processing of Figures 9A and 9B above, with the areas with a colored background indicating "correspondence (related alarms and operations)." However, Figure 10 is an example showing values ​​for a different case from Figures 9A and 9B. According to Figure 10, the above correspondence determined from past performance is performed in response to an alarm, and therefore is likely to be an operation item aimed at improvement, reflecting the fact that this operation has been frequently used at least in past driving operations.

[0047] The above example describes one method for determining correspondence from past cases, but it goes without saying that other response methods are possible. It is recommended that the alarm-operation response model M2 not only determines whether or not an operation has been performed, but also numerically evaluates the number of times it has been performed. It can be said that the more times a similar operation has been performed, the more significant that operation is. Furthermore, it may be possible to reflect the results of this operation and give a high numerical evaluation to those that show a favorable improvement effect. The numbers in Figure 10 represent these evaluation values.

[0048] In this way, the processing of the alarm-operation correspondence model creation function 12 extracts the operations performed while an alarm is being issued, and the operations that are performed with statistical significance while an alarm is being issued are regarded as "operations performed to resolve the alarm (= to improve the value of the monitoring data)," and a correspondence table between alarms and operations (alarm-operation association model) is created.

[0049] Next, the processing and generated contents of the alarm number frequency model creation function 13 for creating the alarm number frequency model M3 in advance will be explained using Fig. 11. In Fig. 11, the alarm number frequency model creation function 13 obtains the alarm data of Fig. 3 and calculates the number of alarms for each time period in the same facility.

[0050] In the example of Figure 11, alarms AL1, AL2, AL3, and AL4 occurred in the same facility in the order shown, and then recovered. In this case, the initial alarm count changed from 0 to 1-2-3-2-1-2-1, and at the same time, the number of 1st alarms was counted as 3, the number of 2nd alarms was counted as 3, and the number of 3rd alarms was counted as 1.

[0051] In this case, the alarm number frequency model M3 is understood as the number of simultaneous alarms on the horizontal axis and the accumulated value of the number of times that number of simultaneous alarms has been experienced on the vertical axis. According to this model, when a case with a large number of simultaneous alarms is experienced, it can be estimated that the abnormal event is complex and that it is an event that is likely to be more difficult to resolve.

[0052] In this way, the processing of the alarm number frequency model creation function 13 creates a frequency distribution of the number of simultaneously issued alarms for each facility obtained from past alarm data.

[0053] Next, the processing and generated contents of the operation amount / monitoring data correspondence model creation function 14 for pre-creating the operation amount / monitoring data correspondence model M4 will be described with reference to Fig. 12. In Fig. 12, the operation amount / monitoring data correspondence model creation function 14 inputs the operation history data D2 in Fig. 4 and the monitoring history data D3 in Fig. 5, and refers to the alarm operation correspondence model M2 in Fig. 10.

[0054] Specifically, first, we will look at the alarm operation correspondence model M2 in Fig. 10 and focus on operations that are thought to be highly related to alarms. In the example of alarm operation correspondence model M2 in Fig. 10, we will focus on, for example, the relationship between alarm A and operation b, the relationship between alarm B and operation a, and the relationship between alarm C and operations b and c. In the following example, we will explain the relationship between alarm A and operation b as a representative example of operations that are thought to be highly related to alarms (operations that are thought to have a high probability of resolving alarm A), but the same will be done sequentially for operations a and c.

[0055] Next, the operation amount / monitoring data correspondence model creation function 14 references the time information of operation b during the issuance of alarm A, and monitors the time series of all monitoring data D3 within a certain appropriate time period from the time of this operation. Three examples of the response of the monitoring data D3 at this time are shown as upper, middle, and lower examples on the left side of the frame in Figure 12. Here, the monitoring data is displayed as analog data A and B, with time on the horizontal axis and the value y of the monitoring data on the vertical axis.

[0056] At this time, the operated variable / monitored data correspondence model creation function 14 extracts the operation direction of operation b (the increase / decrease direction of the operated variable D2e in Figure 4). In the example on the upper left side of the frame in Figure 12, analog data A moves in a decreasing direction after time ts while repeating small fluctuations, and is an example of fluctuation with a time delay after the operation. For analog data like this (especially water quality) that shows the effects of the operation for a long time, the behavior from the time (ts) after the operation when y is no longer a constant is extracted.

[0057] At this time, the operation amount / monitoring data correspondence model M4 stores, for each operation content, the operation direction (amount of change) of operation b and the name and direction of change (amount of change) of the monitoring data at that time as combined information, as shown in Figure 13.

[0058] In the example in the center of the frame in Figure 12, analog data A temporarily increased, but then decreased after time ts, which is also an example of a change with a time delay after the operation. In such a case, it is best to extract the behavior from the first peak after the operation. In this case, too, the operation direction (amount of change) of operation b and the name and direction of change (amount of change) of the monitored data at that time are stored as combined information for each operation.

[0059] In the example at the bottom of the box in Figure 12, analog data B began to decrease immediately after operation a, and is an example of a case where the data changed without any time delay after the operation. For analog data that has an effect immediately after an operation like this (especially hydraulics and equipment status), it is best to extract the behavior immediately after the operation. In this case, too, the operation direction (amount of change) of operation b and the name and direction of change (amount of change) of the monitored data at that time are stored as combined information for each operation.

[0060] In this way, all monitoring data fluctuations are checked in relation to the time of operation after the alarm, including those that do not produce fluctuations. If there is a fluctuation, the operation direction and the fluctuation direction of the process variable are stored as paired information. By determining statistically frequent pairs from the results of this processing, only highly correlated information can be extracted.

[0061] The monitoring data recovery model creation function 15 will be described with reference to Fig. 14. The monitoring data recovery model creation function 15 uses the same data as the operation amount / monitoring data correspondence model creation function 14 in Fig. 12.

[0062] The processing here differs from the processing in the operation quantity / monitoring data correspondence model creation function 14 in Fig. 12 in that it identifies the time when the analog quantity recovered (showed a tendency to recover) as a result of the operation, relative to the correspondence relationship between the monitoring data (analog data) and the operation (example on the left side of the frame in Fig. 12), and other prerequisite processing is the same as the processing in the operation quantity / monitoring data correspondence model creation function 14 in Fig. 12.

[0063] According to the processing examples of the monitoring data recovery model creation function 15 in Figure 14, in the cases in the upper left and center left of the frame (cases where the data moves in the recovery direction with a time delay after operation), the post-operation effects appear over a long period of time in analog data (particularly water quality), so the action time is the time from after operation until the time (ts) when y is no longer a constant. After ts, a line is fitted and the slope is extracted. In the processing example in the lower left of the frame, the post-operation effects appear over a short period of time in analog data (particularly hydraulics), so the line is fitted and the slope is extracted after the immediate peak.

[0064] The monitoring data recovery model M5 created by the monitoring data recovery model creation function 15 as a result of these processes can be understood as a graph with the horizontal axis representing the action time and the vertical axis representing the number of times for each monitoring data, and as a graph with the horizontal axis representing the slope and the vertical axis representing the number of times for each monitoring data.

[0065] FIG. 15 shows a list of configuration examples of the model M created by the model creation unit (11-15) described above. The processing in these sections essentially involves summarizing the feature quantities of inputs from the facility (alarm history data D1, operation history data D2, and monitoring history data D3) for any one or a combination of these inputs. The feature quantities are the number of occurrences per gradient for the monitoring data fluctuation model creation function 11 used to pre-create the monitoring data fluctuation model M1; the association between alarms and operations for the alarm-operation association model creation function 12 used to pre-create the alarm-operation correspondence model M2; the frequency per number of simultaneous alarms for the alarm count frequency model creation function 13 used to pre-create the alarm count frequency model M3; the correspondence between operation quantities and monitoring data for each operation for the operation quantity-monitoring data correspondence model creation function 14 used to pre-create the operation quantity-monitoring data correspondence model M4; and the time at which the monitoring data recovered or the frequency of the gradient for the monitoring data recovery model creation function 15 used to pre-create the monitoring data recovery model M5.

[0066] These model creation units (11-15) can be said to have determined feature quantities from past input data based on past experience. In contrast, the model utilization unit in Fig. 2 can be said to have determined feature quantities included in the models (M1-M5) created by the model creation units (11-15) from the input data at the current time (alarm history data D1, operation history data D2, and monitoring history data D3).

[0067] The feature quantities for the current time calculated by the model utilization unit are as follows: The monitoring data value sudden change detection function 21, which detects sudden changes in monitoring data values ​​using the monitoring data fluctuation model M1, calculates the magnitude of the monitoring data slope as the feature quantity for the current time. The operation content error detection function 22A, which detects errors in the operation contents performed by the operator using the alarm-operation response model M2, calculates the operation contents from the feature quantities stored in the alarm-operation response model M2 as the feature quantity for the current time. The operation non-response detection function 22B, which detects operations not performed by the operator using the alarm-operation response model M2, calculates the presence or absence of an operation from the feature quantities stored in the alarm-operation response model M2 as the feature quantity for the current time. The alarm count increase detection function 23, which detects an increase in the number of alarms using the alarm count frequency model M3, calculates the number of alarms as the feature quantity for the current time. The operation quantity error detection function 24, which detects errors in the operation quantities performed by the operator using the operation quantity-monitoring data response model M4, calculates the magnitude of the operation quantity as the feature quantity for the current time. The post-operation behavior error detection function 25 obtains the post-operation behavior as a feature quantity at the current time.

[0068] In the visualization program 31 of Figure 2, these feature quantities calculated by the model utilization unit can be displayed on the visualization screen individually or in a format that can be compared with the contents held in various models M, thereby providing the operator with information to make decisions.

[0069] Furthermore, the model utilization unit not only finds the feature values ​​at the current time that correspond to the feature values ​​of past performance, but also provides information on the current occurrence of an abnormality and its severity by comparing the past and current feature values.

[0070] Specifically, the monitoring data value sudden change detection function 21 detects a sudden change in the magnitude of the monitoring data slope and can provide information on the scale of the fluctuation based on past fluctuation records by referencing the monitoring data fluctuation model M1. The operation content error detection function 22A can provide information on the operation content itself being incorrect (e.g., operating an incorrect device as a response to an alarm) by referencing the alarm / operation response model M2. The operation non-response detection function 22B can provide information on failure to operate a device that should have been operated when an alarm occurred by referencing the alarm / operation response model M2. The alarm number increase detection function 23 can detect a sudden increase in the number of alarms and can provide information on the scale of the alarm being issued based on past alarm records by referencing the alarm number frequency model M3. The operation volume error detection function 24 can provide information on the magnitude of the operation volume itself being incorrect (e.g., operating a volume of 10 when a volume of 5 should have been operated) by referencing the operation volume / monitoring data correspondence model M4. The post-operation behavior error detection function 25 can provide information on the occurrence of an abnormality in post-operation behavior that does not stabilize by referencing the monitoring data recovery model.

[0071] The monitoring data status deterioration detection function 32 in Fig. 2 inputs current monitoring data and monitors its fluctuations to detect abnormalities in each monitoring data. At this time, the monitoring data status deterioration detection function 32 also inputs alarm and other statuses as real-time data, and also prepares and maintains a list of serious alarms that have been set up in advance by a human. The serious alarm list lists types of alarms that have the greatest impact, and if an alarm in the real-time data is an "alarm that requires immediate attention" that has been set up in advance as a serious alarm, a flag is output.

[0072] In the equipment operation support system of the present invention described in detail above, pre-operation processing involves modeling normal states related to the behavior of alarms, operations, and monitoring data from past operation data. These models are the monitoring data fluctuation model M1, the alarm-operation correspondence model M2, the alarm count frequency model M3, the operation amount-monitoring data correspondence model M4, and the monitoring data recovery model M5.

[0073] Furthermore, as processing during operation, when a large number of alarms are issued using the alarm number frequency model M3 and an abnormality in the behavior of the monitored data is detected using the monitored data fluctuation model M1, a flag is output; when an improper operation is detected using the alarm-operation response model M2 and the operation amount-monitored data response model M4 during actual operation, a flag is output and an operation is suggested; when a poor recovery after an operation is detected using the alarm-operation response model M2 and the monitored data recovery model M5 during actual operation, a flag is output and an operation is suggested; and furthermore, as described below, upon receiving the flag, the event detected as described above is displayed on the visualization screen for the relevant facility.

[0074] According to the equipment operation support system of the present invention described above, operators operate equipment based on their own judgment, and when an equipment alarm occurs, they operate the equipment based on their own judgment. However, by looking at the information provided by the equipment operation support system regarding various conditions and changes in circumstances that accompany this equipment operation, the operators can easily obtain guidelines for the correct solution based on information for understanding the current condition by comparing it with past cases and making decisions to take meaningful measures toward recovery. In this way, the present invention not only simply presents operations, but can also make suggestions for improving the situation when there is a defect in the manual operation made by an unskilled operator based on their own judgment, which is also effective in training operators. [Example]

[0075] In the first embodiment, it was explained that the model utilization unit should be provided with a monitoring data value sudden change detection function 21, an operation content error detection function 22A, an operation incompatibility detection function 22B, an alarm number increase detection function 23, an operation amount error detection function 24, and a post-operation behavior abnormality detection function 25.

[0076] These outputs are obtained by calculating the feature values ​​of each piece of data or the feature values ​​of the correspondence between pieces of data from the input data (alarm history data D1, operation history data D2, monitoring history data D3). In Example 2, we will explain how the model utilization unit can further facilitate the analysis of the equipment status from the relationships between multiple feature values ​​calculated.

[0077] Here, we will explain the relationship between multiple feature quantities obtained by the model utilization unit, detecting alarms and abnormalities in monitoring data, and detecting improper operations, and also explain how to handle poor recovery of monitoring data values ​​after operations.

[0078] First, regarding the detection of alarms and abnormalities in the monitoring data, in the second embodiment, the outputs of the monitoring data value sudden change detection function 21, the alarm count increase detection function 23, and the monitoring data value status deterioration detection function 32 are correlated and understood. Generally, when the value of each monitoring data D3 deteriorates, this is manifested as an increase in alarms, the issuance of a serious alarm, and a worsening trend in the monitoring data D3.

[0079] At this time, the alarm count increase detection function 23 uses the number of alarms, which is real-time data, and the alarm count frequency model M3 as input. When the alarm count, which is real-time data, is referenced to the alarm count frequency model M3, if there are many alarms issued in the same time period at one facility, it is determined that there is an increase in the number of alarms, and a flag is output.

[0080] At this time, the monitoring data status deterioration detection function 32 uses the input real-time data and a list of serious alarms that have been set up in advance by a human. The serious alarm list lists the types of alarms that have the greatest impact, and if an alarm in the real-time data is an "alarm that requires immediate attention" that has been set up in advance as a serious alarm, a flag is output. Furthermore, the monitoring data status deterioration detection function 32 monitors the monitoring data, and when the real-time data shows a worsening trend, it extracts the values ​​of the monitoring data in each facility that are showing a worsening trend and outputs a flag.

[0081] At this time, the monitoring data value sudden change detection function 21 uses real-time data and the monitoring data fluctuation model M1 as input. When the value of the monitoring data, which is real-time data, indicates abnormal monitoring data behavior (for example, a sudden rise or fall, or the possibility of equipment failure or disaster due to turbulence), if each monitoring data value fluctuates with a gradient greater than the maximum gradient of the monitoring data fluctuation model M1 (regardless of whether it was stable before), a flag is output.

[0082] In this way, when detecting abnormalities in alarms and monitoring data, it is effective for the operator to correlate the outputs of the monitoring data value sudden change detection function 21, the alarm count increase detection function 23, and the monitoring data value status deterioration detection function 32 and present them to the operator, helping the operator understand the situation. For this purpose, it is preferable that the visualization circuit 31 compiles these output results and displays them as being related to each other. Note that here, real-time data refers to information that has the same content as the operation history data D2, alarm history data D1, and monitoring history data D3, and is sent to the control desk immediately after an operation is performed, immediately after an alarm is issued, immediately after a sensor acquires a value and converts it into monitoring data.

[0083] Next, we will explain how to detect improper operations based on the relationships between multiple feature quantities obtained by the model utilization unit. In detecting improper operations, we focus on incomplete operations, incorrect operation content, and incorrect operation amount.

[0084] First, regarding non-operational response, the non-operational response detection function 22B uses real-time data and the alarm / operation response model M2 as input. In the non-operational response confirmation process, monitoring data indicating currently active alarms or a worsening trend (pre-alarm) is extracted. Then, by referencing the alarm / operation response model M2, if an operation corresponding to the alarm or deterioration monitoring data is not performed within a certain time, a flag is output and a corresponding operation is suggested.

[0085] Next, the operation error detection function 24 uses real-time data and the alarm / operation response model M2 as input. In the operation error confirmation process, monitoring data indicating currently active alarms or a worsening trend (pre-alarm) is extracted. Then, by referencing the alarm / operation response model M2, if an operation other than the operation corresponding to the alarm or deterioration monitoring data is performed, a flag is output and a corresponding operation is suggested.

[0086] Regarding manipulated variable errors, the manipulated variable error detection function 24 uses real-time data, alarm / operation correspondence model M2, and manipulated variable / monitoring data correspondence model M4 as inputs. The manipulated variable error confirmation process extracts monitoring data that indicates a currently active alarm or a worsening trend (pre-alarm). The direction of deterioration in the extracted monitoring data is then identified (e.g., a falling water level, an increasing residual salt concentration), and the alarm / operation association model M2 is then referenced to confirm that operations corresponding to the alarm or worsening monitoring data have been implemented. The direction of improvement in the direction of deterioration is also determined by referencing the manipulated variable / monitoring data correspondence model M4. If the manipulated variable of the implemented operation is in the opposite direction to the improvement, a flag is output and a manipulated variable in the direction of improvement is proposed.

[0087] In this way, when detecting improper operations, it is effective for the operator to understand the status by relating the outputs of the incompatible operation detection function 22B, the operation content error detection function 24, and the operation amount error detection function 24 and presenting them to the operator. For this purpose, it is preferable that the visualization circuit 31 group these output results together and display them as being related to each other.

[0088] Finally, we will explain how to detect poor recovery of monitoring data after an operation using the relationship between multiple feature quantities obtained in the model utilization section. Detecting poor recovery of monitoring data after an operation uses real-time data, alarm / operation response model M2, and monitoring data recovery model M5 as input. If the monitoring data value after an operation shows no recovery or worsens, it is expected that it will take time to respond, so it is necessary to quickly determine the poor recovery and take measures.

[0089] The post-operation behavior failure detection function 25 extracts monitoring data that indicates a currently occurring alarm or a worsening trend (pre-alarm), and refers to the alarm / operation response model M4 to confirm that an operation corresponding to the alarm or worsening monitoring data has been carried out, or after the corresponding operation has been carried out, it checks the behavior of the relevant monitoring data, and refers to the monitoring data recovery model M5 to see if there is no improvement or if the situation has worsened, outputs a flag and proposes another operation.

[0090] As described above in the second embodiment, the multiple feature quantities determined by the model utilization unit are preferably presented individually so as to be comparable with past history, and also presented collectively (distinguished from other factors) from the viewpoint of alarms, detection of abnormalities in monitoring data, or detection of operational defects. This enables the presentation of information that is easy for operators to analyze events and take optimal action, and makes it possible to configure an equipment operation support system that is also suitable for operator training.

[0091] Furthermore, based on the processing results of incorrect operation content, incompatible operation, and poor behavior after operation, the operation that the operator should perform can be displayed on the display unit, making it possible to create a suggestion-type system. [Example]

[0092] In the third embodiment, a display example of the visualization screen 90 will be described. As described in the first embodiment, the visualization screen 90 can display the feature quantities in the model created using the model creation unit and the feature quantities of the current input in a comparable manner. It can also display the results of determining the state of the current feature quantities. Furthermore, as described in the second embodiment, it can display the processing results of multiple model utilization units in a single display according to a specific purpose.

[0093] In Example 3, we will further explain how to display the information in relation to the configuration of the facility to be monitored. Figure 16 shows an example of the configuration of a water purification plant as an example of a facility. The system is configured such that an intake tower, a sedimentation basin, a filtration basin, a chemical dosing basin, and a drainage basin are arranged in this order from the upstream side of the drinking water supply, and the system after the filtration basin is arranged in two systems.

[0094] FIG. 17 shows that alarm occurrence information is displayed in a color-coded manner for the equipment in the water purification plant where an alarm has occurred (here, the upper series of filtration basins and chemical dosing basins).

[0095] Figure 18 shows an example of a facility where an inappropriate operation occurred (here, the upper series filtration basin), where the cause and suggested action were not taken, and the need to check whether the operation amount was appropriate. In this example, after an alarm was issued, a warning and suggested remedial action were displayed in the event of an inappropriate operation.

[0096] Figure 19 shows the equipment where recovery failure occurred after operation (here, the upper series chemical feeding pond), and displays the equipment where recovery failure occurred after operation was carried out.

[0097] As described above, the visualization screen presents basic information in an appropriate format for operators to make appropriate decisions. This allows even junior operators to make correct decisions, making it an equipment operation support system that can provide information suitable for training. [Explanation of symbols]

[0098] 11: Monitoring data fluctuation model creation function 12: Alarm and operation correlation model creation function 13: Alarm frequency model creation function 14: Model creation function for operation amount and monitoring data 15: Monitoring data recovery model creation function 21: Function to detect sudden changes in monitored data values 22A: Operation error detection function 22B: Operation incompatibility detection function 23: Alarm number increase detection function 24: Operation error detection function 25: Post-operation behavior detection function 31: Monitoring data status deterioration detection function 32: Visualization function 90: Visualization screen M1: Monitoring data fluctuation model M2: Alarm and operation compatible model M3: Alarm frequency model M4: Model compatible with operation volume and monitoring data M5: Monitoring Data Recovery Model

Claims

1. The system comprises an input unit that inputs alarm data when an abnormality is detected from a facility to be monitored and controlled, operation data regarding operations performed by an operator on equipment within the facility, and monitoring data regarding process quantities or states of each part of the facility as input data, a calculation unit that processes the input data, and a display unit that visualizes and displays the processing results of the calculation unit, The calculation unit comprises a model creation unit that creates a model for feature quantities of input data using the input data that was input in the past, a model utilization unit that obtains feature quantities of the input data using the input data that is input at the current time and evaluates the input data using the model, and a visualization unit that visualizes the evaluation results in the model utilization unit and displays them on the display unit.

2. 2. The equipment operation support system according to claim 1, the model creation unit is configured by a model creation unit that calculates a feature value for a single piece of input data and a model creation unit that calculates a feature value between a plurality of pieces of input data; the model creation unit that obtains a feature quantity for the input data alone includes any one of a monitoring data variation model creation function for creating a monitoring data variation model from the monitoring data, an alarm number frequency model creation function for creating an alarm number frequency model from the alarm data, and a monitoring data recovery model creation function for creating a monitoring data recovery model from the monitoring data; An equipment operation support system characterized in that the model creation unit that obtains feature quantities for multiple input data includes either an alarm / operation association model creation function for creating an alarm / operation correspondence model from alarm data and operation data, or an operation quantity / monitoring data correspondence model creation function for creating an operation quantity / monitoring data correspondence model from operation data and monitoring data.

3. 3. The equipment operation support system according to claim 2, an alarm count frequency model creation function that creates a feature value that is the slope of the monitoring data; an alarm count frequency model creation function that creates a feature value that is the number of alarms issued simultaneously; an alarm recovery model creation function that creates a feature value that is the time it takes for the monitoring data to recover or the slope of the recovery; an alarm-operation correlation model creation function that creates a feature value that is the number of operations performed when an alarm is issued; and an operation quantity-monitoring data correspondence model creation function that creates a feature value that is the direction of increase / decrease in the operation quantity and the direction of increase / decrease in the monitoring data.

4. 3. The equipment operation support system according to claim 2, The model utilization unit is provided with any one of a monitoring data value sudden change detection function that detects a sudden change in the monitoring data value using the monitoring data variation model, an operation content error detection function that detects an error in the operation content performed by an operator using the alarm / operation response model, an operation non-response detection function that detects an operation not performed by an operator using the alarm / operation response model, an alarm number increase detection function that detects an increase in the number of alarms using the alarm number frequency model, an operation quantity error detection function that detects an error in the operation quantity performed by an operator using the operation quantity / monitoring data correspondence model, and a post-operation behavior abnormality detection function that detects poor behavior after operation using the monitoring data recovery model.

5. 2. The equipment operation support system according to claim 1, the model creation unit of the calculation unit has a monitoring data fluctuation model creation function for creating a monitoring data fluctuation model from the monitoring data, and an alarm number frequency model creation function for creating an alarm number frequency model from the alarm data, The model utilization unit of the calculation unit has a monitoring data value sudden change detection function that detects a sudden change in the monitoring data value input at the current time point using the monitoring data fluctuation model, and an alarm number frequency increase detection function that detects an increase in the number of alarms at the current time point using the alarm number frequency model, The visualization unit of the calculation unit visualizes the processing results of the monitoring data value sudden change detection function and the processing results of the alarm count frequency increase detection function in a format in which they are associated with each other, and displays them on the display unit.

6. 2. The equipment operation support system according to claim 1, the model creation unit of the calculation unit has an alarm-operation association model creation function for creating an alarm-operation correspondence model from alarm data and operation data, an operation amount-monitoring data correspondence model creation function for creating an operation amount-monitoring data correspondence model from operation data and monitoring data, and a monitoring data recovery model creation function for creating a monitoring data recovery model from the monitoring data; The model utilization unit of the calculation unit has an operation content error detection function that detects an error in the operation content performed by an operator using the alarm / operation correspondence model, an operation incompatibility detection function that detects an operation incompatibility that the operator did not perform using the alarm / operation correspondence model, an operation amount error detection function that detects an error in the operation amount performed by the operator using the operation amount / monitoring data correspondence model, and a post-operation behavior error detection function that detects a post-operation behavior error using the monitoring data recovery model, An equipment operation support system characterized in that the visualization unit of the calculation unit visualizes the processing results of the operation content error detection function, the processing results of the operation incompatibility detection function, and the processing results of the post-operation behavior poor detection function in a format in which they are associated with each other and displayed on the display unit.

7. 7. The equipment operation support system according to claim 6, An equipment operation support system characterized in that the visualization unit of the calculation unit displays on the display unit the operations that the operator should perform based on the processing results of incorrect operation content, incompatible operation, and poor behavior after operation.

8. 2. The equipment operation support system according to claim 1, The facility operation support system is characterized in that the operation data and the monitoring data are data during a period when the alarm data indicates an alarm is being issued.

9. 5. The equipment operation support system according to claim 4, The model utilization unit has a monitoring data condition deterioration detection function that detects that the condition of the monitoring data is deteriorating, and when there is monitoring data that is deteriorating or when an alarm is currently being issued, it refers to the alarm / operation response model to confirm that an operation corresponding to the alarm or the deteriorating monitoring data has been carried out, and after the corresponding operation is carried out, it checks the behavior of the corresponding monitoring data, and refers to the monitoring data recovery model to see if there is any improvement or if the condition has deteriorated, The visualization unit of the calculation unit outputs a flag indicating poor recovery of the monitoring data value after the operation, and proposes that the operation be performed again.

10. 2. The equipment operation support system according to claim 1, An equipment operation support system characterized in that the visualization unit of the calculation unit displays the configuration of multiple pieces of equipment that make up the facility to be monitored and controlled, and identifies and displays the corresponding equipment based on the detected evaluation results.

11. 2. The equipment operation support system according to claim 1, The calculation unit detects an operation error based on the features of the input data or a case where the monitoring data is not recovered within a predetermined period after the operation, and outputs operation data based on a model of the input data to the display unit.

12. Using a computer, alarm data when an abnormality is detected in the facility to be monitored and controlled, operation data regarding operations performed by operators on equipment within the facility, and monitoring data regarding the process quantities or states of each part of the facility are input as input data, the input data are processed, and the processing results of the input data are visualized and displayed, In processing the input data, a model creation process is performed in advance to create a model for the feature quantities of the input data, and the feature quantities are obtained from the input data, and the processing results obtained by referencing the model are then executed on the input data entered at the current time.

Images