Method and unit for dynamically adjusting air compressor system maintenance plan

Abstract

The invention provides a method and a unit for dynamically adjusting a maintenance plan of an air compressor system. The method comprises the following steps: detecting the operation state of each part in the air compressor system to obtain multiple pieces of air compressor operation data; classifying the operation data based on the properties of the operation data, wherein the first type of operation data is data related to pressure detection and / or vibration detection; the second type of operation data is data related to temperature detection; the third type of operation data represents data obtained by checking electric appliance parts of the air compressor system, the first type of operation data is processed through a prediction method based on a time sequence to determine whether maintenance is needed or not within a prediction time period, and / or fault diagnosis is conducted based on a detected potential abnormal result through clustering analysis; aiming at the second type of operation data, adopting clustering analysis to carry out fault diagnosis based on a detected potential abnormal result; and whether the air compressor system needs to be maintained or not is determined according to logic based on rules for the third type of operation data.

Description

Technical Field

[0001] The present invention relates to the technical field of online remote monitoring and fault diagnosis of air compressor systems, and particularly to a method and unit for dynamically adjusting the maintenance plan of air compressor systems. Background Art

[0002] As a key element of modern manufacturing processes, air compressors, as large equipment required for industrial production, are used to store potential energy in storage tanks for future use by pressurizing ambient air. For example, an air compressor system can be used for dust removal within a factory. Therefore, comprehensively grasping, analyzing, and predicting the operating status and health level of air compressor systems, improving the fault diagnosis level, and reasonably arranging repairs or predictive maintenance are necessary bases for ensuring the normal operation of industrial production.

[0003] Currently, for air compressor systems, the system must be able to shut down within 10 seconds to ensure the normal operation of the facility. In addition, the maintenance cost of air compressor systems ranks among the top in all FCM-related systems. So far, regarding the operation of air compressors in factories, operators usually use the upper and lower limit values of each parameter to monitor the operating status. In the prior art, air compressor alarms are generated entirely based on rule-based logic. The selected limit values of some currently available sensors are shown in Table 1 below:

[0004] Table 1

[0005] However, since such inspection methods are difficult to immediately detect potential faults and may not be able to generate alarms in time to optimize the maintenance plan, there is still room for saving in terms of additional costs including maintenance costs and diagnostic costs.

[0006] In addition to the above monitoring of fixed limit values for air compressors, in the prior art, fixed maintenance plans are usually adopted for different components in air compressor systems (see Table 2 below). The resulting problem is that if the frequency of regular maintenance is too high, additional maintenance costs and diagnostic costs will be generated; conversely, if the frequency of regular maintenance is too low, it will negatively affect the performance of air compressor systems. Therefore, it is of great significance to automatically alarm the detected abnormal conditions and potential causes in order to reduce the risk of energy waste and avoid potential shutdowns. Finally, so far, the vast majority of predictive maintenance (PM) functions do not have a feedback loop, and due to the lack of a knowledge base for air compressors, experts in this field need to diagnose similar faults from time to time. Therefore, the existing predictive maintenance functions are not suitable for better control and notification.

[0007] TPM List Database Maintenance Plan Air Filter Differential Pressure Value Every 2000 Hours Oil Filter N / A Oil Separator Differential Pressure Value Every 4000 Hours Motor Bearing Vibration Value Every 16000 Hours Motor Lubrication N / A Rotor Lubrication N / A Pipe Connector N / A Power Contactor Motor Starting Time Every 4000 Hours Heat Exchanger Cooling Water Temperature Every 4000 Hours Ball Joint Vibration Value Every 16000 Hours 8000-Hour Maintenance Kit N / A Rotor Maintenance Vibration Value Every 32000 Hours Oil Separator Alarm Light N / A Oil Pipeline Cleaning N / A Fan Inspection N / A Cooler Cleaning Oil Temperature Every 4000 Hours Motor Maintenance Motor Oil Temperature Every 32000 Hours Inlet Power Cable N / A

[0008] Table 2

[0009] It should be noted that the "Background Art" paragraph is only used to help understand the content of the present invention. Therefore, the content disclosed in the "Background Art" paragraph may include some prior art that is not known to those skilled in the art. The content disclosed in the "Background Art" paragraph does not represent the problems to be solved by such content or one or more embodiments of the present invention, and has been known or recognized by those skilled in the art before the filing of the present invention application. Summary of the Invention

[0010] An object of the present invention is to overcome the above-mentioned defects in the prior art, and to provide a method and unit for dynamically adjusting the maintenance plan formulated based on rules in the original air compressor system, so as to effectively achieve the balance between the too-high and too-low maintenance frequencies of the existing air compressor system, and without affecting the operating performance of each component in the system while saving maintenance costs.

[0011] To achieve the above object, a method for dynamically adjusting the maintenance plan of an air compressor system is provided, including the following steps:

[0012] Detect the operating states of each component in the air compressor system to obtain a plurality of air compressor operation data;

[0013] Classify the air compressor operation data based on the nature thereof, wherein the first type of operation data is data related to pressure detection and / or vibration detection; the second type of operation data is data related to temperature detection; and the third type of operation data represents the data obtained according to the inspection of the electrical components of the air compressor system.

[0014] Among them, for the first type of operation data, a time-series-based prediction method is used for processing to determine whether maintenance of the air compressor system is required within the prediction time period, and / or clustering analysis is adopted to perform fault diagnosis based on the detected potential abnormal results to determine whether maintenance of the air compressor system is required; for the second type of operation data, clustering analysis is adopted to perform fault diagnosis based on the detected potential abnormal results to determine whether maintenance of the air compressor system is required; and for the third type of operation data, a rule-based logic is used to determine whether maintenance of the air compressor system is required.

[0015] According to another aspect of the present invention, a unit for dynamically adjusting the maintenance plan of an air compressor system is provided, including:

[0016] A data detection module that detects the operating states of each component in the air compressor system to obtain a plurality of air compressor operation data;

[0017] A data storage module that stores the air compressor operation data from the data detection module;

[0018] A central processing unit that calls the operation data of the air compressor from the data storage module and executes the method described above to obtain a fault diagnosis result; and

[0019] A monitoring terminal that obtains the operation data of the air compressor and the fault diagnosis result from the central processing unit and displays them for dynamically adjusting the maintenance plan of the air compressor system.

[0020] According to another aspect of the present invention, there is provided a computer-readable storage medium storing a program that, when executed by a central processing unit, enables the central processing unit to execute the method described above. Description of the Drawings

[0021] The above and other aspects of the present invention will be more thoroughly understood and recognized in conjunction with the accompanying drawings, where:

[0022] Figure 1 Shows a flowchart of a method for dynamically adjusting the maintenance plan of an air compressor system according to a preferred embodiment of the present invention;

[0023] Figure 2 Is a schematic diagram for dynamically adjusting the maintenance plan of an air compressor system by performing time series prediction on pressure according to a preferred embodiment of the present invention;

[0024] Figure 3 Is a schematic diagram for diagnosing abnormal data points detected by a vibration sensor based on a clustering algorithm according to a preferred embodiment of the present invention;

[0025] Figures 4A to 4D Is a schematic block diagram for updating the knowledge base of an air compressor system using the logic based on case-based reasoning according to the present invention; and

[0026] Figure 5 Is a schematic block diagram of a unit for dynamically adjusting the maintenance plan of an air compressor system according to the present invention. Detailed Description of the Invention

[0027] In order to make the technical problems to be solved, the technical solutions and the beneficial technical effects of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and exemplary embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and not to limit the protection scope of the present invention.

[0028] Compared with the method of formulating the maintenance plan of the air compressor system using rule-based logic introduced in the previous background art section, the present invention utilizes a time-series based prediction algorithm and an unsupervised clustering algorithm to automatically identify potential faults in the air compressor system and optimize its original fixed maintenance plan, thereby formulating a more reasonable maintenance plan. In addition, a knowledge base based on case-based reasoning logic is developed to assist domain experts in more quickly determining whether it is necessary to shut down and maintain the air compressor system with potential faults and determining corresponding maintenance measures by reducing repetitive diagnostic work and recommendations. In addition, the present invention also develops a unit for dynamically adjusting the maintenance plan of the air compressor system for more convenient monitoring and notification.

[0029] Referring to Figure 1 , which shows a flowchart of a method for dynamically adjusting the maintenance plan of an air compressor system according to a preferred embodiment of the present invention. First, at step S10, the operating states of the components in the air compressor system are detected to obtain a plurality of air compressor operation data. Depending on the type of the monitored component, its operation data can be obtained through existing sensors and stored in a relevant data storage module and subsequently can be called by a central processing unit (see Figure 5 ). Taking the air filter in the air compressor system as an example, pressure sensors with detection heads located at its inlet end and outlet end respectively can be used to sense the pressure difference across the air filter in real time, thereby serving as the operation data of the air filter. Similarly, various operation data listed in Table 1 and Table 2 above can be obtained by means of corresponding vibration sensors, temperature sensors, pressure sensors, etc.

[0030] Next, at step S20, according to the nature of different operation data, they are classified to adopt different fault prediction and diagnosis methods for different types of operation data. From Figure 1As can be seen, the operating data is generally divided into three categories. The first category of operating data is data related to pressure detection and / or vibration detection. For example, the differential pressure of the air filter in the air compressor system, the differential pressure of the oil separator, the vibration decibel value at the motor drive end / non-drive end, the vibration decibel value at the air compressor rotor drive end / non-drive end, etc. When this type of operating data exceeds a limit value that triggers the maintenance of the corresponding components in the air compressor system, it means that the corresponding components should be maintained. The second category of operating data is data related to temperature detection, including but not limited to the outlet temperature of the air compressor, the outlet temperature of the cooling water, the U-phase, V-phase, and W-phase temperatures of the motor coil, etc. In addition, there is a third category of operating data obtained by statistical analysis of the operating conditions of the system, such as the number of motor starts per unit time. Different from the first and second categories of operating data that need to be sensed by various sensors, the third category of operating data is not obtained by detecting the physical quantities of the relevant components in the air compressor system. In the present invention, for the first category of operating data, a time-series-based prediction method can be used to process and notify the operator whether maintenance of the air compressor system is required within the predicted time period, and / or clustering analysis can be used to notify the operator of the detected potential abnormal results; for the second category of operating data, clustering analysis is used to notify the operator of the detected potential abnormal results; and for the third category of operating data, a rule-based logic is used to notify the operator whether maintenance of the air compressor system is required.

[0031] The following combines Figure 1 and Figure 2 to illustrate the process of time-series prediction for the above-mentioned first category of operating data. Taking the differential pressure value of the air filter as an example, in Figure 2 the coordinate system shown, the historical data of the differential pressure value (absolute value) of the air filter measured over a period of time is shown. The abscissa represents time, and the ordinate represents the differential pressure value measured by the pressure sensor at the corresponding time point. As can be seen from the figure, as the air compressor system operates, the above-mentioned differential pressure generally shows a tendency to change towards a limit value. The dotted line with a vertical coordinate of 1.0 bar at the top of the figure represents this limit value. When the differential pressure value of the air filter exceeds this limit value, it means that the air filter should be maintained. In this embodiment, at Figure 1At step S201 shown, historical data of a certain time window length (in this embodiment, it is the past year) is selected (i.e., the data corresponding to Region I, and only the data of the most recent about five weeks is shown for simplicity) to predict the change trend of the differential pressure value within a future prediction time period (in this embodiment, it is 14 days, i.e., the two weeks corresponding to Region II), and at step S301, the predicted differential pressure value is compared with the above limit value (especially to determine whether the predicted value exceeds the limit value) to determine whether it is necessary to send a notice to the operator. For example, if the predicted differential pressure value within 14 days is lower than the acceptable limit value before the originally planned maintenance operation, then the next maintenance operation will be postponed to save maintenance costs. It can be envisioned that in this case, the method returns to step S201 and the actually measured differential pressure value within the next 14 days can be used to update the historical data, and based on the updated historical data, the prediction of the differential pressure value for a future specified date is repeated until the predicted data exceeds the acceptable limit value within the prediction time period. On the contrary, if the predicted differential pressure value within 14 days will exceed the acceptable limit value before a specific maintenance operation, then at S401, the operation result is determined to be abnormal and a report is sent to the operator in advance to maintain the original maintenance operation time or advance the original maintenance operation time based on the fact that the actually measured differential pressure value exceeds the acceptable limit value. It should be noted here that although in this embodiment, the historical data of the past year is used to predict the operation data of the next 14 days, this is not a limitation of the present invention. The operator can select historical data from any time period from the current to the past (for example, since the last maintenance of the relevant component) according to the actual situation and needs to predict the operation data of a future certain time (such as one week, one month, etc.). It should be noted here that as the ratio between the time window length of the selected historical data and the future prediction time period (in this embodiment, it is about 26, i.e., 365 days / 14 days) increases, the prediction of future operation data will be more accurate. In addition, for time series-based prediction, current ARMA (Autoregressive Moving Average) models, ARIMA (Autoregressive Integrated Moving Average) models, neural network models (such as including LSTM, Transformer, Informer, etc.) can be used, which will not be elaborated here.

[0032] The following combines Figure 1 and Figure 3 to illustrate the process of potential fault diagnosis using the clustering algorithm. In Figure 3 the relationship between the vibration decibel value (ordinate) of the motor bearing measured by the vibration sensor and time (abscissa) is schematically shown. It can be seen from the figure that the vibration decibel value of the motor bearing during the operation of the air compressor system is not as Figure 2The pressure difference value of the air filter also shows a certain change trend over the operation time, but fluctuates around a mean value. For normal operation, the vibration decibel value measured at a certain time point should be within a circle centered at this mean value with a limit value r as the radius. In order to distinguish the abnormal data points indicating potential faults of the corresponding components of the air compressor system outside the coverage of this circle from the normal data points indicating the normal operation of the corresponding components, the present invention particularly uses the mean-shift clustering algorithm to process the vibration decibel values of components such as motor bearings ( Figure 1 as described in step S202). More specifically, in step 1), for multiple unclassified data points obtained over a period of time, a data point is randomly selected as the center point; in step 2), all points within the radius r from this center point are obtained as a set, and these points are considered to belong to the same cluster; in step 3), the vectors of all data points within the circular space with a radius of r from this center point are calculated, and the average value of all vectors within this circular space is calculated to obtain a shift vector; in step 4), the center point is moved by this shift vector, that is, the moving direction is the direction of the shift vector, and the moving distance is the modulus of the shift vector to reach the updated position of the center point; then, in step 5), the operations of the above steps 2), 3), and 4) are repeated until the magnitude of the shift vector meets the set threshold requirements. At this time, the position of the center point no longer changes (that is, the center point moves to the place with the maximum density of all data points), and the center point at this time is recorded. Then, the operations of the above steps 1) to 5) are repeated until all data points are classified. Finally, according to each category and the access frequency of each data point, the category with the maximum access frequency is selected as the category to which the current data point belongs, and the label value of the category containing the most data points is specified as 0, the label value of the category containing the second most data points is specified as 1, and so on. The data points in the category with a label value greater than 2 represent abnormal data points outside the normal data points (as shown in the right ellipse in the figure), meaning that there may be potential faults in the corresponding components of the air compressor system that require shutdown maintenance. As Figure 1 shown, after performing mean-shift clustering on the above data, it is judged at step S302 whether the label value representing the category is greater than 2. If the result is yes, then proceed to the following step S402; otherwise, if the result is no, return to step S202 to continue performing mean-shift clustering on the real-time measured operation data. Further, at step S402, it is judged whether the operation data value measured at the corresponding time exceeds the limit value for triggering the maintenance of the corresponding component in the air compressor system. If the result is yes, then turn to step S401 to judge as abnormal and send a notice to the operator; otherwise, if the result is no, then proceed to the optional step S502 shown in the figure.

[0033] For an existing air compressor system, the following operating conditions may exist: during the downtime period, all components operate in a standby state (i.e., not fully loaded state) with a reduced power consumption compared to normal operation (here, the air compressor system does not stop to save power consumption because, according to practical operations, completely shutting down the entire system and then cold starting it usually results in greater energy consumption compared to hot starting and restoring to the fully loaded state in the standby state). In this case, the operating parameters of the corresponding components of the air compressor system appear to be "abnormal" compared to the normal working state on the surface, but there are no potential faults in the actual relevant components. To avoid misdiagnosis due to the above situation, in the optional step S502, it is further determined whether the measurement period classified as abnormal data points by the mean-shift clustering algorithm is during the above-mentioned downtime period, that is, to determine whether the air compressor system is currently in the not fully loaded state. If the result is yes, it means that the above abnormal data points are only caused by the air compressor system operating in the standby state. Accordingly, proceed to step S5021 to output a detection result indicating no abnormality; conversely, if the result is no, it means that the above abnormal data points are not caused by the air compressor system operating in the standby state, but by potential faults in the corresponding components. In this case, proceed to step S5022 to report the abnormal data and potential faults to the domain expert.

[0034] The above uses the vibration decibel value of the motor bearing detected by the vibration sensor as an example to illustrate the detection of potential faults using the clustering algorithm. According to the foregoing content of the present invention, a similar method can also be selected for the detection of potential faults for the pressure difference detected by the pressure sensor. In other words, for the first type of operating data related to pressure and / or vibration detection, a time-series-based prediction method as shown in Figure 2 can be selected for processing and / or a mean-shift clustering analysis as shown in Figure 3 can be selected for the detection of potential faults. For the second type of operating data related to temperature detection, clustering analysis can also be used to notify the operator of the detected potential abnormal results, and its operating principle is similar to the previous description of vibration data and will not be elaborated here. It should be noted that although the mean-shift clustering analysis is used as an example in this embodiment, those skilled in the art can also think of using other clustering analyses such as k-means (k-average), k-Medoids (k-median), DBSCAN (density-based spatial clustering of applications with noise), etc. to process the relevant operating data.

[0035] It should be noted here that when performing clustering analysis on operation data such as temperature, in order to avoid the influence of temperature data of relevant components of the air compressor system during the process from the shutdown state or standby state to the normal operation state on the detection result, it is necessary to preprocess the temperature data to be involved in the clustering analysis. For example, if the system runs from a complete shutdown state to a full load state, the temperature data of the earliest 30 minutes can be removed; on the contrary, if the system runs from a full load state to a complete shutdown state, the temperature data of the last 30 minutes is removed. On the other hand, if the system runs from a standby state (i.e., not fully loaded state) to a full load state, the temperature data of the earliest 15 minutes can be removed; on the contrary, if the system runs from a full load state to a standby state, the temperature data of the last 15 minutes is removed.

[0036] Figure 1 The operation of monitoring the third type of operation data is further shown. In this embodiment, this type of operation data is used for the electrical system inspection of the air compressor system. For example, the number of motor starts within a specified time (such as every hour) formulated based on rules is monitored, and each time the motor starts, the corresponding counter is incremented by 1. For the air compressor system, the motor is usually started to increase the load when the operating load is insufficient. If the load of the air compressor system is detected to switch between high and low too frequently within the specified time, it means that the relevant components may need to be maintained. Accordingly, at step S203, based on the count value of the counter, it is judged whether the number of motor starts within the specified time exceeds a threshold value (3 times in this embodiment, but the present invention is not limited thereto). If the result is yes, it turns to step S2031 to report an abnormality to the operator and remind the operator to perform maintenance in a timely manner; on the contrary, if the result is no, it turns to step S2032 to report a normal result. In this case, the maintenance can be postponed on the basis of the original plan for maintaining the relevant electrical system to save maintenance costs.

[0037] The operating principle of the preferred embodiment of the present invention is described above. Compared with the prior art in which the maintenance plan of the air compressor system is judged solely based on rule-based logic, the present invention classifies the operation data of the corresponding components of the air compressor system differently according to their nature, and respectively adopts a time-series-based prediction algorithm to predict the evolution trend of the operation data, a mean-shift-based clustering algorithm to cluster the operation data still within the acceptable limit value to diagnose abnormal data points, and a rule-based judgment on the regular maintenance plan for different types of operation data, thereby realizing the dynamic adjustment of the original air compressor system maintenance plan. Through the above dynamic adjustment, it is possible to effectively achieve the balance between the too-high and too-low maintenance frequencies of the air compressor system in the prior art, and save maintenance costs without affecting the operating performance of each component in the system at the same time.

[0038] As introduced in the prior art, the predictive maintenance function for the existing air compressor system is not suitable for better control and notification due to the lack of an established knowledge base. Therefore, in addition to dynamically adjusting the maintenance plan for the air compressor system, the present invention also proposes a method for diagnosing abnormal data points in the operation data to achieve efficient maintenance. Referring to Figures 4A to 4D , a schematic block diagram showing the update of the fault database using case-based reasoning is shown. The basic idea of case-based reasoning is to match the current abnormal situation with the problems that have been successfully solved in the past to obtain answers or inspiration. As Figure 4A shown, a knowledge base 100 is constructed using the past fault tracking data and the corresponding problem solutions. The knowledge base 100 includes a fault tracking database 101 and a problem solution database 102. The fault tracking database 101 stores a plurality of historical data entries, and each historical data entry records data vectors u1, u2... u that represent occurred faults or potential fault hazards. n , while the problem solution database 102 stores information representing known faults and their solutions s1, s2... s m , and each historical data entry in the fault tracking database 101 corresponds to one of the information stored in the problem solution database 102. In other words, each historical data entry represents the abnormal data points detected when a corresponding component in the air compressor system fails or has a potential fault hazard, and each historical data entry can find a matching piece of information in the problem solution database 102 to remind the operator of the current fault and countermeasures. From Figure 4A it can be seen that there can be multiple historical data entries corresponding to the information of the same problem solution. For example, in practice, there can be various abnormal detection data indicating that the air filter needs maintenance due to a fault.

[0039] Furthermore, as Figure 4BAs shown, when potential anomalies are detected in the corresponding components of the air compressor system using the mean-shift clustering algorithm as described above, the data entries representing the data points of the above potential anomalies are provided to the fault tracking database 101 for similarity analysis with the existing historical data entries therein, so as to provide several (such as 3, 5, etc.) historical data entries with a similarity exceeding a specific threshold (such as 0.9) and higher than other historical data entries, as well as the corresponding problem solutions to the operator for selection. In this application, the similarity between each historical data entry in the fault tracking database 101 and the data entry representing the potential anomaly data point is compared using cosine similarity, that is, more to distinguish differences in the direction of the data vectors and is insensitive to absolute values. Let the data vectors corresponding to the historical data entries in the fault tracking database 101 be u1, u2... u n , and the data vector corresponding to the data entry representing the potential anomaly data point be v, then the cosine similarity between each data vector is calculated based on the following formula:

[0040]

[0041] where the numerator is the dot product or inner product between the two data vectors, and the denominator is the product of the norms of the two data vectors respectively. For a two-dimensional Euclidean geometric space, the above norm is represented as the Euclidean distance of the data vector from the origin of the space. The range of the cosine value obtained by calculating through the above formula is [-1, 1], where 1 corresponds to completely positive correlation, -1 corresponds to completely negative correlation, and 0 corresponds to completely uncorrelated. It should be noted here that the fault tracking database 101 stores historical data entries representing all types of component failures or potential fault hazards in the air compressor system. Therefore, when performing similarity analysis, it is necessary to select data belonging to the same category as the reported anomaly data point from them. For example, when the reported anomaly data point involves the vibration decibel value of the motor bearing, then all other historical data entries in the fault tracking database 101 that have nothing to do with the above vibration decibel value can be ignored to improve the efficiency of cosine similarity calculation. Take Figure 4B as an example. It is calculated that the data entry representing the above potential anomaly data point is highly similar to the third historical data entry in the fault tracking database 101. Then, after obtaining the relevant prompt, the operator can facilitate the diagnosis of the fault of the corresponding component in the air compressor system and take relevant maintenance measures in a timely manner.

[0042] Furthermore, if the similarity calculated through the above cosine similarity formula does not reach the specified threshold, that is, the data entry representing the above potential anomaly data point is not highly similar to any historical data entry in the fault tracking database 101, then as Figure 4CAt step S30, report the data entry representing the potential abnormal data point to the domain expert for judgment. If the domain expert finally determines that the potential abnormal data point does not indicate a failure or potential failure of the relevant components in the air compressor system, then proceed to step S31 to ignore the data entry and make no updates to the existing knowledge base 100. On the contrary, if the domain expert finally determines that the potential abnormal data point represents a new problem that does not exist in the current problem solution database 102, then proceed to step S32 to add the data entry representing the potential abnormal data point and the problem and solution finally determined by the domain expert to the fault tracking database 101 and the problem solution database 102 respectively for update (as Figure 4D shown, where the data entry representing the potential abnormal data point is represented by the newly added u n+1 and the problem and solution finally determined by the domain expert are represented by the newly added s m+1 ), so as to expand the entire knowledge base 100 for future use. Those skilled in the art can understand that if the domain expert determines that the potential abnormal data point represents an existing problem in the current problem solution database 102 and can be solved using the original solution, then only update the data entry representing the potential abnormal data point to the fault database, and correspond this data entry with the original problem and solution in the problem solution database 102 for future use.

[0043] Corresponding to the method for dynamically adjusting the maintenance plan of the air compressor system in the above embodiment, the present invention also proposes a corresponding unit for dynamically adjusting the maintenance plan of the air compressor system. As Figure 5As shown, the unit 10 includes a data detection module 110 that detects the operating states of various components in the air compressor system to obtain a plurality of air compressor operation data; a data storage module 120 that stores the air compressor operation data from the data detection module 110; a central processing unit 130 that calls the air compressor operation data from the data storage module 120 and executes the foregoing method to obtain a fault diagnosis result; and a monitoring terminal 140. The monitoring terminal 140 obtains the air compressor operation data and the fault diagnosis result from the central processing unit 130 and displays them to the operator and / or domain expert for dynamically adjusting the maintenance plan of the air compressor system. In the present invention, the human-machine interaction interface of the monitoring terminal displays the operation data and monitoring results of various components in the air compressor system in the form of charts or digital quantities, where the charts include but are not limited to curve charts, pie charts, line charts, and the like. The monitoring terminal can be a web page end of a computer or an APP application end of a smart terminal, and all solutions capable of realizing data visualization can be implemented on the monitoring terminal. The unit 10 for dynamically adjusting the maintenance plan of the air compressor system provided in the foregoing embodiment can execute the method provided in any embodiment of the present invention and has the corresponding functions and beneficial effects for executing the method. All the various steps and technical details of the method for dynamically adjusting the maintenance plan of the air compressor system described above can be stored in the unit 10 in the form of software or implemented in the form of a combination of software and hardware. For technical details not described in detail in the foregoing embodiments, reference can be made to the method for dynamically adjusting the maintenance plan of the air compressor system provided in the preferred embodiment of the present invention.

[0044] It should be noted that although in the foregoing embodiment, the prediction of whether the air compressor system needs to be maintained by receiving prediction data and the potential fault diagnosis of abnormal data points are described respectively with the operator and / or domain expert as the objects, the above does not constitute a limitation to the present invention. In practical operations, any person with relevant experience in the art can make a clear and rapid judgment on the maintenance plan of the air compressor system and the possible potential faults based on the prediction results obtained by the time-series-based prediction method and / or the abnormal data points obtained by the clustering algorithm.

[0045] Those skilled in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0046] The above has described in detail, with the aid of the drawings, the feasible but non-limiting embodiments of the method and unit for dynamically adjusting the maintenance plan of the air compressor system according to the present invention. For those of ordinary skill in the art, without departing from the scope and essence of the present disclosure as set forth in the following claims, modifications and supplements to the technology and structure, as well as the recombination of the features in each embodiment, should be regarded as being included within the scope of the present invention. Therefore, these modifications and supplements that can be envisioned under the teachings of the present invention should be regarded as part of the present disclosure. The scope of the present disclosure is defined by the following appended claims and includes equivalent technologies known at the filing date of the present disclosure and equivalent technologies not yet foreseen.

Claims

1. A method for dynamically adjusting the maintenance plan of an air compressor system, comprising the following steps: Detect the operating states of various components in the air compressor system to obtain multiple air compressor operation data; Classify the air compressor operation data based on the nature thereof, where the first type of operation data is data related to pressure detection and / or vibration detection; the second type of operation data is data related to temperature detection; the third type of operation data represents data obtained from inspecting the electrical components of the air compressor system; Among them, for the first type of operation data, use a time-series based prediction method to process it to determine whether maintenance of the air compressor system is required within the prediction time period, and / or adopt clustering analysis to perform fault diagnosis based on the detected potential abnormal results to determine whether maintenance of the air compressor system is required; For the second type of operation data, adopt clustering analysis to perform fault diagnosis based on the detected potential abnormal results to determine whether maintenance of the air compressor system is required; And for the third type of operation data, determine whether maintenance of the air compressor system is required based on rule-based logic.

2. The method according to claim 1, characterized in that, The third type of operation data is the number of starts of the motor in the air compressor system per unit time. If the number of starts of the motor per hour exceeds a threshold, an abnormal result is reported for maintaining the air compressor system; Otherwise, a normal result is output and the number of starts of the motor is continuously monitored.

3. The method according to claim 1 or 2, characterized in that, Use the time-series based prediction method to process the historical data of a certain time window length of the first type of operation data to obtain the prediction data within the prediction time period, and compare the prediction data with the limit value for triggering maintenance of the corresponding components in the air compressor system to determine whether maintenance of the air compressor system is required.

4. The method according to claim 3, characterized in that, If the prediction data exceeds the limit value, report that maintenance of the air compressor system needs to be advanced within the prediction time period; if the prediction data does not exceed the limit value, report that the original planned time for maintaining the air compressor system is postponed, and continuously obtain the real-time first type of operation data within the prediction time period as the updated historical data for subsequent prediction.

5. The method according to claim 1 or 2, characterized in that, Use the mean-shift clustering algorithm to process the first type of operation data and / or the second type of operation data. For the access frequency of each data point, select the category with the largest access frequency as the category to which the current data point belongs, and sequentially assign label values to each category according to the number of data points contained from more to less. Select the category with a label value greater than 2 and further determine whether the abnormal data points therein exceed the limit value for triggering maintenance of the corresponding components in the air compressor system.

6. The method according to claim 5, characterized in that, If the abnormal data points exceed the limit value, report the abnormal data points for fault diagnosis to further determine the cause of the abnormality and whether maintenance of the relevant components of the air compressor system is required; if the abnormal data points do not exceed the limit value, further determine whether the air compressor system is currently in an under-loaded state.

7. The method according to claim 6, characterized in that, If it is determined that the air compressor system is in the incompletely loaded state, a result indicating that the system is normal is output; if it is determined that the air compressor system is not in the incompletely loaded state, the abnormal data points are reported for fault diagnosis to determine the cause of the abnormality and whether maintenance of relevant components of the air compressor system is required.

8. The method according to claim 6 or 7, characterized in that, It also includes improving the knowledge base (100) for maintaining the air compressor system using case-based reasoning logic. The knowledge base (100) includes a fault tracking database (101) and a problem solution database (102). Multiple historical data entries are stored in the fault tracking database (101), and each historical data entry records a data vector representing a occurred fault or a potential fault hidden danger. Information representing known faults and their solutions is stored in the problem solution database (102), and each historical data entry in the fault tracking database (101) corresponds to one piece of information stored in the problem solution database (102).

9. The method according to claim 8, characterized in that, The data entry representing the abnormal data point is provided to the fault tracking database (101) for cosine similarity analysis with the existing historical data entries therein, so as to select a suitable problem solution from multiple historical data entries with a similarity exceeding a specific threshold and higher than other historical data entries and the corresponding problem solutions.

10. The method according to claim 9, characterized in that, If it is determined through cosine similarity analysis that the similarity between the data entry representing the abnormal data point and the existing historical data entries in the fault tracking database (101) does not exceed the specific threshold, the abnormal data point is reported for fault diagnosis. If it is determined that maintenance of the air compressor system is not required, the abnormal data point is ignored and no update is made to the existing knowledge base (100); if it is determined that the abnormal data point represents a new problem that does not exist in the current problem solution database (102), the data entry representing the abnormal data point and the finally judged fault problem and solution are respectively added to the fault tracking database (101) and the problem solution database (102), thereby expanding the knowledge base (100).

11. A unit (10) for dynamically adjusting the maintenance plan of an air compressor system, comprising: A data detection module (110) that detects the operating states of various components in the air compressor system to obtain multiple air compressor operation data; A data storage module (120) that stores the air compressor operation data from the data detection module (110); A central processing unit (130) that calls the air compressor operation data from the data storage module (120) and executes the method according to any one of claims 1 to 10 to obtain a fault diagnosis result; And A monitoring terminal (140) that obtains the air compressor operation data and the fault diagnosis result from the central processing unit (130) and displays them for dynamically adjusting the maintenance plan of the air compressor system.

12. A computer-readable storage medium storing a program, which when executed by a central processing unit enables the central processing unit to execute the method according to any one of claims 1 to 10.

Images