Digital online detection control method and system for compressor cylinder

Abstract

The present application provides a digital online detection control method and system for compressor cylinders, which obtains the candidate operating state events of the compressor cylinders and the prior operating state events associated therewith, and generates abnormal excitation parameters and abnormal influence coefficients according to the operating state vectors of multiple operating state nodes in the prior operating state events, thereby realizing the abnormal probability scoring of the compressor cylinder operating state nodes. Furthermore, by comparing the state vectors of the candidate operating state events with the prior operating state events, and determining the association weights between them in combination with the abnormal influence coefficients, the target association weights that meet the set detection requirements can be screened out from the association weights as the state detection results of the candidate operating state events, thereby improving the accuracy and reliability of the state detection, thereby adaptively controlling the compressor cylinders, effectively improving the operating stability and efficiency of the compressor cylinders, and reducing the risk of failures.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a digital online detection and control method and system for a compressor cylinder. Background Art

[0002] As a core equipment in the industrial field, the stability and efficiency of the compressor's operating status are directly related to the overall performance of the production line and product quality. With the continuous development of industrial automation and intelligent technology, real-time and accurate monitoring and control of the operating status of the compressor cylinder has become the key to improving compressor performance, extending equipment life, and reducing maintenance costs.

[0003] Traditionally, the operating status monitoring of compressor cylinders mainly relies on regular manual inspections, empirical judgments, and offline data analysis. These methods are not only time-consuming and labor-intensive, but also difficult to capture instantaneous changes in the compressor cylinders during operation, resulting in insufficient early warning capabilities for potential failures. In addition, traditional methods often rely on fixed thresholds or rules to judge the operating status, lacking adaptability and flexibility under different operating conditions.

[0004] With the rise of digital technology and big data analysis, online detection and control technologies for compressor cylinders have gradually developed. These technologies collect real-time working data of compressor cylinders, such as pressure, temperature, vibration, etc., and use advanced algorithms to analyze and process these data to achieve real-time monitoring and early warning of the operating status of compressor cylinders.

[0005] However, existing digital online detection and control technologies still have some limitations. On the one hand, these technologies often focus on monitoring a single operating status parameter, while ignoring the correlation and mutual influence between various operating status nodes, resulting in insufficient recognition of complex fault modes. On the other hand, existing technologies usually rely on pre-set models or rules to judge the operating status, lacking adaptability to individual differences between different compressor cylinders and changes in the operating environment. Summary of the invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides a digital online detection and control method for a compressor cylinder, the method comprising:

[0007] Acquire candidate operating state events of the compressor cylinder and each prior operating state event associated with the candidate operating state event, wherein each prior operating state event is: an operating state event that meets preset performance requirements and is generated based on the working data and prior operating records of the compressor cylinder;

[0008] For each of the a priori operating state events, according to the operating state vectors of the multiple operating state nodes included in the a priori operating state event, an abnormal excitation parameter of each operating state node is generated, and according to the abnormal excitation parameter of each operating state node, an abnormal influence coefficient of the corresponding operating state node is determined respectively, wherein the abnormal excitation parameter is used to represent: a probability score that the corresponding operating state node indicates that the compressor cylinder has an abnormal operating condition;

[0009] Determine a plurality of association weights between each priori running state event and the candidate running state event based on the running state vectors of the plurality of running state nodes included in the candidate running state event and the plurality of running state nodes included in the priori running state events, respectively, in combination with the abnormal influence coefficients of the plurality of running state nodes included in the priori running state events;

[0010] Determining, from the association weights, a plurality of target association weights that meet the set detection requirements as the state detection result of the candidate running state event;

[0011] Based on the state detection result of the candidate operating state event, the compressor cylinder is adaptively controlled.

[0012] In a possible implementation manner of the first aspect, generating an abnormal excitation parameter of each of the operating state nodes according to the operating state vectors of the plurality of operating state nodes included in the prior operating state event, and determining an abnormal influence coefficient of a corresponding operating state node respectively according to the abnormal excitation parameter of each of the operating state nodes, includes:

[0013] According to the operation state vectors of the multiple operation state nodes included in the prior operation state event, combined with the event characteristics of the prior operation state event and the prior operation record, an abnormal excitation parameter of each of the operation state nodes is generated;

[0014] Each abnormal excitation parameter is converted into a regularized form to generate an abnormal impact coefficient of each operating state node.

[0015] In a possible implementation manner of the first aspect, the generating of the abnormal excitation parameter of each of the operating state nodes according to the operating state vectors of the plurality of operating state nodes included in the prior operating state event in combination with the event characteristics of the prior operating state event and the prior operating record includes:

[0016] Generate a correlation between each of the operating state nodes and the prior operating state event according to the operating state vectors of the plurality of operating state nodes included in the prior operating state event and the event characteristics of the prior operating state event;

[0017] Determining the abnormal tendency of the priori operating state event according to the priori operating record of the priori operating state event;

[0018] According to each correlation degree and the abnormal tendency degree, an abnormal excitation parameter of each operating state node is determined.

[0019] In a possible implementation manner of the first aspect, the step of determining the abnormal tendency of the priori operating status event based on the priori operating record of the priori operating status event includes:

[0020] In the prior operation record, according to the definition information and event characteristics of the prior operation status event, a data subset directly or indirectly related to the prior operation status event is screened out;

[0021] Performing feature identification on each data item in the screened data subset to generate a data subset with feature identification, wherein the identified features include extreme value features of operating parameters, change rate features of operating parameters, and sequence features of operation events;

[0022] In the data subset with feature identification, searching for an abnormal feature pattern indicating an abnormality, and generating an identified data subset containing the abnormal feature pattern, wherein the abnormal feature pattern is based on a predefined rule or is learned from historical fault data;

[0023] Performing cluster analysis on the identified data subsets containing abnormal feature patterns, dividing the data subsets into different cluster groups according to the similarity of the abnormal feature patterns, generating abnormal data groups clustered by abnormal type, and for each abnormal data group, counting the amount of data contained in the abnormal data group, generating scale statistics results of each abnormal data group, wherein the data amount reflects the scale of each abnormal type appearing in the priori operation record;

[0024] According to the scale statistics of each abnormal data group, the relative abnormal tendency of each abnormal data group is evaluated, and the relative abnormal tendency of each abnormal data group based on the scale is generated;

[0025] Analyze the correlation between each abnormal data group and the operating state factor in the prior operating state event, and adjust the relative abnormal tendency of each abnormal data group based on the scale according to the correlation to generate a corresponding abnormal tendency adjustment value;

[0026] All the abnormal tendency adjustment values ​​are fused and calculated to obtain the abnormal tendency of the priori operating state event, wherein the fusion calculation process adopts a weighted average method.

[0027] In a possible implementation of the first aspect, the determining of multiple association weights between each priori operating state event and the candidate operating state event based on the operating state vectors of the multiple operating state nodes included in the candidate operating state event and the multiple operating state nodes included in the priori operating state events, in combination with abnormal influence coefficients of the multiple operating state nodes included in the priori operating state events, includes:

[0028] Based on the log data corresponding to each of the priori running state events and the candidate running state events, respectively, extracting the running state vectors of the multiple running state nodes included in the priori running state events and the multiple running state nodes included in the candidate running state events, and performing regularization conversion on each running state vector to generate each target running state vector;

[0029] Determining, based on the target operating state vectors of the multiple operating state nodes included in the prior operating state events and the target operating state vectors of the multiple operating state nodes included in the candidate operating state events, the state correlation between the multiple operating state nodes included in the prior operating state events and the multiple operating state nodes included in the candidate operating state events;

[0030] For each of the prior operating status events, multiple association weights between the prior operating status event and the candidate operating status event are determined based on the state correlation between each of the multiple operating status nodes included in the prior operating status event and the multiple operating status nodes included in the candidate operating status event, combined with the abnormal influence coefficients of the multiple operating status nodes included in the prior operating status event.

[0031] In a possible implementation of the first aspect, determining multiple association weights between the priori running state event and the candidate running state event based on the state correlation between each of the multiple running state nodes included in the priori running state event and the multiple running state nodes included in the candidate running state event, combined with abnormal influence coefficients of the multiple running state nodes included in the priori running state event, includes:

[0032] For the multiple operating state nodes included in the candidate operating state event, according to the previously defined selection strategy, determine the target state correlation among the multiple state correlations corresponding to one operating state node, and determine the first association weight between the candidate operating state event and the one prior operating state event according to each target state correlation and the abnormal influence coefficients of the multiple operating state nodes included in the prior operating state event;

[0033] Alternatively, for the multiple operating state nodes included in the prior operating state event, according to the selection strategy, a target state correlation among the multiple state correlations corresponding to one operating state node is determined, and according to each target state correlation and the corresponding abnormal influence coefficient, a second association weight between the candidate operating state event and the one prior operating state event is determined;

[0034] Alternatively, for the multiple operating status nodes included in the candidate operating status event and the prior operating status event, according to the selection strategy, the target state correlation among the multiple state correlations corresponding to an operating status node is determined, and according to each target state correlation and the abnormal impact coefficient of the multiple operating status nodes included in the prior operating status event, the third association weight between the candidate operating status event and the prior operating status event is determined.

[0035] In a possible implementation manner of the first aspect, after determining, from the association weights, a plurality of target association weights that meet the set detection requirement as the status detection result of the candidate running status event, the step further includes:

[0036] Extracting a set number of template running status event combinations from the template running status event sequence;

[0037] According to a plurality of template association weights between two template operation status events included in each template operation status event combination, adjusting the range of the plurality of target association weights;

[0038] The adjusted multiple target association weights are used as the status detection result of the candidate running status event.

[0039] In a possible implementation manner of the first aspect, the range-adjusting the multiple target association weights according to the multiple template association weights between two template operation status events included in each template operation status event combination includes:

[0040] For each template running state event combination, according to the log data respectively corresponding to the two template running state events included in one template running state event combination, the template running state vectors of the multiple running state nodes included in each of the two template running state events are extracted, and according to the template running state vector of each running state node, the multiple template association weights between the two template running state events are determined;

[0041] Determine multiple reference values ​​of a reference range according to each template association weight and the set quantity, wherein the reference value is a threshold value under the reference range, and the reference range is a global range of the template association weights corresponding to each template operation status event combination;

[0042] According to the plurality of reference values ​​and the plurality of target association weights, the adjusted plurality of target association weights are determined.

[0043] In a possible implementation manner of the first aspect, the step of adaptively controlling the compressor cylinder based on the state detection result of the candidate operating state event includes:

[0044] According to the state detection result of the candidate operation state event, the current operation state of the compressor cylinder is evaluated, and preliminary evaluation results of the current operation state of the compressor cylinder in different dimensions are output;

[0045] According to the preliminary evaluation result, the potential problem area existing in the current operating state of the compressor cylinder is determined, and based on the determined potential problem area, a pre-constructed compressor cylinder control strategy knowledge base is queried, and by matching the current potential problem area, a suitable set of candidate control strategies is screened out from the compressor cylinder control strategy knowledge base, wherein the compressor cylinder control strategy knowledge base stores a variety of control strategies for different potential problems, and the control strategies are constructed based on compressor cylinder operating data, prior knowledge, and engineering experience data;

[0046] Performing applicability analysis on the candidate control strategy set, analyzing the applicability of each candidate control strategy in the candidate control strategy set under the current situation, and generating a control strategy subset that has been screened for applicability;

[0047] Prioritize the control strategies according to the subset of control strategies that have been screened for applicability, generate a list of control strategies in order of priority, select the control strategies in order of priority and conduct pre-implementation evaluation, simulate the operation effect after pre-implementation of the control strategies, and generate the evaluation results of the simulated operation effect after pre-implementation of each control strategy;

[0048] According to the evaluation results of the simulated operation effect after the preliminary implementation, the optimal control strategy is selected for actual implementation.

[0049] On the other hand, an embodiment of the present application also provides a digital online detection and control system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0050] Based on the above aspects, the embodiment of the present application obtains the candidate operating state events of the compressor cylinder and the prior operating state events associated therewith, and generates abnormal excitation parameters and abnormal influence coefficients based on the operating state vectors of multiple operating state nodes in the prior operating state events, thereby achieving abnormal probability scoring of the operating state nodes of the compressor cylinder. Furthermore, by comparing the state vectors of the candidate operating state events with those of the prior operating state events, and combining the abnormal influence coefficients, the association weights between them are determined, and the target association weights that meet the set detection requirements can be screened out from the association weights as the state detection results of the candidate operating state events, thereby improving the accuracy and reliability of the state detection. Finally, the compressor cylinder is adaptively controlled based on the state detection results, which effectively improves the operating stability and efficiency of the compressor cylinder and reduces the risk of failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the execution flow of the digital online detection and control method for the compressor cylinder provided in an embodiment of the present application.

[0052] Figure 2 It is a schematic diagram of the hardware architecture of the digital online detection and control system provided in the embodiment of the present application. DETAILED DESCRIPTION

[0053] The present application will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a digital online detection and control method for a compressor cylinder provided by an embodiment of the present application. The digital online detection and control method for a compressor cylinder is introduced in detail below.

[0054] Step S110, obtaining candidate operating status events of the compressor cylinder and each prior operating status event associated with the candidate operating status event, wherein each prior operating status event is: an operating status event that meets preset performance requirements and is generated based on the working data and prior operating records of the compressor cylinder.

[0055] In this embodiment, a compressor cylinder in a large industrial refrigeration system is considered. The working data of the compressor cylinder includes various parameters such as inlet pressure, outlet pressure, cylinder temperature, piston movement speed, workload, etc. The prior operation record is a detailed record of all operating conditions of the compressor cylinder in the past period of time (for example, the past year), including the normal operation period and the record of the minor fault period that has occurred.

[0056] Assume that at a certain moment, the current operating state of the compressor cylinder needs to be evaluated, which is the candidate operating state event. In order to fully evaluate this candidate operating state event, it is necessary to obtain the associated prior operating state events. For example, in the working data, it is found that the current intake pressure is at a specific value, while the outlet pressure fluctuates slightly. By querying the prior operating records, some operating state events that meet the preset performance requirements (such as the refrigeration efficiency is within a certain range, the operating stability reaches a certain standard, etc.) are found. These prior operating state events may include: events in which the outlet pressure has similar fluctuations under similar intake pressures but eventually stabilizes; or operating events with similar temperature change patterns under the same load, etc. The acquisition of these prior operating state events is based on the analysis and screening of a large amount of working data and prior operating records, which can provide a reference for evaluating the current candidate operating state events.

[0057] Step S120, for each of the prior operating status events, based on the operating status vectors of the multiple operating status nodes included in the prior operating status event, generate abnormal excitation parameters for each of the operating status nodes, and based on the abnormal excitation parameters of each of the operating status nodes, respectively determine the abnormal influence coefficient of the corresponding operating status node, the abnormal excitation parameters are used to represent: the probability score that the corresponding operating status node indicates that the compressor cylinder has an abnormal operating condition.

[0058] For each priori operating state event, it is assumed that the operating state nodes included therein include intake pressure node, outlet pressure node, cylinder temperature node, piston motion speed node, etc. The operating state vector is the specific numerical combination of these nodes and their changes over time.

[0059] Taking the intake pressure node as an example, the abnormal excitation parameter is generated based on the operating state vector of the intake pressure node in the prior operating state event, combined with the event characteristics of the prior operating state event (such as the working environment temperature and load conditions at that time) and the prior operating records. Assume that the prior operating records show that under certain specific ambient temperature and load combinations, if the intake pressure deviates from a certain average value by more than a certain range, it may indicate an abnormality. If the operating state vector of the current intake pressure in this prior operating state event shows that it is at the edge of this range that may indicate an abnormality, and the ambient temperature and load conditions at that time also meet the risk combination in the record, then a relatively high abnormal excitation parameter will be generated, indicating that the probability that the intake pressure node indicates an abnormal operating condition of the compressor cylinder is high.

[0060] For the outlet pressure node, if its operating state vector shows that the pressure fluctuates frequently and with a large amplitude, and there are similar fluctuations leading to minor faults in the prior operation records, an abnormal excitation parameter reflecting the possibility of such anomaly will also be generated.

[0061] Then, each abnormal excitation parameter is transformed in a regularized manner to generate an abnormal influence coefficient. For example, all abnormal excitation parameters are transformed according to a certain mathematical formula so that their values ​​are within a specific range (such as between 0 and 1). The converted value is the abnormal influence coefficient. If the abnormal influence coefficient of the abnormal excitation parameter of the intake pressure node is 0.3 after conversion, it means that when considering the entire prior operating state event, the influence of the intake pressure node on the abnormal operating condition of the compressor cylinder is 0.3.

[0062] Step S130, determining multiple association weights between each prior operating state event and the candidate operating state event based on the operating state vectors of the multiple operating state nodes included in the candidate operating state event and the multiple operating state nodes included in the prior operating state events, combined with the abnormal influence coefficients of the multiple operating state nodes included in the prior operating state events.

[0063] Assume that the candidate operating state events include operating state nodes such as intake pressure, outlet pressure, cylinder temperature and piston movement speed, and the prior operating state events also include the same operating state nodes.

[0064] Based on the log data corresponding to each prior operation state event and candidate operation state event (the log data records the detailed information and time series of each operation state node, etc.), the operation state vectors of these operation state nodes in each prior operation state event and the corresponding operation state nodes in the candidate operation state event are extracted, and each operation state vector is converted according to rules to generate each target operation state vector. For example, for the operation state node of intake pressure, its original pressure value is converted into a standardized value in the target operation state vector according to specific rules (such as normalization processing).

[0065] Then, according to the target operating state vectors of the multiple operating state nodes included in each prior operating state event and the target operating state vectors of the multiple operating state nodes included in the candidate operating state event, the state correlation between the multiple operating state nodes included in each prior operating state event and the multiple operating state nodes included in the candidate operating state event is determined. For example, if the target operating state vector of the intake pressure in the prior operating state event is very close to the target operating state vector of the intake pressure in the candidate operating state event, then the state correlation between them is high; if the difference between the two is large, the state correlation is low.

[0066] For each prior operation state event, multiple association weights between the prior operation state event and the candidate operation state event are determined based on the state correlation between the multiple operation state nodes included in the prior operation state event and the multiple operation state nodes included in the candidate operation state event, combined with the abnormal influence coefficients of the multiple operation state nodes included in the prior operation state event. For example, for a prior operation state event, the state correlation between the intake pressure node and the intake pressure node in the candidate operation state event is 0.8, and the abnormal influence coefficient of the intake pressure node is 0.3; the state correlation between the outlet pressure node and the outlet pressure node in the candidate operation state event is 0.6, and the abnormal influence coefficient of the outlet pressure node is 0.2. According to a specific calculation method (such as weighted summation, etc.), an association weight between this prior operation state event and the candidate operation state event can be calculated. Since multiple operation state nodes are involved in the calculation, multiple association weights will be obtained.

[0067] Step S140, determining a plurality of target association weights that meet the set detection requirements from among the association weights as the status detection result of the candidate running status event.

[0068] The detection requirements may be pre-set based on safety standards, performance requirements, etc. of the compressor cylinder. For example, in an industrial refrigeration system, in order to ensure the stable operation and refrigeration effect of the compressor cylinder, some threshold ranges of associated weights are set.

[0069] Assume that a series of correlation weights between prior operating state events and candidate operating state events are obtained. If the correlation weight represents the degree of influence or similarity of the prior operating state event on the candidate operating state event, it will be screened according to the set threshold range. For example, the set requirement is that the correlation weight is between 0.5 and 0.8 to be considered to meet the detection requirements. Among the many correlation weights, those within this range are determined as target correlation weights that meet the set detection requirements. These target correlation weights become the status detection results of the candidate operating state events. If no correlation weight meets this range, it may mean that there is a large difference between the current candidate operating state event and the prior operating state event, which requires further in-depth analysis or may indicate potential abnormal conditions.

[0070] Step S150: adaptively controlling the compressor cylinder based on the state detection result of the candidate operating state event.

[0071] In this embodiment, the current operating state of the compressor cylinder is evaluated based on the state detection result of the candidate operating state event (i.e., the target association weight), and the preliminary evaluation results of the current operating state of the compressor cylinder in different dimensions are output. For example, if the target association weight indicates that the candidate operating state event is highly similar to some normal prior operating state events, the preliminary evaluation result may be that the current operating state is relatively normal, but may need to be fine-tuned in some aspects. If the target association weight is close to the prior operating state event related to the abnormality, the preliminary evaluation result may be that there is a potential risk.

[0072] Based on the preliminary evaluation results, potential problem areas that exist in the current operating state of the compressor cylinder are determined. For example, if the preliminary evaluation results show that there may be problems with air pressure fluctuations, then the outlet pressure is a potential problem area. Based on the identified potential problem areas, query the pre-built compressor cylinder control strategy knowledge base. This knowledge base stores a variety of control strategies for different potential problems. These strategies are built based on compressor cylinder operating data, prior knowledge, and engineering experience data. For example, for the outlet pressure fluctuation problem, the knowledge base may have control strategies such as adjusting the intake valve opening and checking the piston seal.

[0073] Perform applicability analysis on the candidate control strategy set, and analyze the applicability of each candidate control strategy in the current situation. For example, under the current working environment temperature, load and other conditions, some control strategies may not be applicable or the effect is not good. After applicability screening, a subset of control strategies is obtained. Then, the control strategies are prioritized according to the subset of control strategies that have been screened for applicability, and a list of control strategies in priority order is generated. The control strategies are selected in order of priority and pre-implementation evaluation is performed, and the operating effect after pre-implementation of the control strategy is simulated to generate the simulated operating effect evaluation results after pre-implementation of each control strategy. For example, simulate the change in outlet pressure and the change in overall efficiency of the compressor after adjusting the opening of the intake valve. Finally, according to the evaluation results of the simulated operating effect after pre-implementation, the optimal control strategy is selected for actual implementation to achieve adaptive control of the compressor cylinder and ensure its stable and efficient operation.

[0074] Based on the above steps, the embodiment of the present application obtains the candidate operating state events of the compressor cylinder and the prior operating state events associated therewith, and generates abnormal excitation parameters and abnormal influence coefficients according to the operating state vectors of multiple operating state nodes in the prior operating state events, thereby achieving abnormal probability scoring of the operating state nodes of the compressor cylinder. Furthermore, by comparing the state vectors of the candidate operating state events with those of the prior operating state events, and combining the abnormal influence coefficients, the association weights between them are determined, and the target association weights that meet the set detection requirements can be screened out from the association weights as the state detection results of the candidate operating state events, thereby improving the accuracy and reliability of the state detection. Finally, the compressor cylinder is adaptively controlled based on the state detection results, which effectively improves the operating stability and efficiency of the compressor cylinder and reduces the risk of failures.

[0075] In a possible implementation, step S120 includes:

[0076] Step S121 , generating abnormal excitation parameters of each of the operating status nodes according to the operating status vectors of the multiple operating status nodes included in the prior operating status event, combined with the event features of the prior operating status event and the prior operating records.

[0077] Step S122, performing regularization conversion on each abnormal excitation parameter to generate an abnormal influence coefficient of each operating state node.

[0078] In a possible implementation, step S121 includes:

[0079] Step S1211 , generating a correlation between each of the operating status nodes and the prior operating status event according to the operating status vectors of the plurality of operating status nodes included in the prior operating status event and the event features of the prior operating status event.

[0080] Step S1212: determining the abnormal tendency of the priori operating status event according to the priori operating record of the priori operating status event.

[0081] Step S1213, determining the abnormal excitation parameter of each of the operating state nodes according to each correlation and the abnormal tendency.

[0082] In a possible implementation, step S1212 includes:

[0083] Step S1212-1, in the priori operation record, according to the definition information and event characteristics of the priori operation status event, filter out a data subset directly or indirectly related to the priori operation status event.

[0084] Step S1212-2, feature identification is performed on each data item in the screened data subset to generate a data subset with feature identification, wherein the identified features include extreme value features of operating parameters, rate of change features of operating parameters, and sequence features of operation events.

[0085] Step S1212-3, searching for an abnormal feature pattern indicating an abnormality in the data subset with feature identification, and generating an identified data subset containing the abnormal feature pattern, wherein the abnormal feature pattern is based on a predefined rule or is learned from historical fault data.

[0086] Step S1212-4, performing cluster analysis on the identified data subsets containing abnormal feature patterns, dividing the data subsets into different cluster groups according to the similarity of the abnormal feature patterns, generating abnormal data groups clustered by abnormal type, and for each abnormal data group, counting the amount of data contained in the abnormal data group, generating scale statistics results for each abnormal data group, the data amount reflects the scale of each abnormal type in the prior operation record.

[0087] Step S1212-5: evaluate the relative abnormal tendency of each abnormal data group according to the scale statistics of each abnormal data group, and generate the relative abnormal tendency of each abnormal data group based on the scale.

[0088] Step S1212-6, analyzing the correlation between each abnormal data group and the operating status factor in the prior operating status event, and adjusting the relative abnormal tendency of each abnormal data group based on the scale according to the correlation, to generate a corresponding abnormal tendency adjustment value.

[0089] Step S1212-7, all the abnormal tendency adjustment values ​​are fused and calculated to obtain the abnormal tendency of the prior operating state event, wherein the fusion calculation process adopts a weighted average method.

[0090] In this embodiment, for each operation state node in the prior operation state event, such as the intake pressure node, the outlet pressure node, the cylinder temperature node, and the piston movement speed node, the operation state vector includes information such as the value change of these nodes in a specific time period. First, based on the operation state vectors of these operation state nodes in the prior operation state event, combined with the event characteristics of the prior operation state event and the prior operation record, the abnormal excitation parameters of each operation state node are generated.

[0091] First, let's look at how to generate the correlation between each operating state node and the prior operating state event based on the operating state vector and event characteristics. Taking the aforementioned intake pressure node as an example, the event characteristics of the prior operating state event include information such as the workload and ambient temperature at the time. Assume that the numerical change of the intake pressure node in the operating state vector is that the pressure value fluctuates within a relatively narrow range when the workload is stable. The event characteristics of the prior operating state event indicate that the ambient temperature at the time was low. Through the analysis of historical data, it can be seen that under this ambient temperature, the normal fluctuation range of the intake pressure should be narrower. Then, based on the comparative analysis of the operating state vector of the intake pressure node and this event characteristic, the correlation between the intake pressure node and the prior operating state event can be obtained. If the pressure fluctuation range deviates greatly from the normal fluctuation range expected at the ambient temperature, the correlation is low; if it is close to the expected range, the correlation is high. Similarly, for the outlet pressure node, if its fluctuation frequency in the operating state vector matches the workload change law in the event characteristic to a high degree, the correlation is high; otherwise, it is low.

[0092] Then, the abnormal tendency of the prior operating state event is determined based on the prior operating record of the prior operating state event. In the prior operating record of the compressor cylinder, according to the definition information of the prior operating state event (such as a specific operating mode, a specific working time period, etc.) and the event characteristics (such as load type, environmental conditions, etc.), a data subset directly or indirectly related to the prior operating state event is screened. For example, if the prior operating state event is an operation under high load, all operating data under high load are screened. Each data item in the screened data subset is feature-identified to generate a data subset with feature identification, wherein the identified features include extreme value features of operating parameters, rate of change features of operating parameters, and sequential features of operating events. For example, for the intake pressure data item, its maximum value, minimum value, and pressure change rate are marked, and for the operating event, the sequential features of operations such as starting, stopping, and adjusting valves are marked. In the data subset with feature identification, find the abnormal feature pattern representing the abnormality. These abnormal feature patterns may be based on predefined rules (such as the sudden drop of intake pressure exceeding a certain proportion is considered abnormal) or learned from historical fault data (abnormal pressure change pattern summarized by analyzing a large amount of historical fault data). After generating the identified data subset containing abnormal feature patterns, cluster analysis is performed on it. According to the similarity of the abnormal feature patterns, the data subsets are divided into different cluster groups to generate abnormal data groups clustered by abnormal type. For example, the abnormal data related to the sudden drop of intake pressure is classified into one category, and the data related to the abnormal fluctuation of outlet pressure is classified into another category. For each abnormal data group, the amount of data contained in the abnormal data group is counted to generate the scale statistics of each abnormal data group. This data amount reflects the scale of each abnormal type in the prior operation record. If the amount of data of the abnormal type of sudden drop of intake pressure is large, it means that its frequency of occurrence is relatively high. According to the scale statistics of each abnormal data group, the relative abnormal tendency of each abnormal data group is evaluated, and the relative abnormal tendency of each abnormal data group based on scale is generated. For example, if the size of the abnormal data group with a sudden drop in intake pressure is large, its relative abnormal tendency is higher. Then analyze the correlation between each abnormal data group and the operating state factors in the prior operating state event. If the abnormal data group with a sudden drop in intake pressure is directly related to the intake pressure node in the prior operating state event, adjust the relative abnormal tendency of each abnormal data group based on the size to generate the corresponding abnormal tendency adjustment value. For example, if the sudden drop in intake pressure is closely related to the intake pressure node in the prior operating state event, appropriately increase its abnormal tendency adjustment value. Finally, all abnormal tendency adjustment values ​​are fused and calculated to obtain the abnormal tendency of the prior operating state event. Here, the weighted average method is used to assign different weights according to the importance of each abnormal tendency adjustment value for calculation.

[0093] Finally, the abnormal excitation parameters of each operating state node are determined based on the correlations and abnormal tendency. For example, for the intake pressure node, its correlation with the prior operating state event is low, and the abnormal tendency of the prior operating state event is high. Then, according to a specific calculation method (such as the product of the correlation and the abnormal tendency or other functional relationships based on experience and data analysis), the abnormal excitation parameters of the intake pressure node are generated. This abnormal excitation parameter represents the probability score that the intake pressure node indicates that the compressor cylinder has an abnormal operating condition. For other operating state nodes, such as the exhaust pressure node, the cylinder temperature node, and the piston movement speed node, the respective abnormal excitation parameters are determined in the same way by combining their respective correlations and the abnormal tendency of the prior operating state event.

[0094] After obtaining the abnormal excitation parameters of each operating state node, each abnormal excitation parameter is converted according to a rule to generate the abnormal influence coefficient of each operating state node. For example, a conversion rule is set to convert the value range of the abnormal excitation parameter to between 0 and 1 through a mathematical formula. Assuming that the abnormal excitation parameter of the intake pressure node is a certain value, the result obtained after a specific regular conversion (such as dividing by a specific value or through a linear transformation function) is between 0 and 1. This result is the abnormal influence coefficient of the intake pressure node. Similarly, similar regular conversions are performed on the abnormal excitation parameters of other operating state nodes to obtain their respective abnormal influence coefficients. These abnormal influence coefficients can accurately reflect the degree of influence of each operating state node on the abnormal operating condition of the compressor cylinder in the entire prior operating state event.

[0095] In a possible implementation, step S130 includes:

[0096] Step S131, based on the log data corresponding to each of the prior operating status events and the candidate operating status events, respectively, extract the operating status vectors of the multiple operating status nodes included in the prior operating status events and the multiple operating status nodes included in the candidate operating status events, and perform regularization conversion on each operating status vector to generate each target operating status vector.

[0097] Step S132, based on the target operating state vectors of the multiple operating state nodes included in the prior operating state events and the target operating state vectors of the multiple operating state nodes included in the candidate operating state events, determine the state correlation between the multiple operating state nodes included in the prior operating state events and the multiple operating state nodes included in the candidate operating state events.

[0098] Step S133, for each of the prior operating status events, determine multiple association weights between the prior operating status event and the candidate operating status event based on the state correlation between the multiple operating status nodes included in the prior operating status event and the multiple operating status nodes included in the candidate operating status event, combined with the abnormal influence coefficients of the multiple operating status nodes included in the prior operating status event.

[0099] In a possible implementation, step S133 includes:

[0100] Step S1331, for the multiple operating status nodes included in the candidate operating status event, determine the target state correlation among the multiple state correlations corresponding to an operating status node according to a previously defined selection strategy, and determine the first association weight between the candidate operating status event and the prior operating status event according to each target state correlation and the abnormal influence coefficient of the multiple operating status nodes included in the prior operating status event.

[0101] Alternatively, in step S1332, for the multiple operating status nodes included in the prior operating status event, determine the target state correlation among the multiple state correlations corresponding to an operating status node based on the selection strategy, and determine the second association weight between the candidate operating status event and the prior operating status event based on each target state correlation and the corresponding abnormal impact coefficient.

[0102] Alternatively, in step S1333, for the multiple operating status nodes included in the candidate operating status event and the prior operating status event, determine the target state correlation among the multiple state correlations corresponding to an operating status node according to the selection strategy, and determine the third association weight between the candidate operating status event and the prior operating status event according to each target state correlation and the abnormal impact coefficient of the multiple operating status nodes included in the prior operating status event.

[0103] In this embodiment, first, based on the log data corresponding to each prior operation state event and the candidate operation state event, the operation state vectors of multiple operation state nodes included in each prior operation state event and multiple operation state nodes included in the candidate operation state event are extracted, and each operation state vector is converted into a regularity to generate each target operation state vector. Taking the compressor cylinder in the industrial refrigeration system as an example, the log data of the prior operation state event and the candidate operation state event contain a large amount of detailed information about the operation of the compressor cylinder. The operation state nodes in the log data include an intake pressure node, an outlet pressure node, a cylinder temperature node, a piston motion speed node, etc. For the intake pressure node in the prior operation state event, its operation state vector may include an intake pressure value sequence at different time points. Similarly, there are corresponding operation state vectors for the outlet pressure node, the cylinder temperature node, the piston motion speed node, etc. For the candidate operation state event, these operation state nodes also have corresponding operation state vectors. Then these operation state vectors are converted into a regularity to generate the target operation state vector. For example, the operating state vector of the intake pressure node may be normalized to convert its value range to a specific interval, such as between 0 and 1, for subsequent calculation and comparison. Similar regularization conversion methods are also used for the operating state vectors of other operating state nodes to obtain each target operating state vector.

[0104] Next, according to the target operating state vectors of the multiple operating state nodes included in each prior operating state event and the target operating state vectors of the multiple operating state nodes included in the candidate operating state event, the state correlation between the multiple operating state nodes included in each prior operating state event and the multiple operating state nodes included in the candidate operating state event is determined. Taking the intake pressure node as an example, the intake pressure target operating state vector in the prior operating state event and the intake pressure target operating state vector in the candidate operating state event are compared. If the values ​​in the two vectors are very close at the same time point or a similar time point, and the change trend of the values ​​is also similar, then it can be considered that the state correlation between the intake pressure node in the prior operating state event and the intake pressure node in the candidate operating state event is high. On the contrary, if the value difference is large and the change trend is different, the state correlation is low. For the outlet pressure node, cylinder temperature node, piston motion speed node, etc., their respective state correlations are determined in the same way. For example, when comparing the cylinder temperature nodes in the prior operating state event and the candidate operating state event, if the cylinder temperature target operating state vector in the prior operating state event shows that the temperature fluctuates within a stable range, while the cylinder temperature target operating state vector in the candidate operating state event shows that the temperature has a tendency to suddenly increase, then the state correlation between the two is low.

[0105] Then, for each of the priori running state events, multiple association weights between the priori running state event and the candidate running state event are determined based on the state correlation between the multiple running state nodes included in the priori running state event and the multiple running state nodes included in the candidate running state event, combined with the abnormal influence coefficients of the multiple running state nodes included in the priori running state event. There are multiple ways to determine the association weights in this process.

[0106] The first method is to determine the target state correlation among the multiple state correlations corresponding to a running state node according to the previously defined selection strategy for the multiple running state nodes included in the candidate running state event, and determine the first association weight between the candidate running state event and a priori running state event according to the abnormal influence coefficients of the multiple running state nodes included in the priori running state event. For example, for the intake pressure node in the candidate running state event, it has its own state correlation with the intake pressure nodes in the multiple priori running state events. Assuming that there are three priori running state events, the state correlation corresponding to one of the priori running state events is determined as the target state correlation according to the selection strategy (such as selecting the state correlation corresponding to the priori running state event that is most similar to the current running environment). At the same time, the abnormal influence coefficients of the intake pressure node, the outlet pressure node, the cylinder temperature node, and the piston movement speed node in this priori running state event are known. The first association weight between the candidate running state event and the priori running state event is determined by a specific calculation method (such as weighted summation of the target state correlation and each abnormal influence coefficient).

[0107] The second method is to determine the target state correlation among the multiple state correlations corresponding to an operation state node according to the selection strategy for the multiple operation state nodes included in the prior operation state event, and determine the second association weight between the candidate operation state event and the prior operation state event according to the correlation of each target state and the corresponding abnormal influence coefficient. For example, for the outlet pressure node in the prior operation state event, the target state correlation corresponding to the outlet pressure node in the candidate operation state event is determined according to the selection strategy. At the same time, the abnormal influence coefficient of the outlet pressure node in the prior operation state event is considered, and other operation state nodes (such as the intake pressure node, the cylinder temperature node, and the piston movement speed node, etc.) are determined according to their respective determined target state correlations and abnormal influence coefficients, and the second association weight between the candidate operation state event and the prior operation state event is determined by a specific calculation method (such as summing these products or other complex functional relationships).

[0108] The third method is to determine the target state relevance among the multiple state relevances corresponding to one operation state node according to the selection strategy for the multiple operation state nodes included in each of the candidate operation state event and the prior operation state event, and determine the third association weight between the candidate operation state event and the prior operation state event according to the abnormal influence coefficients of the multiple operation state nodes included in the prior operation state event. For example, for the piston movement speed node in the candidate operation state event and the prior operation state event, its target state relevance is determined according to the selection strategy. At the same time, considering the abnormal influence coefficients of the intake pressure node, the outlet pressure node, the cylinder temperature node and the piston movement speed node in the prior operation state event, the third association weight between the candidate operation state event and the prior operation state event is determined according to a predetermined calculation method (such as comprehensively considering the importance of each operation state node, weighted summing the product of the target state relevance and the abnormal influence coefficient, etc.). Through these three methods, the association weight between the prior operation state event and the candidate operation state event can be determined comprehensively and meticulously, thereby providing an important basis for the subsequent state evaluation and control of the compressor cylinder.

[0109] In a possible implementation manner, after step S140, the method further includes:

[0110] Step A110: extracting a set number of template running status event combinations from the template running status event sequence.

[0111] Step A120: adjusting the range of the multiple target association weights according to the multiple template association weights between two template operation status events included in each template operation status event combination.

[0112] Step A130: Using the adjusted multiple target association weights as the status detection result of the candidate running status event.

[0113] In a possible implementation, step A120 includes:

[0114] Step A121, for each template running status event combination, based on the log data respectively corresponding to the two template running status events included in a template running status event combination, extract the template running status vectors of the multiple running status nodes included in each of the two template running status events, and based on the template running status vector of each running status node, determine the multiple template association weights between the two template running status events.

[0115] Step A122, determining multiple reference values ​​of the benchmark range based on the associated weights of each template and the set quantity, wherein the reference value is a threshold value under the benchmark range, and the benchmark range is a global range of the template associated weights corresponding to the combinations of each template operating status event.

[0116] Step A123, determining the adjusted multiple target association weights based on the multiple reference values ​​and the multiple target association weights.

[0117] In this embodiment, after determining multiple target association weights that meet the set detection requirements from each association weight and using them as the state detection results of the candidate operation state events, a set number of template operation state event combinations are extracted from the template operation state event sequence. The template operation state event sequence is a collection of a series of pre-constructed operation state events. These events are constructed based on a large amount of compressor cylinder operation data, prior knowledge, and engineering experience data. They cover the operation state of the compressor cylinder under various typical working conditions. Assuming the set number is 5, 5 template operation state event combinations are selected from this template operation state event sequence. Each template operation state event combination contains two template operation state events, and these two template operation state events are selected from a large number of template operation state events according to specific rules, for example, they may be selected according to the similarity of the operating conditions or according to the order of operation time.

[0118] Then, according to the multiple template association weights between the two template operation state events included in each template operation state event combination, the multiple target association weights are range-adjusted. For each template operation state event combination, the template operation state vectors of the multiple operation state nodes included in each of the two template operation state events are extracted according to the log data corresponding to the two template operation state events included in a template operation state event combination. Taking the compressor cylinder as an example, the operation state node includes the intake pressure node, the outlet pressure node, the cylinder temperature node and the piston movement speed node. For the intake pressure node in each template operation state event, its log data contains the intake pressure value, pressure change rate and other information at different time points, which constitutes the template operation state vector of the intake pressure node. Similarly, there are corresponding template operation state vectors for the outlet pressure node, the cylinder temperature node and the piston movement speed node. According to the template operation state vector of each operation state node, the multiple template association weights between the two template operation state events are determined. For example, for the intake pressure node, by comparing the template operation state vectors of the intake pressure node in the two template operation state events, a specific calculation method (such as calculating the distance between vectors or similarity function, etc.) is used to determine the template association weight corresponding to the intake pressure node. Similar methods are also used to determine the template association weights of other operation state nodes.

[0119] According to the associated weights of each template and the set quantity, multiple reference values ​​of the benchmark range are determined, where the reference value is the threshold value under the benchmark range, and the benchmark range is the global range of the template associated weights corresponding to each template running state event combination. Assume that there are 5 template running state event combinations, and each combination has template associated weights corresponding to multiple running state nodes. Putting all these template associated weights together constitutes a global range. For example, the value range of these template associated weights may be between 0.1 and 0.9, which is the benchmark range. According to the associated weights of each template and the set quantity, a specific statistical method is used to determine the reference value. One possible method is to sort all template associated weights by size, and then select the template associated weight value at a specific position as the reference value according to the set quantity (5). For example, the minimum value, the maximum value, and the middle value (the third value) can be selected as reference values, and these reference values ​​become the threshold values ​​under the benchmark range.

[0120] Finally, based on the multiple reference values ​​and the multiple target association weights, the adjusted multiple target association weights are determined. Assume that three reference values ​​are obtained, which are 0.2, 0.6 and 0.8, respectively, and there are multiple target association weights, for example, the values ​​of the target association weights are 0.3, 0.5 and 0.7, respectively. If the target association weight 0.3 is less than the reference value 0.6, it may be adjusted according to a pre-set adjustment rule (such as adjusting according to a certain proportion). Assuming that the adjustment rule is that when the target association weight is less than the middle reference value (0.6), it is increased by 0.1, then the adjusted target association weight becomes 0.4. For the target association weight 0.5, since it is between the reference values ​​0.2 and 0.6, according to another adjustment rule (such as keeping unchanged), the adjusted target association weight is still 0.5. For the target association weight 0.7, since it is greater than the reference value 0.6, according to another adjustment rule (such as reducing by 0.05), the adjusted target association weight becomes 0.65. In this way, multiple target association weights are adjusted according to multiple reference values, and the adjusted multiple target association weights are used as the final state detection results of the candidate operating state events. Such adjusted state detection results can more accurately reflect the operating state of the compressor cylinder, and provide a more reliable basis for the subsequent adaptive control of the compressor cylinder based on the state detection results.

[0121] Throughout the process, the target association weight is adjusted based on the information in the template operation status event sequence in order to optimize the evaluation of candidate operation status events by integrating more prior knowledge and typical operation conditions. The template association weight is determined by extracting information from the template operation status event combination, and then the reference value is determined based on these template association weights. Finally, the target association weight is adjusted based on the reference value. Each step follows strict technical logic and calculation methods, aiming to improve the accuracy and reliability of compressor cylinder operation status detection, so as to achieve more accurate and effective control and management of compressor cylinders.

[0122] In a possible implementation, step S150 includes:

[0123] Step S151, evaluating the current operating state of the compressor cylinder according to the state detection result of the candidate operating state event, and outputting preliminary evaluation results of the current operating state of the compressor cylinder in different dimensions.

[0124] Step S152: Based on the preliminary evaluation result, determine the potential problem areas existing in the current operating state of the compressor cylinder, and based on the determined potential problem areas, query the pre-constructed compressor cylinder control strategy knowledge base, and screen out a set of applicable candidate control strategies from the compressor cylinder control strategy knowledge base by matching the current potential problem areas. The compressor cylinder control strategy knowledge base stores a variety of control strategies for different potential problems, and the control strategies are constructed based on compressor cylinder operating data, prior knowledge and engineering experience data.

[0125] Step S153 , performing applicability analysis on the candidate control strategy set, analyzing the applicability of each candidate control strategy in the candidate control strategy set under the current situation, and generating a control strategy subset that has been screened for applicability.

[0126] Step S154, prioritize the control strategies according to the control strategy subsets that have been screened for applicability, generate a list of control strategies in order of priority, select the control strategies in order of priority and perform a pre-implementation evaluation, simulate the operating effect after pre-implementation of the control strategy, and generate a simulated operating effect evaluation result after pre-implementation of each control strategy.

[0127] Step S155, selecting the optimal control strategy for actual implementation according to the evaluation results of the simulated operation effect after the preliminary implementation.

[0128] In this embodiment, the current operating state of the compressor cylinder is evaluated according to the state detection results of the candidate operating state events, and the preliminary evaluation results of the current operating state of the compressor cylinder in different dimensions are output. The state detection results contain a series of information related to the operating state of the compressor cylinder, which reflects important contents such as the degree of association between the current operating state and the previous prior operating state events. Taking the operating state nodes such as the intake pressure, outlet pressure, cylinder temperature and piston movement speed of the compressor cylinder as an example, if the state detection results show that the associated weight related to the intake pressure is low and there is a large deviation from the normal prior operating state event, this may mean that there is an abnormality in the intake pressure. For the outlet pressure, if the state detection results show that its associated weight is within the normal range, but the fluctuation frequency has increased, this is also a point that needs attention. From the perspective of cylinder temperature, if its associated weight shows that the temperature value deviates from the temperature range corresponding to the normal prior operating state event, this may indicate that there is a problem in the cooling or heating link. As for the piston movement speed, if the associated weight shows that the speed is unstable or lower than the value during normal operation, it may indicate that the piston assembly has wear or other mechanical problems. Based on the analysis of these state detection results, preliminary evaluation results are output in different dimensions. For example, in the intake pressure dimension, the preliminary evaluation result may be "the intake pressure is unstable and lower than the normal range, and there may be a blockage in the intake pipe or a failure in the intake valve"; in the outlet pressure dimension, it may be "the frequency of outlet pressure fluctuations increases, which may be a change in the compressor load or an exhaust valve failure"; in the cylinder temperature dimension, it is "the temperature is too high, which may be a failure in the cooling system or increased internal friction of the compressor"; in the piston movement speed dimension, it is "the speed is unstable and low, which may be caused by piston wear or insufficient lubrication."

[0129] According to the preliminary evaluation results, the potential problem areas existing in the current operating state of the compressor cylinder are determined, and based on the determined potential problem areas, the pre-built compressor cylinder control strategy knowledge base is queried, and by matching the current potential problem areas, a set of applicable candidate control strategies is screened out from the compressor cylinder control strategy knowledge base. The compressor cylinder control strategy knowledge base is a knowledge treasure house built based on a large amount of compressor cylinder operation data, prior knowledge, and engineering experience data, which stores a variety of control strategies for different potential problems. For example, if the preliminary evaluation results determine that the unstable intake pressure is a potential problem area, then the control strategy related to the intake pressure is searched in the knowledge base. Possible strategies include checking whether the intake pipe is blocked, testing the sealing and opening adjustment function of the intake valve, etc. For the potential problem area of ​​increased frequency of outlet pressure fluctuations, the strategies in the knowledge base may involve checking whether the compressor load is stable, the working state of the exhaust valve, and whether the exhaust pressure needs to be adjusted. When the cylinder temperature is high and is determined as a potential problem area, the control strategies in the knowledge base may include checking the heat dissipation system (such as whether the radiator is blocked, whether the cooling fan is working properly), detecting the lubrication condition inside the compressor to reduce friction, etc. If the piston movement speed is unstable and low, which is a potential problem area, the strategies in the knowledge base may include checking the wear of the piston assembly, whether the lubrication system is supplying oil normally, etc. Through this matching process, a set of candidate control strategies suitable for the current potential problem area is screened out.

[0130] The applicability of the candidate control strategy set is analyzed, and the applicability of each candidate control strategy in the candidate control strategy set under the current situation is analyzed to generate a subset of control strategies that have been screened for applicability. For each candidate control strategy, the actual operating conditions of the current compressor cylinder need to be considered, such as the current workload, ambient temperature, operating time and other factors. Taking the candidate control strategy of checking whether the intake pipe is blocked as an example, if the current workload is low and the deviation of the intake pressure is small, the applicability of this strategy may be low, because a small deviation of the intake pressure under low load may be a normal fluctuation range. For the strategy of testing the sealing of the intake valve, if the intake pressure deviation is large and the intake valve has not been maintained for a long time, then the applicability of this strategy is high in the current situation. For strategies related to the outlet pressure, such as checking whether the compressor load is stable, if the current compressor operating load fluctuates frequently, then the applicability of this strategy is high; but if the load has been relatively stable, the applicability of this strategy is low. For the cooling system inspection strategy, if the cylinder temperature is high and the ambient temperature is also high, then the strategy of checking whether the radiator is blocked and whether the cooling fan is working properly is very suitable; if the ambient temperature is normal and the cylinder temperature is high, it may be necessary to check the lubrication inside the compressor more deeply to reduce friction. At this time, the applicability of this cooling system inspection strategy is relatively low. By analyzing the applicability of each candidate control strategy in the current situation, the control strategies with high applicability are screened out, and a control strategy subset that has been screened for applicability is generated.

[0131] According to the control strategy subsets that have been screened for applicability, the control strategies are prioritized, a list of control strategies in priority order is generated, and the control strategies are selected in order of priority and pre-implemented for evaluation, the operation effect after pre-implementation of the control strategy is simulated, and the evaluation results of the simulated operation effect after pre-implementation of each control strategy are generated. When prioritizing the control strategy subsets, multiple factors need to be considered comprehensively. For example, strategies that directly affect the safe operation of the compressor should be given a higher priority. For example, when the piston movement speed is unstable and low, which may cause serious mechanical failure, the strategy of checking the wear of the piston assembly should be ranked higher. If the implementation cost of a strategy is low and there is a greater possibility of solving the current problem, it should also be given a higher priority. For example, checking the sealing of the intake valve is relatively low in cost and easier to operate than checking whether the entire intake pipe is blocked. If the intake pressure deviation is large, then the strategy of checking the sealing of the intake valve should be ranked higher. After selecting the control strategies in order of priority, a pre-implementation evaluation is performed. For example, for the strategy of checking the tightness of the intake valve, by establishing an operating model of the compressor cylinder, the operating effect after the intake valve tightness is improved is simulated, including the simulation of changes in intake pressure and improvement in overall operating efficiency. For the strategy of checking the cooling system, the change in cylinder temperature after improving the cooling conditions (such as cleaning the radiator and repairing the cooling fan) and the impact on the overall performance of the compressor are simulated. Through these simulations, the simulated operating effect evaluation results after the pre-implementation of each control strategy are generated.

[0132] According to the simulation operation effect evaluation results after pre-implementation, the optimal control strategy is selected for actual implementation. Compare the simulation operation effect evaluation results after pre-implementation of each control strategy, and select the control strategy that can solve the current potential problems while improving the overall operating performance of the compressor cylinder and having the least negative impact as the optimal control strategy. For example, if the simulation operation effect evaluation results after pre-implementation of the strategy for checking the sealing of the intake valve show that the intake pressure can be restored to the normal range and the overall efficiency of the compressor is significantly improved, and the strategy of checking whether the intake pipe is blocked can also improve the intake pressure, but may bring higher costs and longer downtime, then the strategy of checking the sealing of the intake valve is the optimal control strategy. This optimal control strategy is actually implemented to achieve adaptive control of the compressor cylinder, ensure that the compressor cylinder operates in the best state, and improve its operating efficiency, reliability and safety.

[0133] Figure 2 The hardware structure of the digital online detection control system 100 provided in the embodiment of the present application is shown as follows: Figure 2 As shown, the digital online detection and control system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0134] In one possible design, the digital online detection and control system 100 can be a single server or a server group. The server group can be centralized or distributed (for example, the digital online detection and control system 100 can be a distributed system). In some embodiments, the digital online detection and control system 100 can be local or remote. For example, the digital online detection and control system 100 can access information and / or data stored in a machine-readable storage medium 120 via a network. For another example, the digital online detection and control system 100 can be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the digital online detection and control system 100 can be implemented on a digital online detection and control system. By way of example only, the digital online detection and control system can include a private cloud, a semantically related cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any aggregation thereof.

[0135] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data acquired from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the digital online detection control system 100 uses to execute or use to complete the exemplary methods described in this application.

[0136] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the digital online detection and control method for the compressor cylinder in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0137] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the above-mentioned digital online detection and control system 100. The implementation principles and technical effects are similar, and this embodiment will not be repeated here.

[0138] In addition, an embodiment of the present application also provides a readable storage medium, in which computer executable instructions are set. When a processor executes the computer executable instructions, the above-mentioned digital online detection and control method for a compressor cylinder is implemented.

[0139] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A digital online detection and control method for a compressor cylinder, characterized in that: The method comprises: Acquire candidate operating state events of the compressor cylinder and each prior operating state event associated with the candidate operating state event, wherein each prior operating state event is: an operating state event that meets preset performance requirements and is generated based on the working data and prior operating records of the compressor cylinder; For each of the a priori operating state events, according to the operating state vectors of the multiple operating state nodes included in the a priori operating state event, an abnormal excitation parameter of each operating state node is generated, and according to the abnormal excitation parameter of each operating state node, an abnormal influence coefficient of the corresponding operating state node is determined respectively, wherein the abnormal excitation parameter is used to represent: a probability score that the corresponding operating state node indicates that the compressor cylinder has an abnormal operating condition; Determine a plurality of association weights between each priori running state event and the candidate running state event based on the running state vectors of the plurality of running state nodes included in the candidate running state event and the plurality of running state nodes included in the priori running state events, respectively, in combination with the abnormal influence coefficients of the plurality of running state nodes included in the priori running state events; Determining, from the association weights, a plurality of target association weights that meet the set detection requirements as the state detection result of the candidate running state event; Based on the state detection result of the candidate operating state event, adaptively controlling the compressor cylinder; The step of adaptively controlling the compressor cylinder based on the state detection result of the candidate operating state event comprises: According to the state detection result of the candidate operation state event, the current operation state of the compressor cylinder is evaluated, and preliminary evaluation results of the current operation state of the compressor cylinder in different dimensions are output; According to the preliminary evaluation result, the potential problem area existing in the current operating state of the compressor cylinder is determined, and based on the determined potential problem area, a pre-constructed compressor cylinder control strategy knowledge base is queried, and by matching the current potential problem area, a suitable set of candidate control strategies is screened out from the compressor cylinder control strategy knowledge base, wherein the compressor cylinder control strategy knowledge base stores a variety of control strategies for different potential problems, and the control strategies are constructed based on compressor cylinder operating data, prior knowledge, and engineering experience data; Performing applicability analysis on the candidate control strategy set, analyzing the applicability of each candidate control strategy in the candidate control strategy set under the current situation, and generating a control strategy subset that has been screened for applicability; Prioritize the control strategies according to the subset of control strategies that have been screened for applicability, generate a list of control strategies in order of priority, select the control strategies in order of priority and conduct pre-implementation evaluation, simulate the operation effect after pre-implementation of the control strategies, and generate the evaluation results of the simulated operation effect after pre-implementation of each control strategy; According to the evaluation results of the simulated operation effect after the preliminary implementation, the optimal control strategy is selected for actual implementation.

2. The digital online detection and control method for compressor cylinder according to claim 1 is characterized in that: The step of generating an abnormal excitation parameter of each of the operating state nodes according to the operating state vectors of the plurality of operating state nodes included in the prior operating state event, and determining an abnormal influence coefficient of a corresponding operating state node according to the abnormal excitation parameter of each of the operating state nodes, comprises: According to the operation state vectors of the multiple operation state nodes included in the prior operation state event, combined with the event characteristics of the prior operation state event and the prior operation record, an abnormal excitation parameter of each of the operation state nodes is generated; Each abnormal excitation parameter is converted into a regularized form to generate an abnormal impact coefficient of each operating state node.

3. The digital online detection and control method for compressor cylinder according to claim 2 is characterized in that: The generating of abnormal excitation parameters of each of the operation status nodes according to the operation status vectors of the multiple operation status nodes included in the prior operation status event and in combination with the event characteristics of the prior operation status event and the prior operation record comprises: Generate a correlation between each of the operating state nodes and the prior operating state event according to the operating state vectors of the plurality of operating state nodes included in the prior operating state event and the event characteristics of the prior operating state event; Determining the abnormal tendency of the priori operating state event according to the priori operating record of the priori operating state event; According to each correlation degree and the abnormal tendency degree, an abnormal excitation parameter of each operating state node is determined.

4. The digital online detection and control method for compressor cylinder according to claim 3 is characterized in that: The step of determining the abnormal tendency of the priori operating state event based on the priori operating record of the priori operating state event comprises: In the prior operation record, according to the definition information and event characteristics of the prior operation status event, a data subset directly or indirectly related to the prior operation status event is screened out; Performing feature identification on each data item in the screened data subset to generate a data subset with feature identification, wherein the identified features include extreme value features of operating parameters, change rate features of operating parameters, and sequence features of operation events; In the data subset with feature identification, searching for an abnormal feature pattern indicating an abnormality, and generating an identified data subset containing the abnormal feature pattern, wherein the abnormal feature pattern is based on a predefined rule or is learned from historical fault data; Performing cluster analysis on the identified data subsets containing abnormal feature patterns, dividing the data subsets into different cluster groups according to the similarity of the abnormal feature patterns, generating abnormal data groups clustered by abnormal type, and for each abnormal data group, counting the amount of data contained in the abnormal data group, generating scale statistics results of each abnormal data group, wherein the data amount reflects the scale of each abnormal type appearing in the priori operation record; According to the scale statistics of each abnormal data group, the relative abnormal tendency of each abnormal data group is evaluated, and the relative abnormal tendency of each abnormal data group based on the scale is generated; Analyze the correlation between each abnormal data group and the operating state factor in the prior operating state event, and adjust the relative abnormal tendency of each abnormal data group based on the scale according to the correlation to generate a corresponding abnormal tendency adjustment value; All the abnormal tendency adjustment values ​​are fused and calculated to obtain the abnormal tendency of the priori operating state event, wherein the fusion calculation process adopts a weighted average method.

5. The digital online detection and control method for a compressor cylinder according to any one of claims 1 to 4, characterized in that: The determining of multiple association weights between each priori running state event and the candidate running state event based on the running state vectors of the multiple running state nodes included in the candidate running state event and the multiple running state nodes included in the priori running state events, in combination with the abnormal influence coefficients of the multiple running state nodes included in the priori running state events, includes: Based on the log data corresponding to each of the priori running state events and the candidate running state events, respectively, extracting the running state vectors of the multiple running state nodes included in the priori running state events and the multiple running state nodes included in the candidate running state events, and performing regularization conversion on each running state vector to generate each target running state vector; Determining, based on the target operating state vectors of the multiple operating state nodes included in the prior operating state events and the target operating state vectors of the multiple operating state nodes included in the candidate operating state events, the state correlation between the multiple operating state nodes included in the prior operating state events and the multiple operating state nodes included in the candidate operating state events; For each of the prior operating status events, multiple association weights between the prior operating status event and the candidate operating status event are determined based on the state correlation between each of the multiple operating status nodes included in the prior operating status event and the multiple operating status nodes included in the candidate operating status event, combined with the abnormal influence coefficients of the multiple operating status nodes included in the prior operating status event.

6. The digital online detection and control method for compressor cylinders according to claim 5 is characterized in that: The determining of multiple association weights between the priori running state event and the candidate running state event based on the state correlation between each of the multiple running state nodes included in the priori running state event and the multiple running state nodes included in the candidate running state event, in combination with the abnormal influence coefficients of the multiple running state nodes included in the priori running state event, includes: For the multiple operating state nodes included in the candidate operating state event, according to the previously defined selection strategy, determine the target state correlation among the multiple state correlations corresponding to one operating state node, and determine the first association weight between the candidate operating state event and a priori operating state event according to each target state correlation and the abnormal influence coefficients of the multiple operating state nodes included in the priori operating state event; Alternatively, for the multiple operating state nodes included in the prior operating state event, according to the selection strategy, a target state correlation among the multiple state correlations corresponding to an operating state node is determined, and according to each target state correlation and the corresponding abnormal influence coefficient, a second association weight between the candidate operating state event and a prior operating state event is determined; Alternatively, for the multiple operating status nodes included in the candidate operating status event and the prior operating status event, according to the selection strategy, the target state correlation among the multiple state correlations corresponding to an operating status node is determined, and according to each target state correlation and the abnormal impact coefficient of the multiple operating status nodes included in the prior operating status event, the third association weight between the candidate operating status event and the prior operating status event is determined.

7. The digital online detection and control method for a compressor cylinder according to any one of claims 1 to 4, characterized in that: After determining, from among the association weights, a plurality of target association weights that meet the set detection requirements as the state detection result of the candidate running state event, the method further includes: Extracting a set number of template running status event combinations from the template running status event sequence; According to a plurality of template association weights between two template operation status events included in each template operation status event combination, adjusting the range of the plurality of target association weights; The adjusted multiple target association weights are used as the status detection result of the candidate running status event.

8. The digital online detection and control method for compressor cylinders according to claim 7 is characterized in that: The range adjustment of the multiple target association weights according to the multiple template association weights between the two template operation status events included in each template operation status event combination includes: For each template running state event combination, according to the log data respectively corresponding to the two template running state events included in one template running state event combination, the template running state vectors of the multiple running state nodes included in each of the two template running state events are extracted, and according to the template running state vector of each running state node, the multiple template association weights between the two template running state events are determined; Determine multiple reference values ​​of a reference range according to each template association weight and the set quantity, wherein the reference value is a threshold value under the reference range, and the reference range is a global range of the template association weights corresponding to each template operation status event combination; According to the plurality of reference values ​​and the plurality of target association weights, the adjusted plurality of target association weights are determined.

9. A digital online detection and control system, characterized in that: The digital online detection and control system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the digital online detection and control method for compressor cylinders as described in any one of claims 1 to 8.

Images