Axial piston pump state detection method and system based on monitoring data fusion

Abstract

The present application provides an axial piston pump state detection method and system based on monitoring data fusion, which relates to the field of data processing technology. The method includes: obtaining multiple groups of vibration signals; obtaining multiple groups of operating characteristic coefficients; constructing a state identifier; obtaining multiple occurrence coefficients; judging whether multiple occurrence coefficients are greater than the occurrence coefficient threshold, if so, inputting the vibration signal and the operating characteristic coefficient into a typical identification path to obtain a state identification result, if not, inputting the vibration signal and the operating characteristic coefficient into multiple matching special identification paths, outputting the state identification result; obtaining the state detection result. The present application can solve the technical problem in the prior art that it is difficult to perform accurate state detection due to the small number of samples of some detection states. By constructing typical identification branches and special identification branches, typical states and special states are identified, and the technical effect of classifying and detecting the sample size and improving the state detection precision and detection accuracy is achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an axial piston pump state detection method and system based on monitoring data fusion. Background Art

[0002] The axial piston pump is an important component of the hydraulic system, and its operating status directly affects the performance and stability of the entire system.

[0003] Traditional detection methods often rely on sensor monitoring data, such as vibration signals, pressure signals, etc., and perform intelligent detection through machine learning models. This detection method often requires training with a large amount of sample data. However, plunger pumps may have anomalies that have not appeared in history or have occurred less frequently. For this type of anomaly, it is difficult to make accurate predictions because it occurs less frequently, resulting in fewer samples.

[0004] In summary, the prior art has a technical problem that it is difficult to perform accurate state detection due to the small number of samples of some detection states. Summary of the invention

[0005] The purpose of this application is to provide an axial piston pump state detection method and system based on monitoring data fusion, so as to solve the technical problem in the prior art that it is difficult to perform accurate state detection due to the small number of samples of some detection states.

[0006] In view of the above problems, the present application provides an axial piston pump state detection method and system based on monitoring data fusion.

[0007] In a first aspect, the present application provides an axial piston pump state detection method based on monitoring data fusion, wherein the method is implemented by an axial piston pump state detection system based on monitoring data fusion, wherein the method comprises: during the operation process of the axial piston pump to be state detected, collecting vibration signals of multiple pairs of piston pairs to obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; collecting oil suction characteristic information and oil discharge characteristic information of multiple pairs of piston pairs during the oil suction and oil discharge process to obtain multiple groups of operation characteristic information, and processing and calculating the multiple groups of operation characteristic information according to the swash plate angle of the axial piston pump to obtain multiple groups of operation characteristic coefficients; constructing a state identifier based on the state detection data record of the axial piston pump, wherein the state identifier includes a typical identification branch and a special identification branch, wherein the typical identification branch includes multiple typical identification paths, and the special identification branch includes multiple Special identification paths, each special identification path is constructed by a group of special vibration signals and a group of special operating characteristic coefficients, and includes an identified special state; based on the multiple groups of vibration signals and multiple groups of operating characteristic coefficients, data query is performed in the state detection data record to obtain multiple vibration occurrence coefficients and multiple operating characteristic occurrence coefficients, and multiple occurrence coefficients are calculated; respectively determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold, if so, obtain the typical level according to the occurrence coefficient decision, and obtain the analysis quantity by indexing, input the vibration signal and the operating characteristic coefficient into the typical identification path of the analysis quantity, and obtain the state identification result, if not, select multiple matching special identification paths according to the occurrence coefficient, input the vibration signal and the operating characteristic coefficient into multiple matching special identification paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state identification result; based on multiple state identification results, classify and obtain the state detection result of the axial piston pump.

[0008] In the second aspect, the present application also provides an axial piston pump state detection system based on monitoring data fusion, which is used to execute the axial piston pump state detection method based on monitoring data fusion as described in the first aspect, wherein the system includes: a vibration signal acquisition module, which is used to collect vibration signals of multiple pairs of piston pairs during the operation process of the axial piston pump to be state detected, and obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; an operation feature acquisition module, which is used to collect the oil suction feature information and oil discharge feature information of multiple pairs of piston pairs during the oil suction and oil discharge process, and obtain multiple groups of operation feature information, and process and calculate the multiple groups of operation feature information according to the swash plate angle of the axial piston pump to obtain multiple groups of operation feature coefficients; a state identifier construction module, which is used to construct a state identifier based on the state detection data record of the axial piston pump, wherein the state identifier includes a typical identification branch and a special identification branch, and the typical identification branch includes multiple typical identification paths, and the special identification branch includes multiple typical identification paths. The branch includes multiple special identification paths, each of which is constructed by a group of special vibration signals and a group of special operating characteristic coefficients, and includes an identified special state; a data query module is used to perform data query in the state detection data record based on the multiple groups of vibration signals and multiple groups of operating characteristic coefficients, obtain multiple vibration occurrence coefficients and multiple operating characteristic occurrence coefficients, and calculate and obtain multiple occurrence coefficients; a state identification module is used to respectively determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold. If so, obtain the typical level according to the occurrence coefficient decision, and index to obtain the analysis quantity, input the vibration signal and the operating characteristic coefficient into the typical identification path of the analysis quantity, and obtain the state identification result. If not, select multiple matching special identification paths according to the occurrence coefficient, input the vibration signal and the operating characteristic coefficient into multiple matching special identification paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state identification result; a result classification module is used to classify and obtain the state detection result of the axial piston pump based on multiple state identification results.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] During the operation of the axial piston pump to be state-detected, vibration signals of multiple pairs of piston pairs are collected to obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; oil suction characteristic information and oil discharge characteristic information of multiple pairs of piston pairs during the oil suction and oil discharge process are collected to obtain multiple groups of operation characteristic information, and the multiple groups of operation characteristic information are processed and calculated according to the swash plate angle of the axial piston pump to obtain multiple groups of operation characteristic coefficients; based on the state detection data record of the axial piston pump, a state identifier is constructed, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes multiple typical identification paths, the special identification branch includes multiple special identification paths, and each special identification path consists of a group of special vibration signals and a group of special operation characteristic coefficients Construct, and include the special state of the identification; based on the multiple groups of vibration signals and multiple groups of operating characteristic coefficients, perform data query in the state detection data record, obtain multiple vibration occurrence coefficients and multiple operating characteristic occurrence coefficients, and calculate and obtain multiple occurrence coefficients; respectively judge whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold, if so, obtain the typical level according to the occurrence coefficient decision, and obtain the analysis quantity by index, input the vibration signal and the operating characteristic coefficient into the typical recognition path of the analysis quantity, and obtain the state recognition result, if not, select and obtain multiple matching special recognition paths according to the occurrence coefficient, input the vibration signal and the operating characteristic coefficient into multiple matching special recognition paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state recognition result; based on multiple state recognition results, classify and obtain the state detection result of the axial piston pump. By constructing typical recognition branches and special recognition branches, typical states and special states are identified, and state detection based on sample size matching recognition branches is realized, so as to achieve the technical effect of classifying and detecting the sample size and improving the state detection precision and detection accuracy.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0013] Figure 1 This is a flow chart of an axial piston pump state detection method based on monitoring data fusion in this application;

[0014] Figure 2 This is a structural schematic diagram of the axial piston pump state detection system based on monitoring data fusion in this application.

[0015] Explanation of the accompanying drawings: vibration signal acquisition module 11, operation feature acquisition module 12, state identifier construction module 13, data query module 14, state identification module 15, result classification module 16. DETAILED DESCRIPTION

[0016] This application solves the technical problem in the prior art that it is difficult to perform accurate state detection due to the small number of samples of some detection states by providing an axial piston pump state detection method and system based on monitoring data fusion. By constructing typical recognition branches and special recognition branches, typical states and special states are identified, and state detection is performed based on sample size matching recognition branches, achieving the technical effect of performing classification detection based on sample size and improving state detection precision and detection accuracy.

[0017] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0018] Embodiment 1

[0019] Please refer to the attached Figure 1 The present application provides an axial piston pump state detection method based on monitoring data fusion, wherein the method is applied to an axial piston pump state detection system based on monitoring data fusion, and the method specifically comprises the following steps:

[0020] Step 1: During the operation of the axial piston pump to be tested, vibration signals of multiple pairs of piston pairs are collected to obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder;

[0021] Specifically, an axial piston pump usually includes multiple pairs of piston pairs, each pair of piston pairs refers to two opposite piston pairs, and each piston pair includes a piston and a cylinder. During the rotation process, the piston of one piston pair enters the cylinder, and the corresponding piston of the other piston pair on the other side moves away from the cylinder, achieving oil suction on one side and oil discharge on the other side, forming a pump, which is an axial piston pump.

[0022] During the operation of the axial piston pump to be tested for status, friction wear or other mechanical wear reasons may cause changes in oil suction and discharge volumes, resulting in adverse vibration signals. Vibration sensors arranged on the piston pairs are used to collect vibration signals of multiple pairs of piston pairs to obtain multiple groups of vibration signals. The vibration signals can be vibration waveform signals of the piston pairs, providing data support for subsequent status monitoring.

[0023] Step 2: collecting the oil suction characteristic information and the oil discharge characteristic information of the plurality of pairs of piston pairs during the oil suction and oil discharge process, obtaining a plurality of sets of operation characteristic information, and processing and calculating the plurality of sets of operation characteristic information according to the swash plate angle of the axial piston pump, to obtain a plurality of sets of operation characteristic coefficients;

[0024] Specifically, the oil suction characteristic information refers to the oil suction amount of the plunger pair used for oil suction in any pair of plunger pairs, and the oil discharge characteristic information refers to the oil discharge amount of the plunger pair used for oil discharge in any pair of plunger pairs. It can be collected through the existing flow sensor, and then the oil suction amount and oil discharge amount of multiple pairs of plunger pairs are averaged, and the result is multiple sets of operation characteristic information. The swash plate angle of the axial piston pump affects the theoretical volume of the cylinder in the plunger pair. During the operation of the plunger pump, the swash plate angle determines the stroke length of the piston in the cylinder, thereby affecting the amount of oil that can be theoretically sucked or discharged in each stroke, that is, the theoretical cylinder volume. Ideally, the theoretical cylinder volume should be consistent with multiple sets of operation characteristic information. However, due to friction wear or other mechanical wear reasons, the oil suction and discharge amount will change, resulting in abnormal operation status. The ratio of multiple sets of operation characteristic information to the theoretical cylinder volume is used as the multiple sets of operation characteristic coefficients.

[0025] Step 3: Based on the state detection data record of the axial piston pump, a state identifier is constructed, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes a plurality of typical identification paths, the special identification branch includes a plurality of special identification paths, each special identification path is constructed by a set of special vibration signals and a set of special operation characteristic coefficients, and includes a special state of the identification;

[0026] Specifically, the axial piston pump may have various abnormal operating conditions. The status detection data record refers to all historical status detection data after the axial piston pump to be tested begins to serve. Each status detection data in the status detection data record includes a vibration signal, an operating characteristic coefficient and a detection status. The detection status is the operating status of the axial piston pump, such as abnormal piston wear.

[0027] Based on the state detection data records, the same state detection data are clustered, and the occurrence ratio of each type of state detection data is used as the occurrence coefficient. The larger the occurrence coefficient, the higher the probability of occurrence of this type of data, which is a typical detection state. This type of state detection data is used as sample data to construct a typical recognition branch, and the frequently occurring states are detected and analyzed. The typical recognition branch includes multiple typical recognition paths. The state detection data corresponding to a certain proportion of typical detection states are randomly selected, and multiple typical recognition paths are constructed based on the neural network, and the typical recognition branches are integrated.

[0028] State detection data with a small occurrence coefficient can be considered as a special detection state of a small sample. This type of state detection data is used as sample data to construct a special identification branch to detect and analyze special states with a small number of occurrences. The special identification branch includes multiple special identification paths. The data of a small sample is small and cannot be predicted. Therefore, based on the twin network, each state detection data corresponding to the special detection state, that is, a group of special vibration signals and a group of special operation characteristic coefficients in the state detection data are used to construct a path respectively, identify the similarity between the input vibration signal and the operation characteristic coefficient and each state detection data, and thus determine whether the special state corresponding to the state detection data appears.

[0029] Step 4: Based on the multiple groups of vibration signals and the multiple groups of operation characteristic coefficients, data query is performed in the state detection data record to obtain multiple vibration occurrence coefficients and multiple operation characteristic occurrence coefficients, and multiple occurrence coefficients are calculated;

[0030] Specifically, based on the multiple groups of vibration signals and multiple groups of operation characteristic coefficients, the state detection data records are traversed for query, the number of data identical to the multiple groups of vibration signals is extracted as multiple vibration occurrence coefficients, and the number of data identical to the multiple groups of operation characteristic coefficients is obtained as the operation characteristic occurrence coefficients. The multiple vibration occurrence coefficients and the multiple operation characteristic occurrence coefficients are weighted averaged to obtain multiple occurrence coefficients.

[0031] Step 5: respectively determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold value. If so, obtain the typical level according to the occurrence coefficient decision, obtain the analysis quantity by index, input the vibration signal and the operation characteristic coefficient into the typical recognition path of the analysis quantity, and obtain the state recognition result. If not, obtain multiple matching special recognition paths according to the occurrence coefficient selection, input the vibration signal and the operation characteristic coefficient into multiple matching special recognition paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state recognition result;

[0032] Specifically, the occurrence coefficient threshold is set by professional and technical personnel in this field, which is a distinguishing feature between typical detection states and special detection states. The setting value of the occurrence coefficient threshold is relatively small, such as 10 times. If the multiple occurrence coefficients are greater than the occurrence coefficient threshold, it means that the number of occurrences in the state detection data record is relatively high. The typical level is obtained by the occurrence coefficient decision, and the number of analyses is obtained by indexing. The higher the occurrence coefficient, the higher the typical level, and the smaller the number of analyses. That is to say, the higher the typical level, the higher the probability that the corresponding detection state exists in the sample data of each typical identification path, and a smaller number of typical identification paths are used for identification, and the state identification result is output.

[0033] On the contrary, if multiple occurrence coefficients are less than or equal to the occurrence coefficient threshold, they are special data. The vibration signal and the operation characteristic coefficient are input into the special identification path, and the state similarity of the sample data in multiple special identification paths in the detection area is detected to obtain multiple state similarities. The special state with the greatest state similarity is output as the state identification result.

[0034] Step six: Based on the multiple state recognition results, classify and obtain the state detection result of the axial piston pump.

[0035] Specifically, statistics are collected on a plurality of state recognition results, and the state recognition result with the highest occurrence frequency is obtained as the state detection result of the axial piston pump.

[0036] Furthermore, step 2 of this application also includes:

[0037] The oil suction flow information and the oil discharge flow information of multiple pairs of piston pairs during the oil suction and oil discharge processes are collected as the oil suction characteristic information and the oil discharge characteristic information to obtain multiple groups of operation characteristic information; the swash plate angle of the axial piston pump is obtained, and the cylinder volume information under the swash plate angle is obtained; the ratio of the multiple groups of operation characteristic information to the cylinder volume information is calculated to obtain multiple groups of operation characteristic coefficients.

[0038] Specifically, the oil suction characteristic information and the oil discharge characteristic information of multiple pairs of piston pairs in the oil suction and oil discharge process are collected to obtain multiple sets of operation characteristic information. According to the swash plate angle of the axial piston pump, the process of processing and calculating the multiple sets of operation characteristic information is as follows:

[0039] The oil suction flow information and oil discharge flow information of multiple pairs of plunger pairs in the oil suction and oil discharge process are collected through existing flow sensors as oil suction characteristic information and oil discharge characteristic information, and the oil suction characteristic information and oil discharge characteristic information are weighted averaged to obtain multiple sets of operation characteristic information through calculation results. The swash plate angle of the axial piston pump can be obtained through existing angle detection equipment or through the control center of the axial piston pump. The swash plate angle of the plunger pump has a direct impact on the theoretical volume of the cylinder in the plunger pair. When the swash plate angle increases, the stroke length of the plunger in the cylinder will also increase accordingly. In each working cycle, the piston can enter the cylinder more deeply, and the theoretical volume of the cylinder will increase with the increase of the swash plate angle.

[0040] Specifically, professional and technical personnel in this field can extract cylinder volume samples corresponding to different swash plate angle samples in combination with historical data to form a cylinder volume mapping library, and obtain corresponding cylinder volume information based on the swash plate angle in the cylinder volume mapping library. Then, the ratios of the multiple sets of operation feature information and the cylinder volume information are calculated respectively to obtain multiple sets of operation feature coefficients. In this way, the operation feature collection of multiple pairs of plunger pairs is realized, providing data support for subsequent status monitoring.

[0041] Furthermore, step three of this application also includes:

[0042] Based on the state detection data record of the axial piston pump, a plurality of sample state detection data are collected and obtained, each sample state detection data includes a group of sample vibration signals, a group of sample operation characteristic coefficients and a sample detection state; the same data in the plurality of sample state detection data are clustered to obtain a plurality of sample data sets; the proportion of the data in each sample data set in all the sample data is calculated to obtain a plurality of occurrence coefficients; according to the size of the plurality of occurrence coefficients, it is determined whether they are greater than an occurrence coefficient threshold, and the plurality of sample data sets are classified to obtain a typical sample data set and a special sample data set; the typical sample data set is used to construct a typical recognition branch including a plurality of typical recognition paths, and the special sample data set is used to construct a special recognition branch including a plurality of special recognition paths to obtain a state identifier.

[0043] Furthermore, the present application also includes the following steps:

[0044] A first typical construction data set is randomly extracted from the typical sample data set, wherein the first typical construction data set includes at least half of the sample data in the typical sample data set; the sample vibration signal and the sample operation characteristic coefficient in the first typical construction data set are used as input, and the sample detection state is used as output to construct a first typical recognition path; a second typical construction data set is randomly extracted from the typical sample data set again to construct a second typical recognition path, wherein the second typical construction data set is different from the first typical construction data set; and multiple typical recognition paths are continuously constructed to obtain the typical recognition branch.

[0045] Furthermore, the present application also includes the following steps:

[0046] The sample vibration signals and sample operation characteristic coefficients in each different group of special sample data in the special sample data set are used as special vibration signals and special operation characteristic coefficients, and multiple special identification paths are constructed based on the twin network, wherein each special identification path identifies the similarity between the input vibration signal and operation characteristic coefficient and the special vibration signal and special operation characteristic coefficient; each different sample detection state in the special sample data set is used as multiple special states, the multiple special identification paths are identified, and the special identification branches are obtained.

[0047] Specifically, based on the state detection data record of the axial piston pump, the process of building a state identifier is as follows:

[0048] First, based on the state detection data record of the axial piston pump, multiple sample state detection data are collected and obtained, each sample state detection data includes a set of sample vibration signals, a set of sample operation characteristic coefficients and a sample detection state, and the sample detection state refers to the operating state of the axial piston pump, such as abnormal piston wear state. The sample operation characteristic coefficient can be obtained by the method provided in this application, that is, calculating the ratio of the operation characteristic information to the cylinder volume. Then, the same data in the multiple sample state detection data is clustered by existing clustering algorithms such as K-means and hierarchical clustering to obtain multiple sample data sets. The proportion of data in each sample data set in multiple sample state detection data is calculated to obtain multiple occurrence coefficients. It is determined whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold, and the multiple sample data sets are classified. The sample state detection data with an occurrence coefficient greater than the occurrence coefficient threshold are combined to obtain a typical sample data set, and the sample state detection data with an occurrence coefficient less than or equal to the occurrence coefficient threshold are combined to obtain a special sample data set.

[0049] The typical sample data set is used to construct a typical identification branch including multiple typical identification paths. The specific process is: randomly extract multiple sample state detection data in the typical sample data set to construct a first typical construction data set, wherein the first typical construction data set includes at least half of the sample data in the typical sample data set, and the random extraction can be performed by setting a data volume threshold. The sample vibration signal and sample operation characteristic coefficient in the first typical construction data set are used as input, and the sample detection state is used as output. The first typical identification path is constructed based on an existing machine learning model, such as a neural network model. Specifically, the first typical construction data set can be divided into a training set and a verification set to train and verify the first typical identification path, and obtain the first typical identification path trained to convergence. This is a common technical means used by those skilled in the art and will not be expanded here.

[0050] Re-randomly extract a second typical construction data set from the typical sample data set to construct a second typical identification path, wherein the second typical construction data set is different from the first typical construction data set, and the construction method of the second typical identification path is the same as that of the first typical identification path, that is, the sample vibration signal and the sample operation characteristic coefficient in the second typical construction data set are used as input, and the sample detection state is used as output. The second typical identification path is constructed based on the existing machine learning model, which is not elaborated here. And so on, continue to obtain. Continue to randomly extract typical construction data sets from the typical sample data set, construct and obtain multiple typical identification paths, and integrate multiple typical identification paths to obtain the typical identification branch. In this way, the construction of typical identification branches is achieved, which is convenient for the subsequent identification of typical detection states and provides a model basis for the state detection of the plunger pump.

[0051] The special sample data set is used to construct a special identification branch including multiple special identification paths. The specific process is: the sample vibration signals and sample operation characteristic coefficients in each different group of special sample data in the special sample data set are used as special vibration signals and special operation characteristic coefficients, and multiple special identification paths are constructed based on the twin network, wherein each special identification path identifies the similarity between the input vibration signal and operation characteristic coefficient and the special vibration signal and special operation characteristic coefficient.

[0052] In simple terms, each special identification path corresponds to a group of special sample data including special vibration signals and special operation characteristic coefficients. Each special identification path includes two neural networks. One neural network is used to extract data features of special vibration signals and special operation characteristic coefficients, such as extracting features such as amplitude and frequency of vibration signals and numerical features of operation characteristic coefficients. The other neural network is used to extract features of input vibration signals and operation characteristic coefficients, and then perform similarity comparison between input vibration signals and operation characteristic coefficients and special vibration signals and special operation characteristic coefficients to obtain similarity. At the same time, each different sample detection state corresponding to each group of special sample data in the special sample data set is used as multiple special states to identify the multiple special identification paths. The similarity between the input vibration signal and operation characteristic coefficient and the special vibration signal and special operation characteristic coefficient can be used as the output as the state similarity between the current operating state of the plunger pair and multiple special states. Thus, the special identification branches are produced with multiple special identification paths to provide model support for subsequent analysis of special states.

[0053] The state identifier is composed of typical identification branches and special identification branches, which facilitates the identification of typical states and special states and improves the accuracy of state detection.

[0054] Furthermore, step five of this application also includes:

[0055] Obtain a set of sample occurrence coefficients, and evaluate and obtain a set of sample typical levels and a set of sample analysis quantities; construct a typical level classifier based on a decision tree, wherein the size of the sample occurrence coefficient is positively correlated with the size of the sample typical level, and negatively correlated with the size of the sample analysis quantity; based on the typical level classifier, make decision classification on the occurrence coefficients to obtain the typical level, and obtain the analysis quantity through indexing; randomly select a typical identification path of the analysis quantity from the multiple typical identification paths, input a group of vibration signals whose occurrence coefficients are greater than an occurrence coefficient threshold and a group of operation characteristic coefficients, and identify and obtain the typical detection state of the analysis quantity; output the typical detection state with the highest occurrence frequency as the state identification result.

[0056] Specifically, the typical level is obtained according to the occurrence coefficient decision, and the analysis quantity is obtained by indexing, and the vibration signal and the operation characteristic coefficient are input into the typical identification path of the analysis quantity to obtain the state identification result. The process is as follows:

[0057] Based on the state detection data records, a set of sample occurrence coefficients is obtained, and based on the evaluation of the prior art, a set of sample typical levels and a set of sample analysis quantities are obtained, wherein the size of the sample occurrence coefficient is positively correlated with the size of the sample typical level, and negatively correlated with the size of the sample analysis quantity. Based on the relationship between the sample occurrence coefficient and the sample typical level and the sample analysis quantity, a mapping relationship between the sample occurrence coefficient and the sample typical level and the sample analysis quantity can be set by a technician in this field to obtain a set of sample typical levels and a set of sample analysis quantities.

[0058] Based on the decision tree, a typical level classifier is constructed by obtaining a sample occurrence coefficient set and a sample typical level set. Specifically, a multi-layer one-class division node is constructed based on multiple sample occurrence coefficients in the sample occurrence coefficient set through a decision tree algorithm. Specifically, multiple sample occurrence coefficients are used as division thresholds in the multi-layer division node. Each layer of the one-class division node can perform a binary classification decision division on the sample occurrence coefficient input by the lower layer division node, and divide it into a class greater than the division threshold in the node and a class not greater than the division threshold, and input the division result to the upper layer division node. In this way, a multi-layer division decision is performed to obtain multiple division results. Each division result has its corresponding sample typical level in the sample typical level set. The sample typical levels corresponding to the multiple division results are marked, and then a typical level classifier is obtained. Based on the typical level classifier, the occurrence coefficient is classified by decision, the typical level is output, and the analysis quantity is obtained by indexing in the sample analysis quantity set.

[0059] Finally, a typical identification path of the analyzed number is randomly selected from the multiple typical identification paths, a group of vibration signals and a group of operation characteristic coefficients whose occurrence coefficients are greater than the occurrence coefficient threshold are input, and the typical detection states of the analyzed number are output. For example, if the analyzed number is 10, 10 typical identification paths are selected from the multiple typical identification paths to identify the input group of vibration signals and a group of operation characteristic coefficients, and 10 typical detection states are output. A statistical analysis is performed on the typical detection states of the analyzed number, and the typical detection state with the highest occurrence frequency is extracted as the state identification result. In this way, the detection of typical identification states is achieved.

[0060] Furthermore, the present application also includes the following steps:

[0061] Obtain the occurrence coefficients of the special vibration signals and special operation characteristic coefficients in the multiple special identification paths as multiple special occurrence coefficients; take the special identification paths whose special occurrence coefficients are less than or equal to the occurrence coefficients as matching special identification paths to obtain multiple matching special identification paths; input the vibration signals and operation characteristic coefficients into the multiple matching special identification paths respectively to obtain multiple similarities as multiple state similarities; output the special state identified by the matching special identification path corresponding to the maximum state similarity as the state identification result.

[0062] Specifically, the process of obtaining multiple matching special identification paths according to the occurrence coefficient selection, inputting the vibration signal and the operation characteristic coefficient into the multiple matching special identification paths, obtaining multiple state similarities, and outputting the special state with the largest state similarity as the state identification result is as follows:

[0063] Obtain the occurrence coefficients of the special vibration signals and special operation characteristic coefficients in the state detection data records in the multiple special identification paths as multiple special occurrence coefficients. The special identification path with a special occurrence coefficient less than or equal to the occurrence coefficient is used as a matching special identification path, where the occurrence coefficient refers to the multiple occurrence coefficients of the multiple groups of vibration signals and multiple groups of operation characteristic coefficients obtained in the aforementioned step four. In layman's terms, the occurrence coefficient can be understood as the special degree of the current multiple groups of vibration signals and multiple groups of operation characteristic coefficients. The special identification path with a special degree less than or equal to the current detection data is selected as the matching special identification path, and multiple matching special identification paths are obtained to prevent certain special states from being missed due to a small number of occurrences, thereby improving the detection accuracy. The vibration signal and the operation characteristic coefficient are respectively input into multiple matching special identification paths, and multiple similarities are output as multiple state similarities. Finally, the special state identified by the matching special identification path corresponding to the maximum state similarity is output as the state recognition result, thereby realizing the detection of the special state of the axial piston pump, that is, identifying the state that appears less in history, and improving the accuracy and precision of state detection.

[0064] In summary, the axial piston pump state detection method based on monitoring data fusion provided by the present application has the following technical effects:

[0065] During the operation of the axial piston pump to be state-detected, vibration signals of multiple pairs of piston pairs are collected to obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; oil suction characteristic information and oil discharge characteristic information of multiple pairs of piston pairs during the oil suction and oil discharge process are collected to obtain multiple groups of operation characteristic information, and the multiple groups of operation characteristic information are processed and calculated according to the swash plate angle of the axial piston pump to obtain multiple groups of operation characteristic coefficients; based on the state detection data record of the axial piston pump, a state identifier is constructed, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes multiple typical identification paths, the special identification branch includes multiple special identification paths, and each special identification path consists of a group of special vibration signals and a group of special operation characteristic coefficients Construct, and include the special state of the identification; based on the multiple groups of vibration signals and multiple groups of operating characteristic coefficients, perform data query in the state detection data record, obtain multiple vibration occurrence coefficients and multiple operating characteristic occurrence coefficients, and calculate and obtain multiple occurrence coefficients; respectively judge whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold, if so, obtain the typical level according to the occurrence coefficient decision, and obtain the analysis quantity by index, input the vibration signal and the operating characteristic coefficient into the typical recognition path of the analysis quantity, and obtain the state recognition result, if not, select and obtain multiple matching special recognition paths according to the occurrence coefficient, input the vibration signal and the operating characteristic coefficient into multiple matching special recognition paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state recognition result; based on multiple state recognition results, classify and obtain the state detection result of the axial piston pump. By constructing typical recognition branches and special recognition branches, typical states and special states are identified, and state detection based on sample size matching recognition branches is realized, so as to achieve the technical effect of classifying and detecting the sample size and improving the state detection precision and detection accuracy.

[0066] Embodiment 2

[0067] Based on the same inventive concept as the axial piston pump state detection method based on monitoring data fusion in the aforementioned embodiment, the present application also provides an axial piston pump state detection system based on monitoring data fusion, please refer to the attached Figure 2 , the system comprising:

[0068] A vibration signal acquisition module 11 is used to collect vibration signals of multiple pairs of piston pairs during the operation of the axial piston pump to be tested, and obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder;

[0069] The operation characteristic acquisition module 12 is used to acquire the oil suction characteristic information and the oil discharge characteristic information of the plurality of pairs of piston pairs during the oil suction and oil discharge process, obtain a plurality of sets of operation characteristic information, and process and calculate the plurality of sets of operation characteristic information according to the swash plate angle of the axial piston pump to obtain a plurality of sets of operation characteristic coefficients;

[0070] A state identifier building module 13 is used to build a state identifier based on the state detection data record of the axial piston pump, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes a plurality of typical identification paths, the special identification branch includes a plurality of special identification paths, each special identification path is constructed by a set of special vibration signals and a set of special operation characteristic coefficients, and includes a special state of the identification;

[0071] A data query module 14 is used to perform data query in the state detection data record based on the multiple groups of vibration signals and the multiple groups of operation characteristic coefficients, obtain multiple vibration occurrence coefficients and multiple operation characteristic occurrence coefficients, and calculate and obtain multiple occurrence coefficients;

[0072] The state recognition module 15 is used to respectively determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold value. If so, a typical level is obtained according to the occurrence coefficient decision, and the analysis quantity is obtained by indexing, and the vibration signal and the operation characteristic coefficient are input into the typical recognition path of the analysis quantity to obtain the state recognition result. If not, a plurality of matching special recognition paths are obtained according to the occurrence coefficient selection, and the vibration signal and the operation characteristic coefficient are input into the plurality of matching special recognition paths to obtain a plurality of state similarities, and the special state with the largest state similarity is output as the state recognition result;

[0073] The result classification module 16 is used to classify the state detection result of the axial piston pump based on the multiple state recognition results.

[0074] Furthermore, the operation feature collection module 12 in the system is also used for:

[0075] Collecting the oil suction flow information and the oil discharge flow information of multiple pairs of plunger pairs during the oil suction and oil discharge process as the oil suction characteristic information and the oil discharge characteristic information, and obtaining multiple sets of operation characteristic information;

[0076] Acquire the swash plate angle of the axial piston pump, and acquire the cylinder volume information at the swash plate angle;

[0077] The ratios of the multiple sets of operation characteristic information and the cylinder volume information are calculated to obtain multiple sets of operation characteristic coefficients.

[0078] Furthermore, the state identifier construction module 13 in the system is also used for:

[0079] Based on the state detection data record of the axial piston pump, a plurality of sample state detection data are collected and obtained, each sample state detection data includes a set of sample vibration signals, a set of sample operation characteristic coefficients and a sample detection state;

[0080] Clustering the same data in the plurality of sample state detection data to obtain a plurality of sample data sets;

[0081] Calculate the proportion of data in each sample data set in the total sample data to obtain multiple occurrence coefficients;

[0082] According to the magnitudes of the multiple occurrence coefficients, determining whether they are greater than an occurrence coefficient threshold, classifying the multiple sample data sets to obtain a typical sample data set and a special sample data set;

[0083] The typical sample data set is used to construct a typical recognition branch including a plurality of typical recognition paths, and the special sample data set is used to construct a special recognition branch including a plurality of special recognition paths to obtain a state identifier.

[0084] Furthermore, the state identifier construction module 13 in the system is also used for:

[0085] Randomly extracting a first typical constructed data set from the typical sample data set, wherein the first typical constructed data set includes at least half of the sample data in the typical sample data set;

[0086] Using the sample vibration signal and the sample operation characteristic coefficient in the first typical construction data set as input and the sample detection state as output, a first typical identification path is constructed;

[0087] randomly extracting a second typical constructed data set from the typical sample data set again to construct a second typical recognition path, wherein the second typical constructed data set is different from the first typical constructed data set;

[0088] Continue to construct and obtain multiple typical recognition paths to obtain the typical recognition branches.

[0089] Furthermore, the state identifier construction module 13 in the system is also used for:

[0090] Using the sample vibration signal and the sample operation characteristic coefficient in each group of different special sample data in the special sample data set as the special vibration signal and the special operation characteristic coefficient, and constructing a plurality of special recognition paths based on the twin network, wherein each special recognition path recognizes the similarity between the input vibration signal and the operation characteristic coefficient and the special vibration signal and the special operation characteristic coefficient;

[0091] Each different sample detection state in the special sample data set is used as a plurality of special states to identify the plurality of special identification paths to obtain the special identification branches.

[0092] Furthermore, the state identification module 15 in the system is also used for:

[0093] Obtain a set of sample occurrence coefficients, evaluate and obtain a set of sample typical levels and a set of sample analysis quantities, and construct a typical level classifier based on a decision tree, where the size of the sample occurrence coefficient is positively correlated with the size of the sample typical level and negatively correlated with the size of the sample analysis quantity;

[0094] Based on the typical grade classifier, decision classification is performed on the occurrence coefficients to obtain the typical grade, and the analysis quantity is obtained by indexing;

[0095] Randomly selecting the typical identification path of the analyzed number from the multiple typical identification paths, inputting a group of vibration signals with an appearance coefficient greater than an appearance coefficient threshold and a group of operation characteristic coefficients, and identifying and obtaining the typical detection states of the analyzed number;

[0096] The typical detection state with the highest occurrence frequency is output as the state recognition result.

[0097] Furthermore, the state identification module 15 in the system is also used for:

[0098] Acquire occurrence coefficients of special vibration signals and special operation characteristic coefficients in the plurality of special identification paths as a plurality of special occurrence coefficients;

[0099] Taking the special recognition path with a special occurrence coefficient less than or equal to the occurrence coefficient as a matching special recognition path, and obtaining multiple matching special recognition paths;

[0100] The vibration signal and the operation characteristic coefficient are respectively input into a plurality of matching special identification paths to obtain a plurality of similarities as a plurality of state similarities;

[0101] The special state identified by the matching special identification path corresponding to the largest state similarity is output as the state identification result.

[0102] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1The axial piston pump state detection method based on monitoring data fusion and the specific examples in Example 1 are also applicable to the axial piston pump state detection system based on monitoring data fusion in this example. Through the above detailed description of the axial piston pump state detection method based on monitoring data fusion, those skilled in the art can clearly know the axial piston pump state detection system based on monitoring data fusion in this example, so for the sake of brevity of the specification, it will not be described in detail here. For the system disclosed in the example, since it corresponds to the method disclosed in the example, the description is relatively simple, and the relevant parts can be referred to the method description.

[0103] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0104] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. Axial piston pump state detection method based on monitoring data fusion, characterized in that: The method comprises: During the operation of the axial piston pump to be tested, vibration signals of multiple pairs of piston pairs are collected to obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; Collecting oil suction characteristic information and oil discharge characteristic information of multiple pairs of piston pairs during the oil suction and oil discharge process to obtain multiple sets of operation characteristic information, and processing and calculating the multiple sets of operation characteristic information according to the swash plate angle of the axial piston pump to obtain multiple sets of operation characteristic coefficients; Based on the state detection data record of the axial piston pump, a state identifier is constructed, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes a plurality of typical identification paths, the special identification branch includes a plurality of special identification paths, each special identification path is constructed by a set of special vibration signals and a set of special operation characteristic coefficients, and includes a special state of the identification; Based on the multiple groups of vibration signals and the multiple groups of operation characteristic coefficients, a data query is performed in the state detection data record to obtain multiple vibration occurrence coefficients and multiple operation characteristic occurrence coefficients, and multiple occurrence coefficients are calculated, specifically, the state detection data record query is traversed, the number of data identical to the multiple groups of vibration signals is extracted as multiple vibration occurrence coefficients, the number of data consistent with the multiple groups of operation characteristic coefficients is obtained as the operation characteristic occurrence coefficients, and the multiple vibration occurrence coefficients and the multiple operation characteristic occurrence coefficients are weighted averaged to obtain multiple occurrence coefficients; Determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold value respectively; if so, obtain the typical level according to the occurrence coefficient decision, obtain the analysis quantity by index, input the vibration signal and the operation characteristic coefficient into the typical recognition path of the analysis quantity, and obtain the state recognition result; if not, obtain multiple matching special recognition paths according to the occurrence coefficient selection, input the vibration signal and the operation characteristic coefficient into multiple matching special recognition paths, obtain multiple state similarities, and output the special state with the largest state similarity as the state recognition result; Based on the multiple state recognition results, classify and obtain the state detection result of the axial piston pump; The oil suction characteristic information and the oil discharge characteristic information of the plurality of pairs of piston pairs in the oil suction and oil discharge process are collected to obtain a plurality of sets of operation characteristic information, and the plurality of sets of operation characteristic information are processed and calculated according to the swash plate angle of the axial piston pump, including: Collecting the oil suction flow information and the oil discharge flow information of multiple pairs of plunger pairs during the oil suction and oil discharge process as the oil suction characteristic information and the oil discharge characteristic information, and obtaining multiple sets of operation characteristic information; Acquire the swash plate angle of the axial piston pump, and acquire the cylinder volume information at the swash plate angle; Calculating ratios of the plurality of sets of operation characteristic information and the cylinder volume information to obtain a plurality of sets of operation characteristic coefficients; Among them, based on the state detection data record of the axial piston pump, a state identifier is constructed, including: Based on the state detection data record of the axial piston pump, a plurality of sample state detection data are collected and obtained, each sample state detection data includes a set of sample vibration signals, a set of sample operation characteristic coefficients and a sample detection state; Clustering the same data in the plurality of sample state detection data to obtain a plurality of sample data sets; Calculate the proportion of data in each sample data set in the total sample data to obtain multiple occurrence coefficients; According to the magnitudes of the multiple occurrence coefficients, determining whether they are greater than an occurrence coefficient threshold, classifying the multiple sample data sets to obtain a typical sample data set and a special sample data set; Using the typical sample data set, constructing a typical recognition branch including a plurality of typical recognition paths, using the special sample data set, constructing a special recognition branch including a plurality of special recognition paths, and obtaining a state identifier; The typical sample data set is used to construct a typical recognition branch including multiple typical recognition paths, including: Randomly extracting a first typical constructed data set from the typical sample data set, wherein the first typical constructed data set includes at least half of the sample data in the typical sample data set; Using the sample vibration signal and the sample operation characteristic coefficient in the first typical construction data set as input and the sample detection state as output, a first typical identification path is constructed; randomly extracting a second typical constructed data set from the typical sample data set again to construct a second typical recognition path, wherein the second typical constructed data set is different from the first typical constructed data set; Continue to construct and obtain multiple typical recognition paths to obtain the typical recognition branches; The special sample data set is used to construct a special recognition branch including multiple special recognition paths, including: Using the sample vibration signal and the sample operation characteristic coefficient in each group of different special sample data in the special sample data set as the special vibration signal and the special operation characteristic coefficient, and constructing a plurality of special recognition paths based on the twin network, wherein each special recognition path recognizes the similarity between the input vibration signal and the operation characteristic coefficient and the special vibration signal and the special operation characteristic coefficient; Using each different sample detection state in the special sample data set as a plurality of special states, marking the plurality of special identification paths, and obtaining the special identification branches; Among them, the typical level is obtained according to the occurrence coefficient decision, and the analysis quantity is obtained by indexing, and the vibration signal and the operation characteristic coefficient are input into the typical identification path of the analysis quantity to obtain the state identification result, including: Obtain a set of sample occurrence coefficients, evaluate and obtain a set of sample typical levels and a set of sample analysis quantities, and construct a typical level classifier based on a decision tree, where the size of the sample occurrence coefficient is positively correlated with the size of the sample typical level and negatively correlated with the size of the sample analysis quantity; Based on the typical grade classifier, decision classification is performed on the occurrence coefficients to obtain the typical grade, and the analysis quantity is obtained by indexing; Randomly selecting the typical identification path of the analyzed number from the multiple typical identification paths, inputting a group of vibration signals with an appearance coefficient greater than an appearance coefficient threshold and a group of operation characteristic coefficients, and identifying and obtaining the typical detection states of the analyzed number; The typical detection state with the highest occurrence frequency is output as the state recognition result.

2. The method according to claim 1, characterized in that: According to the occurrence coefficient, multiple matching special identification paths are selected, the vibration signal and the operation characteristic coefficient are input into the multiple matching special identification paths, multiple state similarities are obtained, and the special state with the largest state similarity is output as the state identification result, including: Acquire occurrence coefficients of special vibration signals and special operation characteristic coefficients in the plurality of special identification paths as a plurality of special occurrence coefficients; Taking the special recognition path with a special occurrence coefficient less than or equal to the occurrence coefficient as a matching special recognition path, and obtaining multiple matching special recognition paths; The vibration signal and the operation characteristic coefficient are respectively input into a plurality of matching special identification paths to obtain a plurality of similarities as a plurality of state similarities; The special state identified by the matching special identification path corresponding to the largest state similarity is output as the state identification result.

3. The axial piston pump state detection system based on monitoring data fusion is characterized by: For implementing the steps of the method according to any one of claims 1 to 2, the system comprises: A vibration signal acquisition module is used to collect vibration signals of multiple pairs of piston pairs during the operation of the axial piston pump to be tested, and obtain multiple groups of vibration signals, wherein one piston pair in each pair of piston pairs is sucking oil while the other piston pair is discharging oil, and each piston pair includes a piston and a cylinder; An operation characteristic acquisition module is used to acquire the oil suction characteristic information and the oil discharge characteristic information of multiple pairs of piston pairs during the oil suction and oil discharge process, obtain multiple groups of operation characteristic information, and process and calculate the multiple groups of operation characteristic information according to the swash plate angle of the axial piston pump to obtain multiple groups of operation characteristic coefficients; A state identifier building module is used to build a state identifier based on the state detection data record of the axial piston pump, wherein the state identifier includes a typical identification branch and a special identification branch, the typical identification branch includes a plurality of typical identification paths, the special identification branch includes a plurality of special identification paths, each special identification path is constructed by a set of special vibration signals and a set of special operation characteristic coefficients, and includes a special state of the identification; A data query module, for performing data query in the state detection data record based on the multiple groups of vibration signals and the multiple groups of operation characteristic coefficients, obtaining multiple vibration occurrence coefficients and multiple operation characteristic occurrence coefficients, and calculating and obtaining multiple occurrence coefficients, specifically traversing the state detection data record query, extracting the same number of data as the multiple groups of vibration signals as the multiple vibration occurrence coefficients, obtaining the same number of data as the multiple groups of operation characteristic coefficients as the operation characteristic occurrence coefficients, and performing weighted average calculation on the multiple vibration occurrence coefficients and the multiple operation characteristic occurrence coefficients to obtain the multiple occurrence coefficients; A state recognition module is used to determine whether the multiple occurrence coefficients are greater than the occurrence coefficient threshold value. If so, a typical level is obtained according to the occurrence coefficient decision, and the analysis quantity is obtained by indexing, and the vibration signal and the operation characteristic coefficient are input into the typical recognition path of the analysis quantity to obtain a state recognition result. If not, a plurality of matching special recognition paths are obtained according to the occurrence coefficient selection, and the vibration signal and the operation characteristic coefficient are input into the plurality of matching special recognition paths to obtain a plurality of state similarities, and a special state with the largest state similarity is output as the state recognition result; The result classification module is used to classify and obtain the state detection result of the axial piston pump based on multiple state recognition results.

Images