CN113586554A

CN113586554A - Hydraulic motor detection system and method

Info

Publication number: CN113586554A
Application number: CN202110780380.5A
Authority: CN
Inventors: 罗绍卓; 邢柳; 陈龙; 崔楷华; 黄利云
Original assignee: Zoomlion Heavy Industry Science and Technology Co Ltd; Shanghai Zoomlion Piling Machinery Co Ltd
Current assignee: Zoomlion Heavy Industry Science and Technology Co Ltd; Shanghai Zoomlion Piling Machinery Co Ltd
Priority date: 2021-07-09
Filing date: 2021-07-09
Publication date: 2021-11-02

Abstract

The invention discloses a hydraulic motor detection system, which is applied to the quality inspection of a hydraulic motor before delivery, and comprises a sensor for testing parameters of the hydraulic motor, and a judgment device connected with the sensor and receiving parameter data of the hydraulic motor, wherein an unsupervised learning model and a classification model are arranged in the judgment device, the unsupervised learning model is used for establishing a normal parameter interval according to the parameter data and screening abnormal parameter data which are not in the normal interval in the parameter data, and the classification model is used for diagnosing the abnormal parameter data and comparing whether the abnormal parameter data accord with fault characteristics. The invention also discloses a hydraulic motor detection method, which is applied to the quality inspection of the hydraulic motor before leaving a factory, and comprises the step S1 of collecting the test parameter data of the hydraulic motor; step S2, establishing an unsupervised learning model and a classification model according to the parameter data; step S3, inputting parameter data to the established unsupervised learning model and classification model; in step S4, the classification model outputs the corresponding fault type.

Description

Hydraulic motor detection system and method

Technical Field

The invention relates to the technical field of hydraulic motor equipment, in particular to a hydraulic motor detection system and a hydraulic motor detection method.

Background

The hydraulic motor is an actuator of the hydraulic system, which converts the hydraulic pressure energy provided by the hydraulic pump into mechanical energy of its output shaft. In order to ensure the product quality, a series of delivery tests including the necessary items such as a displacement test, a volumetric efficiency test, a variable characteristic test and an external leakage test are required to be carried out before the delivery of the hydraulic motor. And items of spot-testing, such as total efficiency tests, impact tests, overload tests. After the product is off-line, 100% of the products are subjected to necessary test items, and after random sampling or system sampling, a small part of the products are subjected to sampling test items.

However, in the above test items, all the qualified judgments depend on design, production and use experiences, and are greatly influenced by the operation of test workers in the actual operation process. When a problem product occurs, a part of the product needs to be disassembled to confirm the problem, and the cause of the problem is tracked by related technologies and production personnel, so that the working hours and the labor cost are increased. So that the total factory testing cost of the motor accounts for more than 30% of the total cost of the motor at present.

The foregoing description is provided for general background information and is not admitted to be prior art.

Disclosure of Invention

The invention aims to provide a hydraulic motor detection system and method for improving detection efficiency and accuracy.

The invention provides a hydraulic motor detection system, which is applied to the quality inspection of a hydraulic motor before delivery, and comprises a sensor for testing parameters of the hydraulic motor, and a judgment device connected with the sensor and receiving parameter data of the hydraulic motor, wherein an unsupervised learning model and a classification model are arranged in the judgment device, the unsupervised learning model is used for establishing a normal parameter interval according to the parameter data and screening abnormal parameter data which are not in the normal interval in the parameter data, and the classification model is used for diagnosing the abnormal parameter data and comparing whether the abnormal parameter data accord with fault characteristics.

The system further comprises a data storage device and a calculation device which are connected with the judgment device, wherein the data storage module is used for storing the parameter data, and the calculation device is used for optimizing the unsupervised learning model and the classification model according to the parameter data.

Furthermore, the system also comprises a human-computer interaction device which is connected with the judging device and used for displaying the parameter data.

Further, the sensor includes one or more of a pressure sensor, a noise sensor, an oil temperature sensor, a flow sensor, and a vibration sensor.

Further, the unsupervised model is One of One Class SVM, Isolation Forest and Local outer factor, and the classification model is One of Random Forest, SVM, XGboost and collaborative filtering.

The invention also provides a hydraulic motor detection method, which is applied to the quality inspection of the hydraulic motor before leaving the factory and comprises the following steps:

step S1, collecting test parameter data of the hydraulic motor;

step S2, establishing an unsupervised learning model and a classification model according to the parameter data, wherein the unsupervised learning model determines a normal parameter interval according to the parameter data;

step S3, inputting the parameter data to the unsupervised learning model and the classification model which are established, wherein the unsupervised learning model screens out abnormal parameter data which are not in the normal parameter interval in the parameter data, and the classification model compares whether the abnormal parameter data accord with fault characteristics;

step S4, the classification model outputs the corresponding fault type.

Further, in the step S1, one or more of pressure, oil temperature, noise, flow rate and vibration parameters of the hydraulic motor are collected.

Further, the step S3 includes a step S31 and a step S32, and the step S31 is to divide the continuous parameter data into a plurality of pieces of feature data having an overlapping rate of 0.5% to 0.75%; the step S32 is to input the segmented feature data into the unsupervised model and the classification model.

Further, the step S32 includes a step S321 and a step S322, where the step S321 is a preliminary judgment, and abnormal parameter data that is not in the normal parameter interval in the feature data is screened out through the unsupervised model; in the step S322, for final judgment, whether the abnormal parameter data meets the fault characteristics is compared by the classification model.

Further, the method further comprises a step S5 of manually disassembling and confirming the hydraulic motor corresponding to the abnormal parameter data which is in accordance with the fault characteristics in the step S322, and optimizing the unsupervised model and the classification model by using a disassembly conclusion as a sample.

According to the hydraulic motor detection system and method provided by the invention, the parameters of the hydraulic motor are collected, the unsupervised model and the classification model are constructed by using the parameter data, the unsupervised model is used for screening abnormal data different from other data in the parameter data, and the classification model is used for finally judging whether the fault and the fault type exist, so that the machine learning algorithm is used for replacing manpower, the manpower resource requirement and the manpower cost are reduced, the production efficiency is improved, and the detection precision is improved. And the detection precision of the model is continuously improved through a large number of test data iteration models of the fault hydraulic motor.

Drawings

FIG. 1 is a schematic diagram of a connection of a hydraulic motor detection system according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart illustrating a hydraulic motor detection method according to an embodiment of the present invention;

FIG. 3 is a schematic flow chart of an implementation of the hydraulic motor detection method of FIG. 2;

fig. 4 is a schematic flow chart of an implementation of a conventional hydraulic motor detection method.

Detailed Description

The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

Referring to fig. 1, a hydraulic motor detection system according to an embodiment of the present invention includes a sensor 10, a determination device 20, a human-computer interaction device 30, a data storage device 40, a calculation device 50, and a management device 60. After production and off-line, the hydraulic motor (not shown) enters a detection link, normal working conditions are simulated, and the sensor 10 acquires test process parameter data of the hydraulic motor to be tested. In the present embodiment, the sensor 10 includes a pressure sensor, a noise sensor, an oil temperature sensor, a flow sensor, and a vibration sensor. In other embodiments, the sensor 10 may include one or more of the above parameter sensors, and sensors for acquiring other parameter signals may be added according to actual needs.

For acquiring the vibration signal of the hydraulic motor, a non-contact laser vibration sensor is preferred, and a contact vibration sensor or a speed sensor is selected secondarily. The non-contact sensor can reduce the loss of the sensor, reduce the time for assembling and disassembling the sensor and improve the detection working efficiency. For collecting noise signals of the hydraulic motor, an acoustic array sensor is preferred, and a directional acoustic sensor is adopted secondly.

Referring to fig. 1, the determining device 20 and the human-computer interaction device 30 are both located at an edge (a left dotted line in the figure), the data storage device 40, the computing device 50 and the management device 60 are located at a cloud (a right dotted line in the figure), and the determining device 20 and the human-computer interaction device 30 may be integrated in a fixed terminal or a mobile terminal. The judging device 20 is connected to the sensor 10, and is configured to receive the parameter data collected by the sensor 10, and upload the parameter data to the data storage device 40 in the cloud for storage. The calculation means 50 extracts the parameter data from the data storage means 40 and constructs an unsupervised learning model and a classification model, and issues the models to the judgment means 20.

The unsupervised learning model may be One of anomaly detection algorithms such as One Class SVM (support vector machines), Isolation Forest, Local Outlier factor, and the like. The One Class SVM is characterized in that a hyperplane is found to circle out positive examples in a sample, and then the hyperplane is used for making a decision, and the sample in the circle is regarded as a positive sample.

In the Isolation Forest, a feature is randomly selected during detection, and then a section is randomly selected from the maximum value and the minimum value of the selected feature. The training of the whole training set under the algorithm is like a tree, and the algorithm is divided recursively. The number of divisions is equal to the path distance d from the root node to the leaf node. The average of d of all random trees is the final result of our test function.

Local Outlier factor reflects the degree of abnormality of a sample by calculating a numerical value score. This value has the approximate meaning: the average density of the positions of the sample points around a sample point is higher than the density of the positions of the sample points. The more the ratio is greater than 1, the less the density of the location of the point is than the locations of the surrounding samples, and the more likely the point is an outlier.

The classification model may be one of classification algorithms such as Random Forest, SVM, XGBoost (eXtreme Gradient Boosting), collaborative filtering, and the like. Random Forest consists of many decision trees, and there is no correlation between each decision tree. After the forest is built, when a new sample enters, each decision tree is judged respectively, and then a classification result is given out based on a voting method.

The SVM is a binary classification model whose basic model is a linear classifier defined at the largest interval on the feature space, which makes it different from the perceptron. The basic idea of SVM learning is to solve for the separating hyperplane that correctly partitions the training data set and has the largest geometric separation. For linearly separable data sets, there are an infinite number of such hyperplanes (i.e., perceptrons), but the separated hyperplane with the largest geometric separation is unique.

The core algorithm of the XGboost is to continuously add trees, continuously perform feature splitting to grow a tree, and each time a tree is added, actually learn a new function f (x) to fit the residual error predicted last time. When training is completed to obtain k trees, a score of a sample is predicted, namely, according to the characteristics of the sample, a corresponding leaf node is fallen in each tree, and each leaf node corresponds to a score. Finally, the score corresponding to each tree only needs to be added up to be the predicted value of the sample.

The unsupervised learning model in the decision device 20 has been subjected to a number of iterative optimizations by the calculation device 50 to determine the normal parameter interval for each parameter. When the sensor 10 transmits the parameter data to the determination device 20, the determination device 20 first divides the continuous data stream into time interval construction features with an overlap rate of 0.5% to 0.75%, and suggests 1 to 5 seconds, which is a hyper-parameter obtained by mesh optimization of the final model. And comparing the segmented characteristic data with the normal parameter interval through an unsupervised learning model, screening abnormal parameter data which are not in the normal parameter interval, and determining that the hydraulic motor corresponding to the normal parameter data which are not screened is qualified for detection.

And the abnormal parameter data are continuously compared through a classification model to determine whether the abnormal parameter data accord with the fault characteristics and the specific fault characteristics, wherein the faults include but are not limited to valve core clamping stagnation of a variable mechanism, missing installation or abnormal abrasion of a plunger ring of a heart part, abrasion between an oil distribution disc and a rotor, and cavitation erosion of a plunger or a rotating pair. The final diagnosis result is displayed in real time by the human-computer interaction device 30 connected with the judgment device 20, and meanwhile, the human-computer interaction device 30 is also used for controlling the judgment device 20 to select a hydraulic motor which needs to participate in the sampling inspection items such as the total efficiency test, the impact test, the overload test and the like from the abnormal parameter data. And the hydraulic motor corresponding to the abnormal parameter data which do not accord with the fault characteristics and are not selected for sampling inspection in the classification model is also determined to be qualified for inspection.

And (3) manually disassembling the hydraulic motors which accord with the fault characteristics and are selected to participate in the spot inspection in the classification model, inputting the disassembled data into the computing device 50 as training samples, and performing iterative optimization on the unsupervised learning model and the classification model as the training samples. The management device 60 is used for supervising the sensor 10, the judgment device 20, the human-computer interaction device 30, the data storage device 40 and the computing device 50 and providing OTA upgrading function.

The hydraulic motor detection system provided by this embodiment collects parameters of the hydraulic motor through the sensor 10, constructs an unsupervised model and a classification model by using the parameter data, screens abnormal data different from other data in the parameter data by using the unsupervised model, finally judges whether there is a fault and a fault type by using the classification model, replaces manpower with a machine learning algorithm, reduces manpower resource requirements and labor cost, improves production efficiency, and improves detection precision. The non-contact sensor 10 can reduce the loss of the sensor 10, reduce the time for assembling and disassembling the sensor 10, and further improve the detection work efficiency. And the detection precision of the model is continuously improved through a large number of test data iteration models of the fault hydraulic motor.

Referring to fig. 2, the present embodiment further provides a hydraulic motor detection method applied to pre-factory quality inspection of a hydraulic motor, including steps S1 to S5. Step S1 is to collect various parameter data of the hydraulic motor in the simulated working condition on the detection line, where the parameter may be one or more of pressure, oil temperature, noise, flow rate, and vibration, and all the five parameters are collected in this embodiment.

Step S2 is to construct an unsupervised learning model and a classification model from the parameter data, and determine a normal parameter interval of the unsupervised learning model and a classification basis of the classification model, that is, a specific fault feature. The unsupervised learning model can be One of abnormal point detection algorithms such as One Class SVM, Isolation Forest, Local outer factor and the like. The classification model can be one of classification algorithms such as Random Forest, SVM, XGboost, collaborative filtering and the like.

The step S3 includes steps S31 and S32, and the step S31 is to divide the continuous parameter data into a plurality of pieces of feature data having an overlap ratio of 0.5% to 0.75%. Step S32 includes step S321 and step S322, where step S321 is a preliminary determination stage, the segmented feature data is input into the unsupervised learning model, and abnormal parameter data that is not in the normal parameter interval in the feature data is screened out through the unsupervised learning model. Step S322 is a final judgment stage, in which the abnormal parameter data is input into the classification model, the classification model compares whether the abnormal parameter data meets the fault characteristics, and meanwhile, a part of the abnormal parameter data is extracted from the hydraulic motor corresponding to the abnormal parameter data to perform sampling inspection items such as a total efficiency test, an impact test, an overload test, and the like.

Step S4 is to output the corresponding fault type for the classification model, and the hydraulic motor corresponding to the abnormal parameter data that does not match with the characteristic fault type is determined as a qualified product. Step S5 is to perform manual disassembly and verification on the hydraulic motor corresponding to the abnormal parameter data matched with the fault feature in step S322, that is, the unqualified hydraulic motor, and iteratively optimize the unsupervised learning model and the classification model using the disassembly conclusion as sample data, which is used as a reference for the fault feature in the classification model, thereby improving the detection accuracy of the unsupervised learning model and the classification model.

In the conventional hydraulic motor detection method, as shown in fig. 4, after a product is offline, necessary items such as a displacement test, a volumetric efficiency test, a variable characteristic test, an external leakage test and the like are performed by 100%. And randomly sampling products which need to be qualified, and carrying out sampling inspection items such as a total efficiency test, an impact test, an overload test and the like. The necessary inspection and the spot inspection both depend on manual experience, when a problem product occurs, a part of the products need to be disassembled to confirm the problem, and the cause of the problem is tracked by related technologies and production personnel.

As shown in FIG. 3, in the method, after the product is off-line, 100% of parameters are collected, an unsupervised learning model is used as a preliminary judgment, abnormal parameter data is screened out for sampling inspection, and a classification model is used for final judgment. The results of the preliminary judgment and the final judgment are displayed through the human-computer interaction device 30. And finally, the screened unqualified products are manually disassembled to be used as training samples to optimize the unsupervised learning model and the classification model.

Compared with the existing detection method, the hydraulic motor detection system and method provided by the embodiment have the advantages that the parameters of the hydraulic motor are collected through the sensor 10, the unsupervised model and the classification model are built through the parameter data, the unsupervised model is used for screening abnormal data different from other data in the parameter data, the classification model is used for finally judging whether the fault and the fault type exist, the machine learning algorithm is used for replacing manpower, the human resource requirement and the labor cost are reduced, the production efficiency is improved, and the detection precision is improved. The non-contact sensor 10 can reduce the loss of the sensor 10, reduce the time for assembling and disassembling the sensor 10, and further improve the detection work efficiency. And the detection precision of the model is continuously improved through a large number of test data iteration models of the fault hydraulic motor.

In the drawings, the size and relative sizes of layers and regions may be exaggerated for clarity. It will be understood that when an element such as a layer, region or substrate is referred to as being "formed on," "disposed on" or "located on" another element, it can be directly on the other element or intervening elements may also be present. In contrast, when an element is referred to as being "directly formed on" or "directly disposed on" another element, there are no intervening elements present.

In this document, the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", "vertical", "horizontal", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for the purpose of clarity and convenience of description of the technical solutions, and thus, should not be construed as limiting the present invention.

In this document, unless expressly stated or limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms can be understood in a specific case to those of ordinary skill in the art.

As used herein, the ordinal adjectives "first", "second", etc., used to describe an element are merely to distinguish between similar elements and do not imply that the elements so described must be in a given sequence, either temporally, spatially, in ranking, or in any other manner.

As used herein, the meaning of "a plurality" or "a plurality" is two or more unless otherwise specified.

As used herein, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, including not only those elements listed, but also other elements not expressly listed.

The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.

Claims

1. A hydraulic motor detection system is applied to quality inspection of a hydraulic motor before delivery, and is characterized by comprising a sensor for testing parameters of the hydraulic motor and a judging device connected with the sensor and receiving parameter data of the hydraulic motor, wherein an unsupervised learning model and a classification model are arranged in the judging device, the unsupervised learning model is used for establishing a normal parameter interval according to the parameter data and screening abnormal parameter data which are not in the normal interval in the parameter data, and the classification model is used for diagnosing the abnormal parameter data and comparing whether the abnormal parameter data accord with fault characteristics.

2. A hydraulic motor testing system according to claim 1, further comprising data storage means connected to said determining means for storing said parametric data and computing means for optimizing said unsupervised learning model and said classification model based on said parametric data.

3. The hydraulic motor testing system of claim 1, further comprising a human-machine interface coupled to said determining means for displaying said parametric data.

4. The hydraulic motor detection system of claim 1, wherein the sensor comprises one or more of a pressure sensor, a noise sensor, an oil temperature sensor, a flow sensor, a vibration sensor.

5. The hydraulic motor detection system of claim 1, wherein the unsupervised model is One of One Class SVM, Isolation Forest, and Local outer factor, and the classification model is One of Random Forest, SVM, XGBoost, and collaborative filtering.

6. A hydraulic motor detection method is applied to quality inspection before delivery of a hydraulic motor, and is characterized by comprising the following steps:

step S1, collecting test parameter data of the hydraulic motor;

step S4, the classification model outputs the corresponding fault type.

7. The hydraulic motor detection method of claim 6, wherein the step S1 is to collect one or more of pressure, oil temperature, noise, flow rate and vibration parameters of the hydraulic motor.

8. The hydraulic motor testing method as claimed in claim 6, wherein the step S3 includes steps S31 and S32, and the step S31 is to divide the continuous parameter data into a plurality of pieces of feature data having an overlapping rate of 0.5% to 0.75%; the step S32 is to input the segmented feature data into the unsupervised model and the classification model.

9. A hydraulic motor detecting method according to claim 8, wherein the step S32 includes a step S321 and a step S322, the step S321 is a preliminary judgment, and abnormal parameter data which is not in the normal parameter interval in the feature data is screened out by the unsupervised model; in the step S322, for final judgment, whether the abnormal parameter data meets the fault characteristics is compared by the classification model.

10. The hydraulic motor detection method according to claim 9, further comprising a step S5 of performing manual disassembly confirmation on the hydraulic motor corresponding to the abnormal parameter data that meets the fault characteristics in the step S322, and using the disassembly conclusion as a sample to optimize the unsupervised model and the classification model.