Crane fault diagnosis method, system, equipment and medium based on data mining

Abstract

The present application relates to the technical field of crane fault diagnosis, and in particular to a crane fault diagnosis method, system, device and medium based on data mining, wherein the method comprises: collecting time series monitoring data of the crane, inputting the time series monitoring data into a correction model generated by training, and obtaining time series correction data output by the correction model; calculating time series difference data between the time series correction data and the time series monitoring data, obtaining first representative data based on the time series difference data, and obtaining second representative data based on the time series monitoring data; using the first representative data and the second representative data to form comprehensive representative data, inputting the comprehensive representative data into an analysis model generated by training, and obtaining diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal. The present application can improve the intelligence of crane fault diagnosis.

Description

Technical Field

[0001] The present application relates to the technical field of crane fault diagnosis, and in particular to a crane fault diagnosis method, system, equipment and medium based on data mining. Background Art

[0002] Crane is an important engineering equipment, mainly used for lifting, transporting and loading and unloading heavy objects. In order to improve the working efficiency of the crane and extend the service life of the crane, it is necessary to conduct timely diagnosis of the faults occurring in the crane.

[0003] A similar prior art includes a Chinese patent application with publication number CN105174065A, which relates to a crane fault monitoring method and device, wherein the crane fault monitoring method includes a transmission chain breakage fault monitoring method, and the transmission chain breakage fault monitoring method includes the following steps: A1. Collecting the speed signal of the crane motor and the speed signal of the drum respectively to determine the motor speed and the drum speed; A2. If the motor speed and the drum speed are not zero, calculating the transmission ratio of the motor speed and the drum speed, and comparing it with the preset value of the transmission ratio; A3. If the transmission ratio exceeds the preset value of the transmission ratio, determining that a transmission chain breakage fault occurs between the motor and the drum. In addition, similar prior art also includes a Chinese patent application with publication number CN116768059A, which discloses a crane line fault diagnosis system and method, the system includes a display, and a detection module and a detection circuit deployed in the crane control box; the detection circuit is a circuit drawn from the wiring point of the non-control signal of the crane control box, the detection circuit is electrically connected to the detection module, and is used to send the detected non-control signal to the detection module; the detection module is used to perform fault diagnosis on the received non-control signal, and transmit the fault diagnosis result and the received non-control signal to the display; the display is used to parse and display the received fault diagnosis result. However, the fault diagnosis methods in the above two patent applications lack certain intelligence. Summary of the invention

[0004] The application collects the time series monitoring data of the crane, inputs the time series monitoring data into the correction model, obtains the time series correction data output by the correction model, calculates the time series difference data, obtains the first representative data based on the time series difference data, obtains the second representative data based on the time series monitoring data, so as to generate comprehensive representative data, inputs the comprehensive representative data into the analysis model, and obtains the diagnosis result data. The application aims to improve the intelligence of crane fault diagnosis.

[0005] The present application provides a crane fault diagnosis method based on data mining, comprising the following steps:

[0006] The monitoring module collects the time series monitoring data of the crane, transmits the collected time series monitoring data to the diagnosis module through the network module, and after receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training to obtain the time series correction data output by the correction model;

[0007] The diagnosis module calculates time series difference data between the time series correction data and the time series monitoring data, obtains first representative data based on the time series difference data, and the diagnosis module obtains second representative data based on the time series monitoring data;

[0008] The diagnosis module uses the first representative data and the second representative data to form comprehensive representative data, and the diagnosis module inputs the comprehensive representative data into the analysis model generated by training to obtain the diagnosis result data output by the analysis model on whether the time series monitoring data is abnormal.

[0009] As a preferred technical solution of the present application, before the diagnosis module is trained to generate the correction model and the analysis model, the following steps are included:

[0010] The diagnostic module collects a number of time series historical monitoring data from the monitoring module;

[0011] For each collected time series historical monitoring data, the diagnosis module sets identification data corresponding to the time series historical monitoring data, the content of the identification data is a normal situation or a fault situation, and the time series historical monitoring data and the identification data are stored correspondingly.

[0012] As a preferred technical solution of the present application, the diagnostic module training generates a correction model and an analysis model, including the following steps:

[0013] The diagnostic module selects several combinations including identification data of normal conditions from among all the stored combinations of time series historical monitoring data and identification data, extracts several time series historical monitoring data from the selected combinations, and uses the extracted several time series historical monitoring data to train and generate a correction model;

[0014] For each stored combination of time series historical monitoring data and identification data, the diagnosis module inputs the time series historical monitoring data contained in the combination into the correction model generated by training to obtain the time series historical correction data output by the correction model;

[0015] For each time series historical correction data, the diagnosis module calculates time series historical difference data between the time series historical correction data and the time series historical monitoring data corresponding to the time series historical correction data, obtains first historical representative data based on the time series historical difference data, obtains second historical representative data based on the time series historical monitoring data corresponding to the time series historical correction data, and forms comprehensive historical representative data using the first historical representative data and the second historical representative data;

[0016] For each comprehensive historical representative data, the diagnostic module sets label data for the comprehensive historical representative data as to whether the time series historical monitoring data corresponding to the comprehensive historical representative data is abnormal, records the combination of the comprehensive historical representative data and the label data as training data, and the diagnostic module uses all the training data to train and generate an analysis model.

[0017] As a preferred technical solution of the present application, the diagnostic module uses all the training data to train and generate an analysis model, including the following steps:

[0018] The diagnostic module is trained to generate candidate analysis models, and the diagnostic module is trained to generate recognition models. The diagnostic module also determines whether the number of executions of this step has reached a preset execution number threshold. If it has been reached, the candidate analysis model generated by the last training is used as the analysis model and the execution of this step is terminated. If it has not been reached, this step is repeated.

[0019] As a preferred technical solution of the present application, the diagnostic module training generates a candidate analysis model, including the following steps:

[0020] The diagnosis module obtains all the training data, and the diagnosis module determines whether it is the first time to generate the candidate analysis model. If it is the first time to generate the candidate analysis model, the diagnosis module uses all the training data to train and generate the candidate analysis model. If it is not the first time to generate the candidate analysis model, the diagnosis module proceeds to the next step.

[0021] The diagnostic module obtains the recognition model generated by the most recent training, inputs all the training data into the obtained recognition model in sequence, obtains the recognition result data output by the obtained recognition model respectively, and the diagnostic module determines a number of inappropriate training data from all the training data based on all the recognition result data, and uses all the other training data from all the training data except the several inappropriate training data to train and generate a candidate analysis model.

[0022] As a preferred technical solution of the present application, the diagnosis module training generates a recognition model, including the following steps:

[0023] The diagnosis module obtains all the training data, obtains the candidate analysis model generated by the most recent training, and divides all the training data into different training data groups according to the different contents of the included labeled data;

[0024] With respect to each training data group, for each training data in the training data group, the diagnosis module inputs the training data into the acquired candidate analysis model, and obtains the analysis result data output by the acquired candidate analysis model;

[0025] With respect to each training data group, for each training data in the training data group, the diagnosis module calculates the difference between the content of the label data contained in the training data and the content of the analysis result data corresponding to the training data to obtain deviation data, and in each training data group, a number of training data whose content of the corresponding deviation data is less than a preset threshold are divided into a first set, and a number of training data whose content of the corresponding deviation data is greater than or equal to the preset threshold are divided into a second set;

[0026] The diagnostic module sets the content of each training data in the first set as allowed flag data, and uses the training data and the flag data to form practice data. For each training data in the second set, the diagnostic module sets the content of the training data as prohibited flag data, and uses the training data and the flag data to form practice data. The diagnostic module generates a recognition model based on all the practice data training.

[0027] This application also provides a crane fault diagnosis system based on data mining, including the following modules:

[0028] A monitoring module, used for storing a plurality of time series historical monitoring data of the crane, and for collecting the time series monitoring data of the crane, and transmitting the collected time series monitoring data to the diagnosis module through the network module;

[0029] A network module, used for transmitting data between the monitoring module and the diagnosis module;

[0030] The diagnostic module is used to input the time series monitoring data into the correction model generated by training after receiving the time series monitoring data, obtain the time series correction data output by the correction model, and calculate the time series difference data between the time series correction data and the time series monitoring data, obtain the first representative data based on the time series difference data, obtain the second representative data based on the time series monitoring data, and use the first representative data and the second representative data to form comprehensive representative data, input the comprehensive representative data into the analysis model generated by training, and obtain the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal.

[0031] The present invention also provides a device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement any one of the above-mentioned methods when executing the computer program.

[0032] The present invention also provides a medium, wherein the medium stores program instructions, wherein when the program instructions are executed, the device where the medium is located is controlled to execute any one of the above methods.

[0033] Compared with the prior art, the beneficial effects of the present application are at least as follows:

[0034] In the technical solution provided by the present application, first, the monitoring module collects the time series monitoring data of the crane, transmits the collected time series monitoring data to the diagnosis module through the network module, and after receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training, and obtains the time series correction data output by the correction model. Secondly, the diagnosis module calculates the time series difference data between the time series correction data and the time series monitoring data, obtains the first representative data based on the time series difference data, and the diagnosis module obtains the second representative data based on the time series monitoring data. Finally, the diagnosis module uses the first representative data and the second representative data to form comprehensive representative data, and the diagnosis module inputs the comprehensive representative data into the analysis model generated by training, and obtains the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal. Through this application, not only can the intelligent level of crane fault diagnosis be significantly improved, but also the accuracy of the crane fault diagnosis results can be ensured, so that the crane fault can be discovered in time and the loss caused by the crane fault can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0036] Figure 1 This is a flow chart of a crane fault diagnosis method based on data mining in an embodiment of the present application;

[0037] Figure 2 A flowchart of a method for training and generating a correction model and an analysis model in an embodiment of the present application;

[0038] Figure 3 A flowchart of a method for training and generating a candidate analysis model in an embodiment of the present application;

[0039] Figure 4Schematic diagram of a crane fault diagnosis system based on data mining in an embodiment of the present application. DETAILED DESCRIPTION

[0040] The embodiments of the present application provide a crane fault diagnosis method, system, device and medium based on data mining. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0041] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 The crane fault diagnosis method based on data mining in the embodiment of the present application includes the following main steps:

[0042] Step 1: The monitoring module collects the time series monitoring data of the crane, and transmits the collected time series monitoring data to the diagnosis module through the network module. After receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training to obtain the time series correction data output by the correction model;

[0043] Step 2: The diagnosis module calculates time series difference data between the time series correction data and the time series monitoring data, obtains first representative data based on the time series difference data, and obtains second representative data based on the time series monitoring data;

[0044] Step 3: The diagnosis module uses the first representative data and the second representative data to form comprehensive representative data, and the diagnosis module inputs the comprehensive representative data into the analysis model generated by training to obtain the diagnosis result data output by the analysis model on whether the time series monitoring data is abnormal.

[0045] Specifically, in order to improve the intelligence level of crane fault diagnosis, it is considered to use machine learning, one of the multiple core technologies of data mining, to perform crane fault diagnosis. In step 1, the monitoring module collects the current time series monitoring data of the crane. For the sake of ease of understanding, for example, a vibration sensor is set on the key component of the crane. The time series monitoring data can be the vibration acceleration data of the most recent minute collected by the vibration sensor. The monitoring module transmits the collected time series monitoring data to the diagnosis module through the network module. After receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training to obtain the time series correction data output by the correction model. It should be noted that the time series correction data is the data collected by the vibration sensor under the normal working conditions of the crane. The normal working condition is also the fault-free working condition. How to train and generate the correction model will be introduced below. In step 2, the diagnosis module calculates the time series difference data between the time series correction data and the time series monitoring data, and obtains the first representative data based on the time series difference data. For ease of understanding, for example, the time series correction data, the time series monitoring data, and the time series difference data are displayed in the form of a waveform diagram. It is assumed that in the waveform diagram corresponding to the time series correction data, all parts are displayed as waveform changes, the horizontal axis represents time, and the vertical axis represents vibration acceleration data. It is assumed that in the waveform diagram corresponding to the time series monitoring data, the front part is displayed as waveform changes, and the rear part is displayed as an almost straight line, the horizontal axis represents time, and the vertical axis represents vibration acceleration data. Speed ​​data, then in the waveform diagram corresponding to the time series difference data, the front part is displayed as an almost straight line, and the rear part is displayed as a waveform change. The horizontal axis represents time, and the vertical axis represents the difference data of the two vibration acceleration data at the corresponding time points. Specifically, the average amplitude of the front part and the average amplitude of the rear part can be calculated, and the difference between the two average amplitudes is used as the first representative data. The diagnostic module also obtains the second representative data based on the time series monitoring data, and does not limit the method of obtaining the second representative data. For example, the neural network in the prior art can be used to extract the local features of the time series monitoring data as the second representative data, which will not be elaborated here. In step 3, the diagnosis module uses the first representative data and the second representative data to form comprehensive representative data. The first representative data and the second representative data can be expressed in the form of vectors. The comprehensive vector is obtained by simply merging the two vectors. The dimension of the comprehensive vector is the sum of the dimensions of the two vectors. The comprehensive vector is the comprehensive representative data. The diagnosis module inputs the comprehensive representative data into the trained analysis model to obtain the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal. How to train and generate the analysis model will be described below. For ease of understanding, for example, the diagnostic result data is 0.7, and 0.7 corresponds to the probability that the time series monitoring data is abnormal. Since it is greater than the threshold value of 0.6, it is considered that the crane has a fault.

[0046] Furthermore, before the diagnosis module is trained to generate the correction model and the analysis model, the following steps are included:

[0047] Step 1: The diagnosis module collects several time series historical monitoring data from the monitoring module;

[0048] Step 2: For each collected time series historical monitoring data, the diagnosis module sets identification data corresponding to the time series historical monitoring data. The content of the identification data is a normal situation or a fault situation, and the time series historical monitoring data and the identification data are stored correspondingly.

[0049] Specifically, the preparatory work performed by the diagnosis module before training and generating the correction model and the analysis model is introduced. In step 1, because the monitoring module can store a certain amount of time series historical monitoring data, the diagnosis module can collect a number of time series historical monitoring data from the monitoring module. The time series historical monitoring data refers to, for example, data collected by a vibration sensor in the past. In step 2, for each collected time series historical monitoring data, the diagnosis module sets identification data corresponding to the time series historical monitoring data. The content of the identification data is true, and can be a normal situation or a fault situation. The normal situation corresponds to the normal working condition of the crane, and the fault situation corresponds to the faulty working condition of the crane. The diagnosis module stores the time series historical monitoring data and the identification data accordingly.

[0050] Furthermore, the diagnosis module is trained to generate a correction model and an analysis model, including the following steps:

[0051] Step 1: The diagnosis module selects several combinations of the identification data in which the content is normal from all the stored combinations of the time series historical monitoring data and the identification data, extracts several time series historical monitoring data from the selected combinations, and uses the extracted several time series historical monitoring data to train and generate a correction model;

[0052] Step 2: For each stored combination of time series historical monitoring data and identification data, the diagnosis module inputs the time series historical monitoring data contained in the combination into the correction model generated by training to obtain the time series historical correction data output by the correction model;

[0053] Step 3: For each time series historical correction data, the diagnosis module calculates the time series historical difference data between the time series historical correction data and the time series historical monitoring data corresponding to the time series historical correction data, obtains the first historical representative data based on the time series historical difference data, obtains the second historical representative data based on the time series historical monitoring data corresponding to the time series historical correction data, and forms the comprehensive historical representative data using the first historical representative data and the second historical representative data;

[0054] Step 4: For each comprehensive historical representative data, the diagnosis module sets label data for the comprehensive historical representative data as to whether the time series historical monitoring data corresponding to the comprehensive historical representative data is abnormal, and records the combination of the comprehensive historical representative data and the label data as training data, and the diagnosis module uses all the training data to train and generate an analysis model.

[0055] Specifically, refer to Figure 2 As shown, the main steps of training the diagnosis module to generate the correction model and the analysis model are introduced. In the first step, the diagnosis module first selects several combinations containing identification data of the normal situation from all the stored combinations consisting of time series historical monitoring data and identification data, and then extracts several time series historical monitoring data from the selected several combinations, so that the diagnosis module can use the extracted several time series historical monitoring data to train and generate the correction model. This is done to enable the correction model to fully learn the characteristics and laws of the data collected by the vibration sensor under the normal working conditions of the crane. In the second step, for each stored combination consisting of time series historical monitoring data and identification data, the diagnosis module inputs the time series historical monitoring data contained in the combination into the correction model generated by the training, and obtains the time series historical correction data output by the correction model. In order to understand the meaning of the time series historical correction data, you can refer to the meaning of the time series correction data. In step 3, for each time series historical correction data, the diagnosis module calculates the time series historical difference data between the time series historical correction data and the corresponding time series historical monitoring data, the corresponding time series historical monitoring data refers to the time series historical monitoring data corresponding to the time series historical correction data, the diagnosis module obtains the first historical representative data based on the time series historical difference data, the method for obtaining the first representative data has been introduced above, and the method for obtaining the first historical representative data is the same as this method, the diagnosis module obtains the second historical representative data according to the time series historical monitoring data corresponding to the time series historical correction data, the method for obtaining the second representative data has also been introduced above, and the method for obtaining the second historical representative data is the same as this method, the diagnosis module uses the first historical representative data and the second historical representative data to form comprehensive historical representative data, the method for obtaining comprehensive representative data has also been introduced above, and the method for obtaining comprehensive historical representative data is the same as this method. In step 4, for each comprehensive historical representative data, the diagnosis module sets label data for the comprehensive historical representative data. The content of the label data indicates whether the time series historical monitoring data corresponding to the comprehensive historical representative data is abnormal. The content of the label data is real and can be set manually. The diagnosis module records the combination of the comprehensive historical representative data and the label data as training data. Thereafter, the diagnosis module can use all the training data to train and generate an analysis model.

[0056] Furthermore, the diagnosis module uses all the training data to train and generate an analysis model, including the following steps:

[0057] The diagnostic module trains to generate candidate analysis models, and the diagnostic module trains to generate recognition models. The diagnostic module also determines whether the number of executions of this step has reached a preset execution number threshold. If it has been reached, the candidate analysis model generated by the last training will be used as the analysis model and the execution of this step will be terminated. If it has not been reached, this step will be repeated.

[0058] Specifically, the main steps of the diagnosis module using all the training data to train and generate an analysis model are introduced. First, the diagnosis module trains to generate a candidate analysis model. The following article will introduce how to train and generate a candidate analysis model. Secondly, the diagnosis module trains to generate a recognition model. The following article will introduce how to train and generate a recognition model. Finally, the diagnosis module determines whether the execution number of this step has reached a preset execution number threshold. The execution number threshold is set according to the actual application scenario. If the execution number threshold has been reached, the candidate analysis model generated by the last training is used as the analysis model, and this step is terminated. If the execution number threshold has not been reached, this step is repeated.

[0059] Furthermore, the diagnosis module trains and generates a candidate analysis model, including the following steps:

[0060] Step 1: The diagnosis module obtains all the training data, and determines whether it is the first time to generate a candidate analysis model. If it is the first time to generate a candidate analysis model, the diagnosis module uses all the training data to train and generate the candidate analysis model. If it is not the first time to generate a candidate analysis model, the diagnosis module proceeds to the next step.

[0061] Step 2: The diagnosis module obtains the recognition model generated by the most recent training, inputs all the training data into the obtained recognition model in sequence, obtains the recognition result data output by the obtained recognition model respectively, and the diagnosis module determines a number of inappropriate training data in all the training data based on all the recognition result data, and uses all the other training data except the several inappropriate training data in all the training data to train and generate a candidate analysis model.

[0062] Specifically, refer to Figure 3As shown, it is introduced how the diagnosis module trains and generates candidate analysis models in the process of training and generating analysis models. In step 1, the diagnosis module obtains all the training data and determines whether it is the first time to generate a candidate analysis model. If it is the first time to generate a candidate analysis model, because the recognition model has not been trained and generated, all the training data are used to train and generate the candidate analysis model. The candidate analysis model has the same function as the analysis model. The process of training and generating the candidate analysis model is a prior art, so it is not repeated. If it is not the first time to generate a candidate analysis model, continue to the next step, that is, step 2. In step 2, the diagnosis module obtains the recognition model generated by the most recent training, inputs all the training data into the obtained recognition model in sequence, and obtains the recognition result data output by the obtained recognition model respectively. The content of the recognition result data indicates whether the corresponding training data should be allowed to be used to train and generate the candidate analysis model. The diagnosis module determines a number of inappropriate training data from all the training data based on all the recognition result data. Inappropriate training data refers to training data that should not be allowed to be used to train and generate the candidate analysis model. Therefore, the diagnosis module uses all the other training data except a number of inappropriate training data from all the training data to train and generate the candidate analysis model.

[0063] Furthermore, the diagnosis module is trained to generate a recognition model, including the following steps:

[0064] Step 1: The diagnosis module obtains all the training data, obtains the candidate analysis model generated by the most recent training, and divides all the training data into different training data groups according to the different contents of the included labeled data;

[0065] Step 2: For each training data group, for each training data in the training data group, the diagnosis module inputs the training data into the obtained candidate analysis model to obtain the analysis result data output by the obtained candidate analysis model;

[0066] Step 3: For each training data group, for each training data in the training data group, the diagnosis module calculates the difference between the content of the label data contained in the training data and the content of the analysis result data corresponding to the training data to obtain deviation data, and in each training data group, a number of training data whose content of the corresponding deviation data is less than a preset threshold are divided into a first set, and a number of training data whose content of the corresponding deviation data is greater than or equal to the preset threshold are divided into a second set;

[0067] Step 4: The diagnostic module sets the content of each training data in the first set as allowed flag data, and uses the training data and the flag data to form practice data. For each training data in the second set, the diagnostic module sets the content of the training data as prohibited flag data, and uses the training data and the flag data to form practice data. The diagnostic module generates a recognition model based on all the practice data training.

[0068] Specifically, it is introduced how the diagnosis module trains and generates the recognition model in the process of training and generating the analysis model. In step 1, the diagnosis module obtains all the training data, obtains the candidate analysis model generated by the most recent training, and the diagnosis module divides all the training data into different training data groups according to the different contents of the included label data. For ease of understanding, for example, there are training data 1 to training data 4, the content of the label data contained in training data 1 is 0.7, the content of the label data contained in training data 2 is 0.4, the content of the label data contained in training data 3 is 0.8, and the content of the label data contained in training data 4 is 0.5, then training data 1 and training data 3 should be divided into the same training data group, because they correspond to the crane failure, and training data 2 and training data 4 should be divided into the same training data group, because they correspond to the crane not failing. In step 2, for each training data group, for each training data in the training data group, the diagnosis module inputs the training data into the obtained candidate analysis model, and obtains the analysis result data output by the obtained candidate analysis model. In step 3, for each training data group, for each training data in the training data group, the diagnosis module calculates the difference between the content of the labeled data contained in the training data and the content of the analysis result data corresponding to the training data, and obtains the deviation data. For example, if the content of the labeled data is 0.8 and the content of the analysis result data is 0.7, then the content of the deviation data is 0.1. Thereafter, in each training data group, the diagnosis module divides several training data whose corresponding deviation data content is less than a preset threshold into a first set. The training data in the first set are considered to be allowed to be used for training and generating candidate analysis models, and divides several training data whose corresponding deviation data content is greater than or equal to the preset threshold into a second set. The training data in the second set are considered not to be allowed to be used for training and generating candidate analysis models. The preset threshold is set according to the actual application scenario. In step 4, the diagnostic module sets the content of each training data in the first set as allowed flag data, and uses the training data and the flag data to form practice data. The diagnostic module sets the content of each training data in the second set as prohibited flag data, and uses the training data and the flag data to form practice data. After that, the diagnostic module can generate a recognition model based on all the practice data. The process of training and generating a recognition model is a prior art and will not be described in detail herein.

[0069] In summary, first, the monitoring module collects the time series monitoring data of the crane, transmits the collected time series monitoring data to the diagnosis module through the network module, and after receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training, and obtains the time series correction data output by the correction model. Secondly, the diagnosis module calculates the time series difference data between the time series correction data and the time series monitoring data, obtains the first representative data based on the time series difference data, and the diagnosis module obtains the second representative data based on the time series monitoring data. Finally, the diagnosis module uses the first representative data and the second representative data to form comprehensive representative data, and the diagnosis module inputs the comprehensive representative data into the analysis model generated by training, and obtains the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal. Through this application, not only can the intelligent level of crane fault diagnosis be significantly improved, but also the accuracy of the crane fault diagnosis results can be ensured, so that crane faults can be discovered in a timely manner and the losses caused by crane faults can be reduced.

[0070] According to another aspect of the embodiment of the present application, refer to Figure 4 As shown, the present application also provides a crane fault diagnosis system based on data mining, including a monitoring module, a network module, and a diagnosis module, so as to implement the crane fault diagnosis method based on data mining described above.

[0071] The functions of each module are as follows:

[0072] A monitoring module, used for storing a plurality of time series historical monitoring data of the crane, and for collecting the time series monitoring data of the crane, and transmitting the collected time series monitoring data to the diagnosis module through the network module;

[0073] A network module, used for transmitting data between the monitoring module and the diagnosis module;

[0074] The diagnostic module is used to input the time series monitoring data into the correction model generated by training after receiving the time series monitoring data, obtain the time series correction data output by the correction model, and calculate the time series difference data between the time series correction data and the time series monitoring data, obtain the first representative data based on the time series difference data, obtain the second representative data based on the time series monitoring data, and use the first representative data and the second representative data to form comprehensive representative data, input the comprehensive representative data into the analysis model generated by training, and obtain the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal.

[0075] According to another aspect of an embodiment of the present invention, there is further provided a device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement any one of the above methods when executing the computer program.

[0076] According to another aspect of an embodiment of the present invention, a medium is provided, wherein the medium stores program instructions, wherein when the program instructions are executed, a device where the medium is located is controlled to execute any one of the above methods.

[0077] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0078] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0079] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A crane fault diagnosis method based on data mining, characterized in that: The method comprises the following steps: The monitoring module collects the time series monitoring data of the crane, and transmits the collected time series monitoring data to the diagnosis module through the network module. After receiving the time series monitoring data, the diagnosis module inputs the time series monitoring data into the correction model generated by training, and obtains the time series correction data output by the correction model; The diagnosis module calculates time series difference data between the time series correction data and the time series monitoring data, obtains first representative data based on the time series difference data, and the diagnosis module obtains second representative data based on the time series monitoring data; The diagnosis module uses the first representative data and the second representative data to form comprehensive representative data, and the diagnosis module inputs the comprehensive representative data into the analysis model generated by training to obtain the diagnosis result data output by the analysis model regarding whether the time series monitoring data is abnormal; Before the diagnosis module is trained to generate the correction model and the analysis model, the following steps are included: The diagnostic module collects a number of time series historical monitoring data from the monitoring module; For each collected time series historical monitoring data, the diagnostic module sets identification data corresponding to the time series historical monitoring data, the content of the identification data is a normal situation or a fault situation, and the time series historical monitoring data and the identification data are stored correspondingly; The diagnostic module training generates a correction model and an analysis model, including the following steps: The diagnostic module selects several combinations including identification data of normal conditions from among all the stored combinations of time series historical monitoring data and identification data, extracts several time series historical monitoring data from the selected combinations, and uses the extracted several time series historical monitoring data to train and generate a correction model; For each stored combination of time series historical monitoring data and identification data, the diagnosis module inputs the time series historical monitoring data contained in the combination into the correction model generated by training to obtain the time series historical correction data output by the correction model; For each time series historical correction data, the diagnosis module calculates time series historical difference data between the time series historical correction data and the time series historical monitoring data corresponding to the time series historical correction data, obtains first historical representative data based on the time series historical difference data, obtains second historical representative data based on the time series historical monitoring data corresponding to the time series historical correction data, and forms comprehensive historical representative data using the first historical representative data and the second historical representative data; For each comprehensive historical representative data, the diagnostic module sets label data for the comprehensive historical representative data as to whether the time series historical monitoring data corresponding to the comprehensive historical representative data is abnormal, records the combination of the comprehensive historical representative data and the label data as training data, and the diagnostic module uses all the training data to train and generate an analysis model.

2. The method according to claim 1, characterized in that The diagnostic module uses all the training data to train and generate an analysis model, including the following steps: The diagnostic module is trained to generate candidate analysis models, and the diagnostic module is trained to generate recognition models. The diagnostic module also determines whether the number of executions of this step has reached a preset execution number threshold. If it has been reached, the candidate analysis model generated by the last training is used as the analysis model and the execution of this step is terminated. If it has not been reached, this step is repeated.

3. The method according to claim 2, characterized in that The diagnosis module is trained to generate a candidate analysis model, including the following steps: The diagnosis module obtains all the training data, and the diagnosis module determines whether it is the first time to generate the candidate analysis model. If it is the first time to generate the candidate analysis model, the diagnosis module uses all the training data to train and generate the candidate analysis model. If it is not the first time to generate the candidate analysis model, the diagnosis module proceeds to the next step. The diagnostic module obtains the recognition model generated by the most recent training, inputs all the training data into the obtained recognition model in sequence, obtains the recognition result data output by the obtained recognition model respectively, and the diagnostic module determines a number of inappropriate training data from all the training data based on all the recognition result data, and uses all the other training data from all the training data except the several inappropriate training data to train and generate a candidate analysis model.

4. The method according to claim 2, characterized in that: The diagnosis module is trained to generate a recognition model, including the following steps: The diagnosis module obtains all the training data, obtains the candidate analysis model generated by the most recent training, and divides all the training data into different training data groups according to the different contents of the included labeled data; With respect to each training data group, for each training data in the training data group, the diagnosis module inputs the training data into the acquired candidate analysis model, and obtains the analysis result data output by the acquired candidate analysis model; With respect to each training data group, for each training data in the training data group, the diagnosis module calculates the difference between the content of the label data contained in the training data and the content of the analysis result data corresponding to the training data to obtain deviation data, and in each training data group, a number of training data whose content of the corresponding deviation data is less than a preset threshold are divided into a first set, and a number of training data whose content of the corresponding deviation data is greater than or equal to the preset threshold are divided into a second set; The diagnostic module sets the content of each training data in the first set as allowed flag data, and uses the training data and the flag data to form practice data. For each training data in the second set, the diagnostic module sets the content of the training data as prohibited flag data, and uses the training data and the flag data to form practice data. The diagnostic module generates a recognition model based on all the practice data training.

5. A crane fault diagnosis system based on data mining, used to implement the method according to any one of claims 1 to 4, characterized in that: Includes the following modules: A monitoring module, used for storing a plurality of time series historical monitoring data of the crane, and for collecting the time series monitoring data of the crane, and transmitting the collected time series monitoring data to the diagnosis module through the network module; A network module, used for transmitting data between the monitoring module and the diagnosis module; The diagnostic module is used to input the time series monitoring data into the correction model generated by training after receiving the time series monitoring data, obtain the time series correction data output by the correction model, and calculate the time series difference data between the time series correction data and the time series monitoring data, obtain the first representative data based on the time series difference data, obtain the second representative data based on the time series monitoring data, and use the first representative data and the second representative data to form comprehensive representative data, input the comprehensive representative data into the analysis model generated by training, and obtain the diagnostic result data output by the analysis model on whether the time series monitoring data is abnormal.

6. A device, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the method according to any one of claims 1 to 4 when executing the computer program.

7. A medium, characterized in that The medium stores program instructions, wherein when the program instructions are executed, the device where the medium is located is controlled to execute the method according to any one of claims 1 to 4.

Images