Fault diagnosis method, system and equipment of numerical control machine tool spindle and medium

Abstract

The embodiment of the invention relates to the technical field of numerical control machine tools, and discloses a fault diagnosis method, system and device for a spindle of a numerical control machine tool and a medium. The fault diagnosis method for the numerical control machine tool spindle comprises the steps of obtaining a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a numerical control machine tool; extracting a frequency domain signal in the real-time vibration signal; feature extraction is carried out on the frequency domain signal through a deep belief network, high-dimensional features of the frequency domain signal are obtained, and the high-dimensional features are used for indicating a target fault mode of the main shaft bearing; and selecting a preset fault mode matched with the target fault mode from a pre-constructed main shaft fault knowledge base containing a plurality of preset fault modes, and taking a fault solution strategy corresponding to the selected preset fault mode in the main shaft fault knowledge base as a fault solution strategy corresponding to the target fault mode. According to the fault diagnosis method for the numerical control machine tool spindle, the accuracy of fault diagnosis of the numerical control machine tool spindle can be improved, and a fault solving strategy is provided in time.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of CNC machine tools, and in particular to a fault diagnosis method, system, device and medium for a spindle of a CNC machine tool. Background Art

[0002] As key equipment for modern intelligent manufacturing, CNC machine tools are the cornerstone for high-precision processing and manufacturing, and play the role of "the mother of industry" in the development of industrial automation. The spindle system of CNC machine tools often fails due to long-term operation, which affects the performance of the machine tools, reduces product quality, and even causes the machine tools to fail to work properly. The spindle, as the "joint" of the mechanical transmission of CNC machine tools, is the core supporting component for the smooth operation of the spindle of CNC machine tools. Therefore, status monitoring and fault diagnosis of the machine tool spindle system are the key to ensuring the reliable operation of the entire CNC machine tool.

[0003] Currently, most spindle systems are monitored and diagnosed manually through regular maintenance and upkeep, which is not only time-consuming and labor-intensive, but also the diagnostic results are subject to the uncertainty of manual experience. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a fault diagnosis method, system, device and medium for a CNC machine tool spindle, which can improve the accuracy of CNC machine tool spindle fault diagnosis and provide fault resolution strategies in a timely manner to ensure the reliable operation of the CNC machine tool.

[0005] In order to solve the above technical problems, an embodiment of the present application provides a fault diagnosis method for a spindle of a CNC machine tool, comprising the following steps:

[0006] Obtain the frequency domain signal corresponding to the real-time vibration signal of the spindle bearing of the CNC machine tool;

[0007] Extracting features of the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, wherein the high-dimensional features are used to indicate a target fault mode of the spindle bearing;

[0008] The deep belief network includes a plurality of stacked restricted Boltzmann machines and a BP classifier, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the plurality of stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network, and the features extracted by each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation;

[0009] A preset fault mode matching the target fault mode is selected from a pre-built spindle fault knowledge base containing multiple preset fault modes, and a fault resolution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base is used as the fault resolution strategy corresponding to the target fault mode.

[0010] In some optional embodiments, the step of acquiring a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool includes:

[0011] Based on the empirical mode decomposition method, the real-time vibration signal is decomposed to obtain a plurality of intrinsic mode function components of the real-time vibration signal;

[0012] Envelope detection is performed on each eigenmode function component through Hilbert transform to obtain frequency domain signals of different time scales of the real-time vibration signal.

[0013] In some optional embodiments, selecting a preset fault mode matching the target fault mode from a pre-built spindle fault knowledge base containing multiple preset fault modes includes:

[0014] Based on a concept similarity algorithm, a semantic mapping relationship between the target fault mode and the plurality of preset fault modes is established;

[0015] According to the semantic mapping relationship between the target fault mode and the multiple preset fault modes, a preset fault mode having the highest similarity to the target fault mode is selected from the multiple preset fault modes.

[0016] In some optional embodiments, the spindle fault knowledge base is constructed by the following steps:

[0017] Constructing an ontology class of the spindle fault knowledge base, wherein the ontology class includes preset fault modes and fault resolution strategies of the spindle bearing;

[0018] Defining object properties of all objects in the ontology class to associate objects that have corresponding relationships;

[0019] Define the data attributes of all objects in the ontology class.

[0020] In some optional embodiments, the entity class further includes the fault type, fault location, fault phenomenon and fault cause of the main shaft bearing, and the method further includes:

[0021] According to the determined target fault mode and the fault resolution strategy corresponding to the target fault mode, the fault type, fault location, fault phenomenon and fault cause corresponding to the target fault mode are obtained.

[0022] In some optional embodiments, a preset failure mode of the main shaft bearing corresponds to multiple failure causes, and obtaining the failure cause corresponding to the target failure mode according to the target failure mode and the failure solution strategy corresponding to the target failure mode includes:

[0023] Based on the preset SWRL rules, knowledge reasoning is performed on the target fault mode and the fault resolution strategy corresponding to the target fault mode to obtain the fault cause corresponding to the target fault mode.

[0024] In some optional embodiments, the target failure mode is obtained by classifying the high-dimensional features based on a random forest algorithm.

[0025] The embodiment of the present application also provides a fault diagnosis system for a CNC machine tool spindle, comprising:

[0026] A signal acquisition module is used to obtain a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool;

[0027] A feature extraction module, used for performing feature extraction on the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, wherein the high-dimensional features are used to indicate a target fault mode of the spindle bearing;

[0028] The deep belief network includes a plurality of stacked restricted Boltzmann machines and a BP classifier, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the plurality of stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network, and the features extracted by each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation;

[0029] A fault diagnosis module is used to select a preset fault mode that matches the target fault mode from a pre-built spindle fault knowledge base containing multiple preset fault modes, and use the fault solution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base as the fault solution strategy corresponding to the target fault mode.

[0030] An embodiment of the present application also provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned CNC machine tool spindle fault diagnosis method.

[0031] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, wherein the computer program implements the above-mentioned CNC machine tool spindle fault diagnosis method when executed by a processor.

[0032] The fault diagnosis method of a CNC machine tool spindle provided in the embodiment of the present application has at least the following beneficial effects:

[0033] Since the most prone to failure and the most difficult to detect failure in the spindle system of a CNC machine tool often comes from the bearing part, the real-time vibration signal of the spindle bearing of the CNC machine tool is collected in this application, and the frequency domain signal that accurately reflects the spindle vibration failure is extracted. The frequency domain signal is subjected to feature extraction through a deep confidence network, which can fully exploit the high-dimensional features of the spindle bearing failure, and the high-dimensional features are used to indicate the target failure mode of the spindle bearing. A preset failure mode that matches the target failure mode can be selected from a pre-built spindle failure knowledge base containing multiple preset failure modes, and the fault resolution strategy corresponding to the selected preset failure mode in the spindle failure knowledge base is used as the fault resolution strategy corresponding to the target failure mode.

[0034] The deep belief network includes multiple stacked restricted Boltzmann machines, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the multiple stacked restricted Boltzmann machines is the feature transmission direction of the deep belief network. The extracted features of each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so that the high-dimensional features of the spindle bearing fault can be fully mined. The BP classifier can correct the network parameters of the deep belief network through reverse error propagation, further improving the accuracy of the extracted high-dimensional features, with high efficiency and accuracy. At the same time, the fault resolution strategy corresponding to the fault mode can also be determined, so as to timely handle the spindle bearing fault of the CNC machine tool and ensure the stable operation of the CNC machine tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] One or more embodiments are exemplarily described by the pictures in the corresponding drawings, and these exemplary descriptions do not constitute limitations on the embodiments.

[0036] Figure 1 It is a flow chart of a fault diagnosis method for a CNC machine tool spindle provided according to an embodiment of the present application;

[0037] Figure 2 is a schematic diagram of a DBN model structure provided according to an embodiment of the present application;

[0038] Figure 3 It is a flow chart of constructing a spindle fault knowledge base according to an embodiment of the present application;

[0039] Figure 4 is a schematic diagram of an ontology class of a spindle fault knowledge base provided according to an embodiment of the present application;

[0040] Figure 5 is a characteristic schematic diagram of a statistical parameter provided according to an embodiment of the present application;

[0041] Figure 6 is a schematic diagram of a model output result provided according to an embodiment of the present application;

[0042] Figure 7 is a schematic diagram of a fault diagnosis system for a CNC machine tool spindle provided according to an embodiment of the present application;

[0043] Figure 8 It is a structural diagram of a computer device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0045] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. However, it will be appreciated by those skilled in the art that in the present application, many technical details are proposed in order to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The division of the following embodiments is for the convenience of description, and the specific implementation of the present application should not be construed as any limitation, and the various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0046] In order to avoid downtime caused by spindle failure of CNC machine tools and the impact on product quality, it is crucial to monitor the state of CNC machine tool spindle systems and identify faults in the early stage. With the development of intelligent sensing technology and artificial intelligence theory, it has become a research hotspot to focus on monitoring and analyzing the state data of machine tool spindles to achieve accurate and intelligent fault diagnosis. For example, based on data-driven quantitative analysis methods, sensors collect relevant information and data representing the fault state of machine tool equipment, analyze and process the data through mechanism analysis, signal processing methods and mathematical statistics methods, extract key fault features and identify fault types. Qualitative analysis methods based on knowledge modeling can achieve efficient semantic mapping and knowledge reasoning while realizing domain knowledge sharing. On the basis of analyzing the fault type of machine tool spindle based on data-driven analysis, potential implicit knowledge can be inferred, thereby giving the cause of the fault, fault repair methods and maintenance measures, thereby reducing the probability of fault occurrence.

[0047] Based on the above, by associating the data-driven fault type identification results with the instances in the constructed CNC machine tool spindle fault knowledge base, we can further explore the deep-seated causes of the fault and provide specific fault maintenance strategies through knowledge reasoning. Therefore, the embodiment of the present application combines the quantitative analysis method based on data-driven with the qualitative analysis method based on knowledge modeling to achieve in-depth and comprehensive intelligent fault diagnosis and health management of CNC machine tool spindle systems.

[0048] An embodiment of the present application relates to a fault diagnosis method for a spindle of a CNC machine tool. The implementation details of the fault diagnosis method for a spindle of a CNC machine tool of this embodiment are specifically described below. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this solution.

[0049] The specific process of the fault diagnosis method of the CNC machine tool spindle of this embodiment can be as follows: Figure 1 As shown, including:

[0050] Step 101, obtaining a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool.

[0051] Specifically, as a basic component of CNC machine tools, the spindle plays an important role in the movement of parts processing, and its working state directly affects the efficiency and quality of product processing. The spindle system is generally composed of a rotating shaft, bearings, a broach system, a rotor, a stator, a spindle housing and other components. The mechanical failure of the spindle system is often caused by the wear and aging of the above components, improper use, and changes in the equipment structure. The most prone to failure and the most difficult to find failure in the spindle system often comes from the rolling bearing part. The frequency of bearing failure is high and it is also difficult to be diagnosed by the machine tool self-diagnosis system. The long-term operation of the spindle causes the high-speed operation of the bearing to wear, which causes the bearing clearance to increase and deform. In addition, if the maintenance of the bearing components is not paid attention to, pitting, cracks and other defects will appear on the surface (inner ring, outer ring, rolling element) due to poor lubrication and excessive load. Therefore, the quality of the bearing directly determines the working life and normal operation of the machine tool spindle system, and also affects the overall operating performance of the CNC machine tool.

[0052] In the specific implementation, when the machine tool spindle bearing fails, it is often accompanied by a highly random vibration signal, and this type of nonlinear vibration signal contains rich fault information and characteristics. Therefore, in order to fully perceive and grasp the type of bearing failure, this embodiment chooses to install a vibration sensor module on the machine tool spindle part to obtain a real-time vibration signal that can fully reflect the characteristics of the bearing failure.

[0053] By performing signal analysis and processing on the collected real-time vibration signals, sensitive characteristic information (i.e., frequency domain signals) in the real-time vibration signals can be further obtained, laying a data foundation for subsequent fault type identification research.

[0054] In one example, based on the empirical mode decomposition method, the real-time vibration signal is decomposed to obtain multiple intrinsic mode function components of the real-time vibration signal; each intrinsic mode function component is envelope-detected by Hilbert transform to obtain frequency domain signals of different time scales of the real-time vibration signal. Since the vibration signal collected by the sensor of the CNC machine tool spindle system contains more noise and exhibits nonlinear properties of frequent oscillation and instability when a failure occurs, conventional signal processing methods cannot adapt to this complexity. Therefore, this embodiment realizes the decomposition of nonlinear multidimensional data (real-time vibration signal of spindle bearing) and the extraction of frequency domain signals through empirical mode decomposition (EMD).

[0055] EMD is an analysis method that can adaptively process non-stationary and nonlinear signals. EMD does not require the specification of basis functions, and can flexibly decompose signals into frequency band components of different scales and extract its intrinsic mode function from the time domain signal itself. The specific decomposition process is as follows:

[0056]

[0057] Among them, x(t) represents the real-time vibration signal, t represents the sampling time of the real-time vibration signal, I represents the kth IMF component containing different time scales, and r(t) represents the residual trend term of the signal.

[0058] In the processing of CNC machine tool spindle vibration signals, EMD is used to obtain the intrinsic mode function components IMF, and then envelope detection is performed through Hilbert transform to obtain frequency domains of different time scales. These information components can more accurately characterize the spindle fault characteristics. The specific implementation process of EMD mainly includes the following steps: (1) decomposing the vibration signal into local extreme value lines of different scales; (2) constructing envelope detection lines through local extreme value lines; (3) averaging the envelope lines to obtain IMF components; (4) repeating the iteration until only the trend term is retained in the signal, thereby extracting the vibration intrinsic mode components of the original signal and reconstructing them to obtain the vibration signal frequency domain signal.

[0059] Step 102: extract features of the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, where the high-dimensional features are used to indicate a target fault mode of the spindle bearing.

[0060] Specifically, Deep belief networks (DBN) are deep learning networks that can fully mine data features. They are multi-layer neural network generation models formed by stacking multiple restricted Boltzmann machines (RBM). DBN can start from the underlying raw data, optimize the weight parameters of the deep neural network through layer-by-layer greedy learning algorithm training, realize the complex mapping from low-dimensional input to high-dimensional feature data output, and effectively mine the essential characteristics of faults in the raw data of vibration signals.

[0061] In the specific implementation, the DBN learning process mainly consists of unsupervised forward stacking RBM learning and supervised backward fine-tuning RBM parameter training. The model structure and training process of DBN are as follows: Figure 2 As shown, the deep belief network includes multiple stacked restricted Boltzmann machines and a BP classifier, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the multiple stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network, and the extracted features of each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation.

[0062] Among them, h represents the hidden layer, v represents the visual layer, and the two are connected by the weight parameter w. The input raw data passes through the first visual layer input and forms a low-level RBM with the first hidden layer. The output of this layer is used as the visual layer input of the second RBM, and forms the second RBM with the second hidden layer h2. And so on, the stacking forms a DBN deep learning network. The DBN network parameters are initialized through forward layer-by-layer training, and then the BP (Back Propagation) classifier is used to fine-tune the parameters through reverse error propagation correction, so as to obtain more representative high-dimensional features from the low-level data through continuous model training.

[0063] In one example, the target fault mode is obtained by classifying high-dimensional features based on the random forest algorithm. After extracting the high-dimensional features of the frequency domain signal through DBN, the obtained high-dimensional feature set is randomly sampled and grouped to construct a training sample set, which is input into the random forest algorithm model, and the recognition result of the fault mode (i.e., the target fault mode) is output through majority voting.

[0064] The Random Forest algorithm (RF) is a supervised machine learning algorithm used to solve classification or regression problems. The algorithm uses decision trees as the basic learner and uses decision tree voting to establish the results. The main principle is to randomly sample k sample sets with replacement from the N sample training sets, and randomly select features from these sample sets to generate decision trees to form a random forest. Each decision tree is used to perform pattern discrimination on the new sample to be tested, and the discrimination results of all decision trees are combined to determine the final classification result through selection voting. The implementation process of the random forest algorithm model mainly includes the following steps: (1) m samples are randomly selected with replacement from the original N training sample data sets using the bootstrap method as the sampling subset, and the sampling is repeated multiple times. (2) Decision trees are established in the sampling subset for training, and attribute features are randomly selected from the feature set to calculate the optimal features as tree nodes. (3) Each decision tree grows by node splitting until it can no longer be divided, thereby generating n classification and regression decision trees. (4) The test set is input into the random forest model composed of n decision trees to obtain n results, and the final result is determined by majority voting.

[0065] Step 103 , selecting a preset fault mode that matches the target fault mode from a pre-built spindle fault knowledge base that includes multiple preset fault modes, and using the fault resolution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base as the fault resolution strategy corresponding to the target fault mode.

[0066] Specifically, this embodiment uses the ontology knowledge modeling method to construct a spindle fault knowledge base. The spindle fault knowledge base contains a variety of preset fault modes of spindle bearings and corresponding fault resolution strategies. By matching the target fault mode with these multiple preset fault modes, a preset fault mode that matches the target fault mode can be obtained, and then the fault resolution strategy corresponding to the target fault mode can be determined.

[0067] In one example, a preset fault mode that matches a target fault mode is obtained in the following manner: based on a concept similarity algorithm, a semantic mapping relationship between the target fault mode and multiple preset fault modes is established; based on the semantic mapping relationship between the target fault mode and multiple preset fault modes, a preset fault mode with the highest similarity to the target fault mode is selected from the multiple preset fault modes.

[0068] The construction process of the spindle fault knowledge base is as follows: Figure 3As shown in the figure, the knowledge base is applied to the fault diagnosis of CNC machine tool spindles, and the ontology is constructed mainly around the knowledge of spindle component structure and fault mechanism. First, the knowledge in the field of CNC machine tool spindle bearings is obtained through literature, expert experience and fault cases, and an information dictionary is established based on the ontology formal language to lay the corpus foundation for ontology construction. Then, the protege ontology construction tool is used to build the spindle fault diagnosis ontology model, that is, to define the ontology concept class and its subclasses, define the object attributes and data attributes of the ontology class, define the attribute constraint range and the relationship between the class, create the class instance, build the diagnostic reasoning rules and combine the instance to complete the ontology consistency check and evaluation, and feedback to improve the ontology if the consistency check fails. Among them, the ontology is stored and shared in OWL files, and the knowledge base is formed through instance semantic mapping, and fault diagnosis is realized based on knowledge reasoning.

[0069] Therefore, the construction of the spindle fault knowledge base mainly includes the following parts:

[0070] (1) Construct the ontology class of the spindle fault knowledge base, which includes the preset failure modes and fault resolution strategies of the spindle bearings.

[0071] Ontology construction includes five basic primitives, namely, class, class attribute, axiom, relationship and instance. By analyzing the structural characteristics and fault mechanism of machine tool spindle bearings, the bearing fault diagnosis domain ontology (Fault Diagnosis Ontology, FDO) is defined. FDO =<C,A,R,I,X> , where C represents the set of concept classes, A represents the set of attributes, R represents the set of relationships between classes, I represents the set of class instances, and X represents the set of axioms.

[0072] The structural characteristics and fault mechanism of the bearing are analyzed and summarized into top-level parent classes: fault location, fault characteristics, fault phenomenon, fault cause, fault type, and fault diagnosis and maintenance strategy. The fault location refers to the specific location where the fault phenomenon and fault repair occur in the bearing; the fault phenomenon refers to the description of the external manifestation when the bearing fails; the fault cause refers to the cause of the specific fault of the machine tool spindle bearing; the fault type refers to the classification of the fault; the fault diagnosis and maintenance strategy refers to the specific and effective measures taken to solve the fault (equivalent to the fault solution strategy); the fault characteristic refers to the main characteristics of each component that characterizes the bearing fault extracted after the analysis of the fault signal collected by the sensor (equivalent to the fault mode). Based on the above core elements, the top-level parent class of the spindle bearing fault domain ontology is determined, and the ontology construction tool protege is used to manually construct the parent class hierarchy. The parent class can contain subclasses, and its subclasses are further determined based on the above five top-level classes. Figure 4 A schematic diagram showing an ontology class and the relationship between ontology classes.

[0073] In an example, the ontology class also includes the fault type, fault location, fault phenomenon and fault cause of the spindle bearing. After determining the target fault mode and the fault solution strategy corresponding to the target fault mode, the fault type, fault location, fault phenomenon and fault cause corresponding to the target fault mode can be obtained based on the determined target fault mode and the fault solution strategy of the target fault mode.

[0074] In one example, a preset failure mode of a spindle bearing corresponds to multiple failure causes. In this case, in order to determine the failure cause corresponding to the target failure mode, it is necessary to perform knowledge reasoning on the target failure mode and the fault resolution strategy corresponding to the target failure mode based on the preset SWRL rules to obtain the failure cause corresponding to the target failure mode.

[0075] (2) Define the object properties of all objects in the ontology class to associate the ontology classes that have corresponding relationships and define the data properties of all objects in the ontology class.

[0076] Specifically, the semantic relationship and instance properties between ontology classes need to be described by defining class attributes. Attributes are mainly constrained by axioms, domains, and ranges. In the bearing fault diagnosis ontology, object attributes and data attributes are defined to express the association between classes and the data characteristics of class instances. For example, the object attribute "has_Cause" is used to associate the fault phenomenon with the fault cause. For example, the attribute domain of "pitting corrosion has_Cause poor lubrication in the main bearing" is the fault phenomenon class, and the range is the fault cause class. In addition, data attributes are added in protege through the DataTypeProperties option to describe the data constraints of the instance. For example, "has_temperature" indicates that the bearing seat has temperature data characteristics, and its domain is "fault location" and the range is "float".

[0077] In order to realize knowledge reasoning, it is also necessary to instantiate the constructed ontology class and complete the transformation into a knowledge base. In addition to having all the attributes of the class, the specific instance of the class can also realize knowledge reasoning by semantically associating with other instances through object attributes. There may be semantic conflicts and errors in the knowledge base after the spindle fault ontology is instantiated. It is necessary to use the built-in reasoning engine in the protege tool to perform consistency detection to ensure the logical consistency of the knowledge base structure and the correctness of the knowledge description. Assuming that the diameter damage of the inner ring of the bearing generally exceeds 0.17mm, it is considered that pitting fault has occurred. Once pitting occurs, it is identified as an inner ring fault.

[0078] The following is a consistency test of the knowledge base for the bearing fault instance: the "testInstance1" test instance, the inner ring component instance Inner_Ring1 and the pitting fault phenomenon instance "Pitting_phenomenon1" are constructed. The inner ring component of "testInstance1" is inner_Ring1, that is, "testInstance1 has_Componentinner_Ring1", and the pitting phenomenon occurs, that is, "has_Cause Pitting_phenomenon1". The equivalence class of the inner ring fault class (Inner_RingFault) is added through the Euwntfo button as "(has_Component someInner_Ring)and(has_Cause some Pitting)". After selecting the HemiT reasoning engine and clicking Start Reasoner to start reasoning on the instance, it is found that the test instance is automatically attributed to an instance of the Inner_RingFault class, which is the same as the expected result. At the same time, the fault resolution measure that can be taken is Maintenance1, which proves that the semantic logic of the knowledge base is correct and the consistency test passes. After the knowledge base is verified, it can be saved in the form of an OWL file for cross-platform sharing.

[0079] In this embodiment, the semantic mapping method is used to semantically associate the results of signal processing, feature extraction and fault identification with the fault phenomenon instances in the ontology knowledge base to establish a mapping between them. The semantic mapping process can be expressed as: f = map(,R,[instance,},(result,}). Among them, f represents the mapping relationship between I and R, instance represents the set of fault phenomenon instances, and rusult represents the set of fault identification results in the data-driven diagnosis method. The concept similarity algorithm is used to establish a semantic mapping relationship between the two, and the ontology knowledge reasoning is combined with the constructed SWRL rules to obtain the fault cause and maintenance solution (fault resolution strategy) that matches the fault diagnosis instance (preset fault mode).

[0080] In this embodiment, since the most prone to failure and the most difficult to detect failure in the spindle system of a CNC machine tool often comes from the bearing part, the real-time vibration signal of the spindle bearing of the CNC machine tool is collected in this application, and the frequency domain signal that accurately reflects the spindle vibration failure is extracted. The frequency domain signal is subjected to feature extraction through a deep confidence network, which can fully exploit the high-dimensional features of the spindle bearing failure, and the high-dimensional features are used to indicate the target failure mode of the spindle bearing. A preset failure mode that matches the target failure mode can be selected from a pre-constructed spindle failure knowledge base containing multiple preset failure modes, and the fault resolution strategy corresponding to the selected preset failure mode in the spindle failure knowledge base is used as the fault resolution strategy corresponding to the target failure mode. The deep belief network includes multiple stacked restricted Boltzmann machines, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the multiple stacked restricted Boltzmann machines is the feature transmission direction of the deep belief network. The extracted features of each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so that the high-dimensional features of the spindle bearing fault can be fully mined. The BP classifier can correct the network parameters of the deep belief network through reverse error propagation, further improving the accuracy of the extracted high-dimensional features, with high efficiency and accuracy. At the same time, the fault resolution strategy corresponding to the fault mode can also be determined, so as to timely handle the spindle bearing fault of the CNC machine tool and ensure the stable operation of the CNC machine tool.

[0081] In one embodiment, the present application provides an experimental scheme for verifying the effectiveness and adaptability of the above-mentioned CNC machine tool spindle fault diagnosis method, as described below:

[0082] This embodiment uses an acceleration sensor to collect experimental sample data based on the SQI-MFS mechanical comprehensive fault simulation experimental platform. The experimental sample is a bearing, model MB ER-16K, which can simulate the working state of the bearing under healthy or different degrees of fault under different working conditions. In order to reflect the fault phenomena occurring in the actual operation of the bearing as comprehensively as possible, experimental data of three types of faults (inner ring, outer ring, rolling element) and fault diameters of 0.1778mm, 0.335mm, and 0.554mm were collected to form a sample data set corresponding to 10 working conditions. On this basis, bearing vibration signal collection tasks at speeds of 1000r / min, 1400r / min, and 1800r / min were set. Each task collects 100 groups of samples at a sampling frequency of 15kHz, which are divided into a model training set and a test set according to a ratio of 7:3, where each group of samples has 2000 data points. The experimental data sets of the three tasks are shown in Table 1:

[0083] Table 1

[0084]

[0085]

[0086] (1) After obtaining the original vibration signal data set according to the above requirements, the EMD is used to decompose the training sample signals in Task 1 to Task 3. The time domain signals of the first four IMF components, the Hilbert envelope spectrum of the four IMF components, and the Hilbert marginal spectrum are used to calculate 18 statistical parameters (range, mean, standard deviation, kurtosis, energy, energy entropy, kurtosis, crest factor, pulse factor, shape factor, skewness, power spectrum entropy, singular spectrum entropy, approximate entropy, sample entropy, fuzzy entropy, permutation entropy and envelope entropy). A total of 162 statistical features are obtained. Taking the inner race fault sample data at 1000 r / min as an example, the 162 statistical features extracted are as follows: Figure 5 shown.

[0087] (2) Input each group of original statistical feature data sets into the DBN network for iterative training to achieve adaptive mining of deep features. The hidden layer of the network is set to 5 layers, the number of nodes in the first layer is 10, the second layer is 8, the third layer is 5, and the RBM learning rate is set to 0.001. After setting the parameters, the model finally extracts 100 deep features that can better reflect bearing faults through iterative optimization. The training output results are as follows: Figure 6 Shown ， It can be seen from the figure that the feature discrimination and distribution characteristics are good.

[0088] (3) The deep features mined by the DBN network are input into RF for fault mode classification. First, the number of random forest decision trees is set to 80, the maximum depth of the decision tree is 12, and the number of optimization iterations of RF hyperparameter cross-validation is 5. Then, 40% of each working condition and its fault state are randomly selected as training data for RF training. After the training is completed, the actual fault label classification is performed on the test samples.

[0089] In addition to the RF classification method, classic machine learning classifiers (SVM, KNN, Softmax) are selected to identify and classify the fault states of Task 1, Task 2, and Task 3, respectively, and the diagnosis results of each model under different working conditions are obtained, as shown in Table 2. The experimental results show that the fault diagnosis models constructed based on the four classifiers can achieve effective fault pattern recognition and classification. Among them, the model constructed based on the RF classifier can achieve fault diagnosis performance that is significantly better than the models constructed by other classifiers. The fault diagnosis accuracy rates under Tasks 1-3 can reach 92.75%, 91.25%, and 94.79%, respectively.

[0090] Table 2

[0091]

[0092] To further verify the effectiveness of the fault diagnosis method for the CNC machine tool spindle of this embodiment, this embodiment conducts a bearing fault diagnosis experiment based on an actual CNC machine tool. The original data is collected by an acceleration sensor and a high-speed data acquisition card for verification, and four bearing states (normal, inner ring fault, outer ring fault, and rolling element fault) are set. The fault diagnosis method proposed in this embodiment is used to identify the fault state, and an average fault diagnosis accuracy rate of 92% can be achieved.

[0093] In addition, the above fault diagnosis results are semantically mapped with the instances in the spindle fault knowledge base of this embodiment, and the concept similarity is used to match the instances. Taking the "outer ring fault result (Outerringmalfunction)" of data-driven fault diagnosis as an example, the concept similarity Sim(cg2c1)=0.87 with the instance "quter_ringdefect" constructed in the knowledge base, so it is judged that there is a semantic mapping relationship between the two. Next, the relevant attributes of the fault instance are edited and input into the ontology knowledge base for knowledge reasoning. Here, a custom SWRL rule is constructed, that is, the "Pitting" class in the fault phenomenon and the "Maintenance2" class in the fault resolution strategy class are defined to include the same fault cause "Fault_Cause", then the two can be semantically associated through "hasMaintenance". After constructing the class instance and editing the rules, the protege reasoning engine is started for knowledge reasoning, which can realize the fault diagnosis, maintenance and strategy guidance of the fault phenomenon. After reasoning, the pitting corrosion phenomenon of the bearing outer ring "hasMaintenance (has maintenance strategy)" maintenance strategy 2. The specific content of maintenance strategy 2 is to improve the bearing processing quality, control the load, ensure lubrication and heat dissipation. In summary, it realizes the maintenance strategy support for specific faults.

[0094] Based on the above, the embodiment of the present application also uses C# language, WPF form application framework, and SqlSever database technology to develop a CNC machine tool spindle fault diagnosis and health service management service system. The system mainly includes three modules: First, the machine tool bearing status monitoring and diagnosis is based on the data collected by various sensors, referred to as the data-driven monitoring and fault identification module, which mainly completes the functions of data preservation, abnormal data alarm, dynamic display data, fault data processing and analysis, fault diagnosis, etc. The second is based on the CNC machine tool spindle fault ontology and rule reasoning knowledge base, which mainly includes the input of examples, fault knowledge query and retrieval, etc. The third is to combine the data-driven machine tool spindle bearing fault state identification results with ontology semantics to improve the knowledge service library of healthy operation management, mainly including linkage with the protege software background, iterative improvement of the ontology knowledge base built before, and can also provide knowledge retrieval services according to the fault phenomenon, and provide decision references for possible fault cause analysis and fault maintenance strategies, so as to provide guarantee for the healthy operation of CNC machine tool spindle equipment.

[0095] The step division of the above methods is only for the purpose of clear description. When implemented, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this patent; adding insignificant modifications to the algorithm or process or introducing insignificant designs without changing the core design of the algorithm and process are all within the scope of protection of this patent.

[0096] Another embodiment of the present application relates to a fault diagnosis system for a CNC machine tool spindle. The implementation details of the fault diagnosis system for the CNC machine tool spindle of this embodiment are specifically described below. The following content is only for the convenience of understanding the implementation details provided, and is not necessary for the implementation of this solution. The schematic diagram of the fault diagnosis system for the CNC machine tool spindle of this embodiment can be as follows: Figure 7 As shown, it includes: a signal acquisition module 701, a feature extraction module 702 and a fault diagnosis module 703.

[0097] Specifically, the signal acquisition module 701 is used to obtain a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool.

[0098] A feature extraction module 702 is used to extract features from the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, where the high-dimensional features are used to indicate a target fault mode of the spindle bearing;

[0099] Among them, the deep belief network includes multiple stacked restricted Boltzmann machines and a BP classifier. Each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the multiple stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network. The extracted features of each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation.

[0100] The fault diagnosis module 703 is used to select a preset fault mode that matches the target fault mode from a pre-built spindle fault knowledge base containing multiple preset fault modes, and use the fault solution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base as the fault solution strategy corresponding to the target fault mode.

[0101] In one example, the signal acquisition module 701 is also used to decompose the real-time vibration signal based on the empirical mode decomposition method to obtain multiple intrinsic mode function components of the real-time vibration signal; and perform envelope detection on each intrinsic mode function component through Hilbert transform to obtain frequency domain signals of different time scales of the real-time vibration signal.

[0102] In one example, the fault diagnosis module 703 is also used to establish a semantic mapping relationship between the target fault mode and multiple preset fault modes based on a concept similarity algorithm; according to the semantic mapping relationship between the target fault mode and multiple preset fault modes, select the preset fault mode with the highest similarity to the target fault mode from the multiple preset fault modes.

[0103] It is not difficult to find that this embodiment is a system embodiment corresponding to the above method embodiment, and this embodiment can be implemented in conjunction with the above method embodiment. The relevant technical details and technical effects mentioned in the above embodiment are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiment.

[0104] It is worth mentioning that all modules involved in this embodiment are logic modules. In practical applications, a logic unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application, but this does not mean that there are no other units in this embodiment.

[0105] Another embodiment of the present application relates to a computer device, such as Figure 8As shown, it includes: at least one processor 801; and a memory 802 that is communicatively connected to the at least one processor 801; wherein the memory 802 stores instructions that can be executed by the at least one processor 801, and the instructions are executed by the at least one processor 801 so that the at least one processor 801 can execute the fault diagnosis method of the CNC machine tool spindle in the above-mentioned embodiments.

[0106] Among them, the memory and the processor are connected in a bus manner, and the bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium via an antenna, and further, the antenna also receives data and transmits the data to the processor.

[0107] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0108] Another embodiment of the present application relates to a computer-readable storage medium storing a computer program, which implements the above method embodiment when executed by a processor.

[0109] That is, those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including a number of instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor 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 (Read-Only Memory, referred to as: ROM), random access memory (Random Access Memory, referred to as: RAM), disk or optical disk and other media that can store program codes.

[0110] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A method for diagnosing a spindle fault of a CNC machine tool, characterized in that: The method comprises: Obtain the frequency domain signal corresponding to the real-time vibration signal of the spindle bearing of the CNC machine tool; Extracting features of the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, wherein the high-dimensional features are used to indicate a target fault mode of the spindle bearing; The deep belief network includes a plurality of stacked restricted Boltzmann machines and a BP classifier, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the plurality of stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network, and the features extracted by each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation; A preset fault mode matching the target fault mode is selected from a pre-built spindle fault knowledge base containing multiple preset fault modes, and a fault resolution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base is used as the fault resolution strategy corresponding to the target fault mode.

2. The fault diagnosis method for a CNC machine tool spindle according to claim 1, characterized in that: The method of obtaining a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool comprises: Based on the empirical mode decomposition method, the real-time vibration signal is decomposed to obtain a plurality of intrinsic mode function components of the real-time vibration signal; Envelope detection is performed on each eigenmode function component through Hilbert transform to obtain frequency domain signals of different time scales of the real-time vibration signal.

3. The fault diagnosis method for a CNC machine tool spindle according to claim 2, characterized in that: The selecting a preset fault mode matching the target fault mode from a pre-built spindle fault knowledge base containing a plurality of preset fault modes comprises: Based on a concept similarity algorithm, a semantic mapping relationship between the target fault mode and the plurality of preset fault modes is established; According to the semantic mapping relationship between the target fault mode and the multiple preset fault modes, a preset fault mode having the highest similarity to the target fault mode is selected from the multiple preset fault modes.

4. The fault diagnosis method for a CNC machine tool spindle according to claim 1, characterized in that: The spindle fault knowledge base is constructed by the following steps: Constructing an ontology class of the spindle fault knowledge base, wherein the ontology class includes preset fault modes and fault resolution strategies of the spindle bearing; Defining object properties of all objects in the ontology class to associate objects that have corresponding relationships; Define the data attributes of all objects in the ontology class.

5. The fault diagnosis method for a CNC machine tool spindle according to claim 4, characterized in that: The entity class also includes the fault type, fault location, fault phenomenon and fault cause of the main shaft bearing, and the method also includes: According to the determined target fault mode and the fault resolution strategy corresponding to the target fault mode, the fault type, fault location, fault phenomenon and fault cause corresponding to the target fault mode are obtained.

6. The fault diagnosis method for a CNC machine tool spindle according to claim 5, characterized in that: A preset fault mode of the main shaft bearing corresponds to multiple fault causes, and obtaining the fault cause corresponding to the target fault mode according to the target fault mode and the fault solution strategy corresponding to the target fault mode includes: Based on the preset SWRL rules, knowledge reasoning is performed on the target fault mode and the fault resolution strategy corresponding to the target fault mode to obtain the fault cause corresponding to the target fault mode.

7. The fault diagnosis method for a CNC machine tool spindle according to any one of claims 1 to 6, characterized in that: The target failure mode is obtained by classifying the high-dimensional features based on a random forest algorithm.

8. A fault diagnosis system for a CNC machine tool spindle, characterized in that: The device comprises: A signal acquisition module is used to obtain a frequency domain signal corresponding to a real-time vibration signal of a spindle bearing of a CNC machine tool; A feature extraction module, used for performing feature extraction on the frequency domain signal through a deep belief network to obtain high-dimensional features of the frequency domain signal, wherein the high-dimensional features are used to indicate a target fault mode of the spindle bearing; The deep belief network includes a plurality of stacked restricted Boltzmann machines and a BP classifier, each restricted Boltzmann machine is used to extract the features of the frequency domain signal, and the direction from the bottom layer to the top layer of the plurality of stacked restricted Boltzmann machines is the feature transfer direction of the deep belief network, and the features extracted by each restricted Boltzmann machine are the input features of the next restricted Boltzmann machine, so as to mine the high-dimensional features of the frequency domain signal; the BP classifier is used to correct the network parameters of the deep belief network through reverse error propagation; A fault diagnosis module is used to select a preset fault mode that matches the target fault mode from a pre-built spindle fault knowledge base containing multiple preset fault modes, and use the fault solution strategy corresponding to the selected preset fault mode in the spindle fault knowledge base as the fault solution strategy corresponding to the target fault mode.

9. A computer device, characterized in that: include: at least one processor; And, a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the fault diagnosis method for the spindle of a CNC machine tool as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the fault diagnosis method for a CNC machine tool spindle according to any one of claims 1 to 7 is implemented.

Images