Electric tool fault type identification method and system based on data driving

Abstract

The invention discloses an electric tool fault type identification method and system based on data driving, which are used in the field of fault identification, and the method comprises the following steps: collecting operation data of an electric nail gun under different environment conditions and operation states; processing including time domain analysis and frequency domain analysis is executed on the operation data, and a waveform index and a rough principal component analysis method are used for extracting a feature set for distinguishing fault types in an analysis processing result; constructing a fault type identification model by using a support vector machine combination of K neighbors; performing fault type detection and recognition on the electric nail gun by using the trained fault type recognition model; and obtaining the fault type of the electric nail gun according to the analysis result. According to the invention, fault type detection and identification are carried out by using the model to quickly and accurately determine the fault type of the electric nail gun, the efficiency and accuracy of fault diagnosis are improved, and the cause and the feature of the fault are obtained by comparing the difference between different fault types.

Description

Technical Field

[0001] The present invention relates to the field of fault identification, and in particular, to a method and system for identifying the fault type of an electric tool based on data-driven. Background Art

[0002] In many industries such as decoration and construction, electric nail guns are a key basic tool that is essential to the construction process. With the rapid advancement of technology, product testing methods are constantly being updated, gradually eliminating labor-intensive manual testing methods. Manual testing not only consumes a lot of manpower and material resources, but the duration and accuracy of the test are often affected by the status of the executor. On the contrary, the automated control testing platform has gradually become the mainstream with its high efficiency, cost-effectiveness and high precision, promoting the development of testing technology from traditional manual operation to automation and intelligence. However, in this transformation process, some problems have also been exposed. On the one hand, although the requirements for electric nail gun fault detection and identification remain unchanged, complex operating conditions often cause the automation system to be unable to accurately identify and classify the faults of the power tool, affecting the automatic start-stop control and recording of the experiment. On the other hand, the large amount of data collected during the test is often only stored but not fully utilized, resulting in the problem of data overload. Therefore, there is an urgent need for a method that can effectively utilize the collected data and combine it with the specific needs of the factory for power tool fault identification. An automated fault identification model is constructed through data analysis to meet the industrial needs of the automated testing platform in the power tool field for fault identification.

[0003] Traditional fault diagnosis methods often only focus on a single data dimension (usually the time domain or frequency domain) and lack a comprehensive understanding of the equipment's working status and potential faults. This single perspective limits the model's ability to distinguish different fault types in complex situations. Old methods often rely on experience or simple statistical methods for feature selection, and do not fully utilize all the information in the data, causing the model to ignore subtle differences in fault signals, increasing the risk of misdiagnosis and missed diagnosis. Due to over-reliance on performance on specific data sets, models in existing technologies have weak generalization capabilities on unseen data and are prone to overfitting, resulting in reduced performance in practical applications. Existing models often lack sufficient flexibility to adapt to different fault diagnosis scenarios, and parameter selection and adjustment often rely on experience and cannot automatically adapt to changes in data.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In order to overcome the above problems, the present invention aims to propose a data-driven electric tool fault type identification method and system, with the aim of solving the problem that the old methods often rely on experience selection or simple statistical methods in feature selection, and do not make full use of all the information in the data, resulting in the model may ignore subtle differences in fault signals, increasing the risk of misdiagnosis and missed diagnosis. Due to over-reliance on performance on specific data sets, the models in the prior art have weak generalization capabilities on unseen data and are prone to overfitting, resulting in performance degradation in practical applications. Existing models often lack sufficient flexibility to adapt to different fault diagnosis scenarios, and parameter selection and adjustment often rely on experience and cannot automatically adapt to changes in data.

[0006] To this end, the specific technical solution adopted by the present invention is as follows:

[0007] According to one aspect of the present invention, a method for identifying a power tool fault type based on data drive is provided, and the method for identifying a power tool fault type comprises the following steps:

[0008] S1. Collect the operation data of the electric nail gun under different environmental conditions and operating states;

[0009] S2. Performing time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis and processing results using waveform indicators and rough principal component analysis;

[0010] S3. Based on the extracted feature set and the preset task requirements, a fault type recognition model is constructed using a K-nearest neighbor support vector machine combination, and the fault type recognition model is trained using a data set of known fault types;

[0011] S4. Use the trained fault type recognition model to detect and identify the fault type of the electric nail gun, and compare and analyze the differences between different fault types based on the recognition results;

[0012] S5. Obtain the fault type of the electric nail gun based on the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

[0013] Optionally, performing processing including time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis processing results using waveform indicators and rough principal component analysis method comprises the following steps:

[0014] S21, performing time domain analysis on the operation data, calculating statistics of the operation data, and extracting waveform indicators;

[0015] S22, converting the operation data from the time domain to the frequency domain using fast Fourier transform, and extracting spectrum features;

[0016] S23, constructing an initial feature set using waveform indicators and spectrum features;

[0017] S24, using the conditional reduction attribute in the rough set theory, evaluating the feature set to obtain the feature set of the attribute;

[0018] S25, calculating a heuristic function value for each feature in the attribute feature set, and evaluating the contribution to the fault type classification according to the heuristic function value;

[0019] S26. Use the rough principal component analysis method to process the attribute feature set and combine it with the waveform index to form the final feature set.

[0020] Optionally, converting the operation data from the time domain to the frequency domain using a fast Fourier transform and extracting the spectrum features comprises the following steps:

[0021] S221, setting the data window and sampling frequency of the operation data, and performing time domain preprocessing;

[0022] S222, performing fast Fourier transform on the pre-processed operation data to convert the time domain signal of the operation data into a frequency domain signal;

[0023] S223, analyzing the spectrum obtained by fast Fourier transform to identify the frequency components in the signal;

[0024] S224, calculating the amplitude of the frequency component to determine the relative strength of different frequency components in the signal;

[0025] S225. Extracting spectrum features from the amplitude of each frequency component.

[0026] Optionally, the conditional reduction attribute in the rough set theory is used to evaluate the feature set, and obtaining the feature set of the attribute includes the following steps:

[0027] S241, constructing a decision table according to the initial feature set, wherein the frequency domain features in the decision table are used as condition attributes and the corresponding fault types are used as decision attributes;

[0028] S242. Use the attribute importance measurement method in rough set theory to evaluate the contribution of each condition attribute in the decision table to the decision attribute;

[0029] S243, identifying and removing redundant attributes whose contribution to the decision category in the contribution degree is less than a preset threshold, to obtain a simplified attribute subset;

[0030] S244. Use the simplified attribute subset to construct a new decision table, sort the conditional attributes by importance, and obtain a feature set of the attributes.

[0031] Optionally, using the attribute importance measurement method in the rough set theory, evaluating the contribution of each condition attribute in the decision table to the decision attribute includes the following steps:

[0032] S2421. Calculate the dependency of the decision table;

[0033] S2422. For each condition attribute in the decision table, determine a set of data objects classified according to the attribute;

[0034] S2423, calculating the change in the dependency after removing a conditional attribute in the data object set, and calculating the importance of each conditional attribute;

[0035] S2424, performing normalization processing on different condition attributes according to the calculation result of the importance of each condition attribute;

[0036] Among them, the dependence formula for calculating the decision table is:

[0037]

[0038] In the formula, U represents the set of all objects in the decision table;

[0039] C represents the conditional attribute set;

[0040] D represents the decision attribute set;

[0041] U / C represents the equivalent partitioning of the set U of all objects in the decision table according to the conditional attribute set C;

[0042] POS C (X) represents the positive domain of the decision attribute set D relative to the condition attribute set C;

[0043] |PoS C (X)| represents the number of objects in the positive domain;

[0044] |U| represents the total number of objects in the universe U;

[0045] γC(D) represents the dependency of the decision table.

[0046] Optionally, calculating a heuristic function value for each feature in the attribute feature set, and evaluating the contribution to the fault type classification according to the heuristic function value includes the following steps:

[0047] S251, defining a heuristic evaluation function according to the attribute feature set;

[0048] S252. For each feature in the attribute feature set, calculate the heuristic function value of each feature;

[0049] S253, evaluating the contribution of each feature to the fault type classification according to the heuristic function value of each feature;

[0050] S254, a preset contribution threshold is set, and when the contribution of the screening feature to the fault type classification exceeds the preset contribution threshold, an optimized feature set is formed.

[0051] Optionally, using a rough principal component analysis method to process the attribute feature set and combining the waveform index to form a final feature set includes the following steps:

[0052] S261, performing a rough principal component analysis method on the optimized feature set to identify component information of the optimized feature set;

[0053] S262, selecting component information representing data variation according to the result of the rough principal component analysis method;

[0054] S263, constructing a new feature set using the selected component information;

[0055] S264. Combine the waveform indicator with the new feature set, and set the new feature set combined with the waveform indicator as the final feature set.

[0056] Optionally, according to the extracted feature set and the preset task requirements, a fault type recognition model is constructed by using a K-nearest neighbor support vector machine combination, and the fault type recognition model is trained by using a data set of known fault types, including the following steps:

[0057] S31, obtaining a final feature set, and combining the feature set with a known fault type label to form a complete data set;

[0058] S32, determining the K value in the K nearest neighbor algorithm and the kernel function and regularization parameter of the support vector machine;

[0059] S33, using support vector machine to build a fault type recognition model, and integrating the output of K nearest neighbor algorithm into the decision of support vector machine;

[0060] S34, using the similarity weight coefficient calculated by the K nearest neighbor algorithm to optimize and adjust the prediction output of the fault type recognition model, and using the weighted voting mechanism to ensure that the final recognition result reflects both local information and global decision-making;

[0061] S35. Evaluate the performance of the fault type identification model using the test data set;

[0062] S36. According to the results of the performance evaluation, adjust the K value of the fault type identification model and the kernel function and regularization parameter of the support vector machine.

[0063] Optionally, using a trained fault type recognition model to detect and recognize fault types of the electric nail gun, and performing comparative analysis on the differences between different fault types based on the recognition results includes the following steps:

[0064] S41, deploying the trained fault type recognition model to the actual detection environment;

[0065] S42. In a testing environment, collect operation data of the electric nail gun under different working conditions;

[0066] S43, inputting the collected operation data into the fault type identification model, and using the fault type identification model to identify known fault type features, and outputting a specific fault category;

[0067] S44. Based on the output fault categories, use data visualization tools to display the distribution of different fault types and write a detailed diagnostic report.

[0068] Optionally, obtaining the fault type of the electric nail gun according to the analysis result, providing a decision basis and explanation information, taking maintenance measures according to the fault situation, and classifying and managing the fault type of the electric nail gun include the following steps:

[0069] S51, using the fault type recognition model trained in step 1 to detect the electric nail gun and identify the fault type;

[0070] S52, different fault types, analyze the differences and characteristics between each fault;

[0071] S53. Provide maintenance suggestions based on the fault type and analysis results of different fault types;

[0072] S54. Implement corresponding maintenance measures according to the maintenance suggestions provided;

[0073] S55. Classify and record the identified and processed fault types, and feed back the feedback information after maintenance and the fault processing results to the fault type identification model, and classify and manage the fault types of the electric nail gun.

[0074] According to another aspect of the present invention, there is also provided a data-driven electric tool fault type identification system, the system comprising: a data collection module, a feature extraction module, a model building module, a fault detection and identification module and a maintenance management module;

[0075] The data collection module is connected to the model building module through the feature extraction module, and the model building module is connected to the maintenance management module through the fault detection and identification module;

[0076] A data collection module, used for collecting operation data of the electric nail gun under different environmental conditions and operation states;

[0077] A feature extraction module is used to perform time domain analysis and frequency domain analysis on the operation data, and extract a feature set for distinguishing fault types from the analysis and processing results using waveform indicators and rough principal component analysis;

[0078] A model building module is used to build a fault type recognition model based on the extracted feature set and the preset task requirements using a K-nearest neighbor support vector machine combination, and train the fault type recognition model using a data set of known fault types;

[0079] The fault detection and identification module is used to detect and identify the fault type of the electric nail gun using the trained fault type identification model, and to compare and analyze the differences between different fault types based on the identification results;

[0080] The maintenance management module is used to obtain the fault type of the electric nail gun based on the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

[0081] Compared with the prior art, this application has the following beneficial effects:

[0082] 1. The present invention comprehensively understands the working status and potential faults of the equipment by simultaneously focusing on the time domain and frequency domain characteristics of the operation data. The time domain analysis reveals the fluctuation pattern and statistical properties of the signal, while the frequency domain analysis provides the frequency components and energy distribution of the signal. The combination of these two aspects of information can more accurately identify and distinguish different fault types. The extraction of waveform indicators and spectrum features enables the model to capture subtle differences in fault signals, reflecting the basic form and internal structure of the signal, which is crucial for understanding the normal and abnormal states of the equipment. The rough set theory is used to evaluate and simplify the feature set, which helps to remove redundant and irrelevant features, reduce the computational burden, and improve the efficiency of model training and the accuracy of classification. By calculating the heuristic function value, the contribution of each feature to fault classification is evaluated to ensure that the model focuses on the most informative features. The rough principal component analysis method deeply processes the feature set, reduces the feature dimension by extracting the components that best represent the data variation, and retains the most critical information for fault identification. By synthesizing and optimizing the feature set, not only the performance of the fault identification model is improved, but also the interpretability of the model is enhanced. The model can make judgments based on features with rich physical meanings, making the fault diagnosis process more transparent and credible.

[0083] 2. The K-nearest neighbor algorithm of the present invention performs well in processing local data structures and can identify training samples that are most similar to test samples, while the support vector machine has an advantage in finding the optimal separation boundary between different categories. By combining the two, the model can not only effectively process local data structures, but also make accurate classification decisions on a global scale. By integrating the output of the K-nearest neighbor into the decision of the support vector machine, the decision boundary can be adjusted when necessary to make it closer to the distribution of actual data, thereby improving the accuracy of classification. By carefully selecting the K value and the regularization parameter of the support vector machine, the model can be effectively prevented from being too complex and causing errors. The overfitting problem caused by this algorithm is solved, and the performance of the model on unseen data is enhanced. The optimal K value and support vector machine parameters are determined through cross-validation to ensure that the model can maintain high performance on different data sets and enhance the generalization ability of the model. According to different task requirements and data characteristics, the parameters of K-nearest neighbor and support vector machine (such as K value, kernel function type and regularization parameter) are flexibly selected and adjusted to enable the model to better adapt to different fault diagnosis scenarios. The weighted voting mechanism enables the model to consider both local similarity information and global classification boundaries when making the final decision, thus enhancing the model's ability to process complex data.

[0084] 3. The present invention utilizes a trained model to detect and identify fault types, quickly and accurately determine the fault type of the electric nail gun, and significantly improves the efficiency and accuracy of fault diagnosis. By comparing the differences between different fault types, a deeper understanding of the causes and characteristics of the faults is achieved, providing a scientific basis for formulating maintenance strategies. According to the specific analysis results of the fault types, more accurate maintenance suggestions are provided to ensure the effectiveness and pertinence of maintenance measures. The feedback information and fault handling results after maintenance are fed back to the fault type identification model for continuous learning and optimization of the model to enhance its diagnostic performance. Through systematic fault classification and recording, a complete fault database is established to facilitate tracking, analysis and management of fault information. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] The above characteristics, features and advantages of the present invention and their implementation methods and methods will become more clearly understood in conjunction with the following description of the embodiments, which are described in detail in conjunction with the accompanying drawings. Herein, a schematic diagram is shown:

[0086] Figure 1 is a flow chart of a method for identifying a fault type of an electric tool based on data driving according to an embodiment of the present invention;

[0087] Figure 2 The invention is a block diagram of a data-driven electric tool fault type identification system according to an embodiment of the invention.

[0088] In the figure:

[0089] 1. Data collection module; 2. Feature extraction module; 3. Model building module; 4. Fault detection and identification module; 5. Maintenance management module. DETAILED DESCRIPTION

[0090] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0091] According to an embodiment of the present invention, a method and system for identifying electric tool fault types based on data driving are provided.

[0092] The present invention is now described in detail with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a method for identifying a power tool fault type based on data driving is provided, and the method for identifying a power tool fault type comprises the following steps:

[0093] S1. Collect the operation data of the electric nail gun under different environmental conditions and operating states.

[0094] It should be explained that first, the objectives and requirements of data collection need to be clarified, such as fault detection or performance evaluation, and key data types (such as temperature and vibration) need to be determined. Then, a data collection plan should be designed, including the selection and installation of appropriate sensors (temperature sensors, accelerometers, etc.) and the development or configuration of software systems for data reading, storage and transmission. The data collection process should cover the performance of electric nail guns under various environmental conditions and operating states, and the data should be properly labeled for easy analysis. The collected data needs to be cleaned and preprocessed (such as normalization) to improve its quality and usability. Finally, the integrity and validity of the data should be verified through preliminary analysis, and the collection plan should be adjusted as needed to ensure that the resulting data set can support subsequent feature extraction, model training and fault type identification.

[0095] S2. Performing time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis and processing results using waveform indicators and rough principal component analysis.

[0096] Preferably, performing processing including time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis processing results using waveform indicators and rough principal component analysis method comprises the following steps:

[0097] S21. Perform time domain analysis on the operation data, calculate the statistics of the operation data, and extract waveform indicators (usually including calculating basic statistics of the data, such as mean, standard deviation, maximum and minimum values, etc. and waveform indicators, such as waveform factor, pulse factor, etc. The extraction of waveform indicators can capture the characteristics of the time series for subsequent identification of fault types).

[0098] S22. Use fast Fourier transform to convert the operation data from the time domain to the frequency domain and extract spectrum features (extracting spectrum features, which are particularly critical for distinguishing fault types that differ in frequency distribution).

[0099] Preferably, converting the operating data from the time domain to the frequency domain using a fast Fourier transform and extracting the spectrum features comprises the following steps:

[0100] S221, setting the data window and sampling frequency of the operation data, and performing time domain preprocessing (first determine the sampling frequency of the operation data, set it according to the highest frequency of the signal you want to capture, according to the Nyquist theorem, the sampling frequency should be at least twice the highest frequency of the signal, then select an appropriate data window to help reduce the edge effect of the fast Fourier transform result);

[0101] S222, performing fast Fourier transform on the preprocessed operation data, converting the time domain signal of the operation data into a frequency domain signal (using a mathematical software package or a programming language library (such as NumPy or MATLAB in Python) to perform fast Fourier transform, taking the time series data as input, applying the FFT algorithm, and obtaining the frequency domain signal);

[0102] S223. Analyze the spectrum graph obtained by fast Fourier transform and identify the frequency components in the signal (use a visualization tool to draw a spectrum graph of the fast Fourier transform result, which will show the frequency components and their strengths of the signal. By observing the spectrum graph, the main frequency components in the signal can be identified. For example, a fault signal will show an abnormal peak at a specific frequency);

[0103] S224, calculating the amplitude of the frequency component, determining the relative strength of different frequency components in the signal (from the FFT result, extracting the amplitude of each frequency component, the amplitude represents the relative strength of the frequency component in the signal, and calculating the absolute value of each point from the fast Fourier transform result);

[0104] S225. Extract spectral features from the amplitude of each frequency component (based on the obtained frequency component amplitude, a variety of spectral features can be extracted, such as the main frequency, frequency bandwidth, frequency band with concentrated energy, etc. The features are calculated, for example, calculating the average amplitude of all significant frequency components, or finding the frequency range with the most concentrated energy).

[0105] S23. Construct an initial feature set using waveform indicators and spectrum features.

[0106] In addition, the following explanations are required: constructing the initial feature set involves combining the waveform indicators and spectral features extracted in the previous steps. First, it is necessary to ensure that the waveform indicators and spectral features have been accurately calculated and recorded. Then the indicators and features are summarized into a unified data structure, such as a table or vector, where each row represents a data point and each column represents a feature. To achieve this step, programming languages ​​​​(such as Python) and data processing libraries (such as Pandas or NumPy) are used to merge the two types of features together. It is important to maintain data consistency and ensure that each feature is processed and standardized in the same way. In addition, in order to improve the efficiency and accuracy of subsequent model training, feature selection or dimensionality reduction processing is performed, such as using principal component analysis (PCA) or selecting K most important features. The process of merging and preprocessing is a key step in forming a data structure that can effectively reflect the characteristics of the operating data and provide input for the fault identification model.

[0107] S24. Use the conditional reduction attribute in the rough set theory to evaluate the feature set and obtain the feature set of attributes (convert the initial feature set into a decision table, where the conditional attributes are frequency domain features and the decision attributes are the corresponding fault types, usually implemented through data processing software to ensure that the characteristics of each instance or data point correctly correspond to its fault type).

[0108] Preferably, the conditional reduction attribute in the rough set theory is used to evaluate the feature set, and obtaining the feature set of the attribute includes the following steps:

[0109] S241, constructing a decision table according to the initial feature set, wherein the frequency domain features in the decision table are used as conditional attributes and the corresponding fault types are used as decision attributes (the dependency of the entire decision table is calculated, i.e., to what extent the decision attribute depends on the conditional attribute, which is achieved through statistical analysis software);

[0110] S242. Use the attribute importance measurement method in the rough set theory to evaluate the contribution of each condition attribute in the decision table to the decision attribute (implemented through specialized software tools or self-written programs, using Python, by defining functions to calculate the dependency and attribute importance, first calculate the dependency of the entire decision table as a baseline, then remove each condition attribute one by one, recalculate the dependency, in order to measure the change in the dependency after the condition attribute is removed, and thus determine its importance);

[0111] S243, identifying and removing redundant attributes whose contribution to the decision category is less than a preset threshold in the contribution degree, and obtaining a simplified attribute subset (implemented in a programming environment such as Python, by comparing the importance of each attribute with a preset threshold to determine whether to retain the attribute, and after removing redundant attributes, what remains is a conditional attribute subset with a higher contribution degree to the fault type);

[0112] S244. Use the simplified attribute subset to build a new decision table, sort by the importance of the conditional attributes, and obtain the feature set of the attributes (the conditional attributes can be sorted according to their contribution to the fault type, with the most contributing attributes placed in front. This new decision table is used for in-depth data analysis or as input for machine learning model training. In Python, the Pandas library can be used to process data and rebuild the decision table, and the sorting can be performed based on the attribute importance calculated previously).

[0113] Preferably, using the attribute importance measurement method in the rough set theory, evaluating the contribution of each condition attribute to the decision attribute in the decision table includes the following steps:

[0114] S2421. Calculate the dependency of the decision table;

[0115] Among them, the dependence formula for calculating the decision table is:

[0116]

[0117] In the formula, U represents the set of all objects in the decision table;

[0118] C represents the conditional attribute set;

[0119] D represents the decision attribute set;

[0120] U / C represents the equivalent partitioning of the set U of all objects in the decision table according to the conditional attribute set C;

[0121] POS C (X) represents the positive domain of the decision attribute set D relative to the condition attribute set C. The positive domain includes all objects whose values ​​of the decision attribute set D can be uniquely determined by the condition attribute set C;

[0122] |POS C (X)| represents the number of objects in the positive domain;

[0123] |U| represents the total number of objects in the universe U;

[0124] γC(D) represents the dependency of the decision table.

[0125] S2422. For each condition attribute in the decision table, determine a set of data objects classified according to the attribute (determine the classification effect for each condition attribute, i.e., group the data objects according to the attribute);

[0126] S2423. Calculate the change in dependency after removing a conditional attribute in the data object set, and calculate the importance of each conditional attribute (evaluate the importance of an attribute by removing an attribute and observing the change in dependency, and test the impact of each conditional attribute);

[0127] S2424. According to the calculation result of the importance of each condition attribute, normalize the different condition attributes.

[0128] S25. Calculate a heuristic function value for each feature in the attribute feature set, and evaluate the contribution to the fault type classification according to the heuristic function value.

[0129] Preferably, calculating the heuristic function value for each feature in the attribute feature set, and evaluating the contribution to the fault type classification according to the heuristic function value comprises the following steps:

[0130] S251. Define a heuristic evaluation function according to the attribute feature set (define an evaluation function that can reflect the contribution of the feature to the fault type classification. The function is designed and written based on the statistical properties of the feature, such as information gain, Gini impurity, correlation coefficient, etc. The input is the feature value and the output is the evaluation score of the feature);

[0131] S252. For each feature in the attribute feature set, the heuristic function value of each feature is calculated (for each feature in the attribute feature set, the function value thereof is calculated using the defined heuristic evaluation function, which requires writing a loop or using vectorized operations, iterating each feature, calling the evaluation function, and obtaining an evaluation score for each feature);

[0132] S253. According to the heuristic function value of each feature, the contribution of the feature to the fault type classification is evaluated (according to the heuristic function value of each feature, its contribution to the fault type classification is evaluated. The contribution is determined by comparing the function value. The larger the function value, the higher the contribution of the feature to the classification. This is achieved by sorting or classifying the function values ​​so that you can intuitively see which features are more important).

[0133] S254, preset contribution threshold, when the contribution of the screening feature to the fault type classification exceeds the preset contribution threshold, and form an optimized feature set (set the contribution threshold, only retain those features whose contribution exceeds the threshold, so as to form an optimized feature set, and achieve it by comparing the evaluation score of each feature with the threshold for screening. In actual operation, this usually involves setting filtering conditions, which are then applied to the entire feature set).

[0134] S26. Use the rough principal component analysis method to process the attribute feature set and combine it with the waveform index to form the final feature set.

[0135] Preferably, using the rough principal component analysis method to process the attribute feature set and combining the waveform index to form the final feature set includes the following steps:

[0136] S261, performing a rough principal component analysis method on the optimized feature set to identify component information of the optimized feature set (applying a rough principal component analysis method (Rough PCA) to the optimized feature set involves using data analysis software or programming languages, such as PCA-related functions or libraries in MATLAB or Python. Rough PCA is an extension of traditional PCA, taking into account uncertainty of data, taking the optimized feature set as input, and then performing PCA to extract principal components);

[0137] S262. According to the result of the rough principal component analysis method, select the component information representing the data variation (the PCA results obtained in the analysis, especially the proportion of data variation explained by each principal component, usually select the first few principal components that explain most of the data variation. This process can be assisted by viewing the PCA variance explanation ratio chart, and select the principal component whose cumulative contribution rate reaches a certain threshold, such as 90%);

[0138] S263, constructing a new feature set using the selected component information (constructing a new feature set based on the selected representative components, projecting the original feature set into the new space formed by the selected principal components, using the obtained projection values ​​as the new feature set, and using the results of the PCA transformation to obtain new feature values);

[0139] S264. Combining the waveform indicators with the new feature set, and setting the new feature set combined with the waveform indicators as the final feature set (combining the waveform indicators with the new feature set obtained by rough PCA processing is accomplished by simply incorporating the waveform indicators as additional features into the feature set formed by PCA).

[0140] It needs to be explained that, for example: Electric nail guns are commonly used tools in carpentry, decoration and other fields. They can complete the nail driving work quickly and efficiently. However, after long-term use or improper maintenance, electric nail guns will malfunction, such as poor firing, nail jamming and other problems. In order to ensure work efficiency and worker safety, it is very important to diagnose and repair electric nail gun malfunctions in a timely manner.

[0141] In a laboratory environment, the operation data of a batch of electric nail guns under normal and various simulated fault conditions (such as motor damage, poor battery contact, wear of mechanical parts, etc.) are collected. The data mainly include vibration signals and sound signals, which can reflect the working status of the electric nail gun. Time domain analysis is performed on the collected operation data, and statistics (mean, standard deviation, etc.) and waveform indicators (waveform factor, pulse factor, etc.) are calculated to capture the characteristics of normal and abnormal working conditions of the electric nail gun. Fast Fourier transform (FFT) is used to convert the operation data from time domain to frequency domain, and spectral features are extracted. Through spectral analysis, the changes in frequency components caused by specific faults can be identified. For example, motor damage leads to enhanced vibration of specific frequencies. The results of time domain and frequency domain analysis are combined to construct an initial feature set containing waveform indicators and spectral features. Rough set theory and heuristic function evaluation method are used to evaluate and optimize the feature set, and the features that are most helpful in distinguishing fault types are screened out. Rough principal component analysis is performed on the optimized feature set to extract the main components of the data, and the components are combined with waveform indicators to form the final fault diagnosis feature set.

[0142] S3. Based on the extracted feature set and the preset task requirements, a fault type recognition model is constructed using a K-nearest neighbor support vector machine combination, and the fault type recognition model is trained using a data set of known fault types.

[0143] Preferably, according to the extracted feature set and the preset task requirements, a fault type recognition model is constructed by using a K-nearest neighbor support vector machine combination, and the fault type recognition model is trained by using a data set of known fault types, including the following steps:

[0144] S31, obtain the final feature set, and combine the feature set with the known fault type label to form a complete data set (it is necessary to obtain the final optimized feature set from the previous step, and each data point of the feature set needs to have a corresponding fault type label. The data set formed in this way contains both feature information and the label of each fault type, forming a complete supervised learning data set);

[0145] S32, determine the K value in the K nearest neighbor algorithm and the kernel function and regularization parameter of the support vector machine (the key parameter of the K nearest neighbor algorithm is the K value, that is, the number of nearest neighbors considered when making classification decisions, and the key parameters of the support vector machine include the kernel function type, such as linear kernel, polynomial kernel, radial basis function RBF kernel, etc., and regularization parameter, control the complexity of the model, prevent overfitting, and the selection of parameters is usually determined by cross-validation to ensure that the generalization performance of the model on unseen data is optimal);

[0146] S33. Use support vector machine to build a fault type recognition model, and integrate the output of K nearest neighbor algorithm into the decision of support vector machine (use SVM to build a basic fault type recognition model, SVM is trained to distinguish different types of faults, and then, integrate the output of K nearest neighbor algorithm into the decision process of SVM. Specifically, K nearest neighbor algorithm can find K nearest neighbors of each test sample in feature space, and the labels and distance information of K nearest neighbors can be used to adjust the decision boundary of SVM or perform post-processing);

[0147] Among them, the prediction output calculation formula that integrates the output of the K nearest neighbor algorithm into the decision of the support vector machine is:

[0148] Y′=∑Y i W s ;

[0149]

[0150] In the formula, Y′ represents the predicted output of the feature set;

[0151] Y i 1D output representing the feature set;

[0152] W s Represents the weight of each fault type in the feature set;

[0153] n s Indicates the number of fault types;

[0154] K represents the distance between the nearest neighbors of the test sample and the training sample.

[0155] S34. Optimize and adjust the prediction output of the fault type recognition model using the similarity weight coefficient calculated by the K nearest neighbor algorithm, and use the weighted voting mechanism to ensure that the final recognition result reflects both local information and global decisions (the prediction output of the SVM is optimized by the similarity weight coefficient calculated by the K nearest neighbor algorithm, which is implemented by the weighted voting mechanism, that is, according to the fault type and corresponding distance of the K nearest neighbors of the test sample, it is converted into a similarity weight to adjust the prediction of the SVM. The final decision not only considers the global decision boundary provided by the SVM, but also reflects the local similarity information);

[0156] S35. Use the test data set to evaluate the performance of the fault type recognition model (use a part of the data as the test set to evaluate the performance of the fault type recognition model. Common evaluation indicators include accuracy, recall, F1 score, etc. The evaluation can be completed through cross-validation, which helps to estimate the generalization ability of the model on new data);

[0157] S36. According to the results of the performance evaluation, adjust the K value of the fault type identification model and the kernel function and regularization parameter of the support vector machine.

[0158] S4. Use the trained fault type recognition model to detect and identify the fault category of the electric nail gun, and compare and analyze the differences between different fault types based on the recognition results.

[0159] Preferably, using the trained fault type recognition model to detect and identify the fault type of the electric nail gun, and performing comparative analysis on the differences between different fault types based on the recognition results includes the following steps:

[0160] S41. Deploy the trained fault type identification model to the actual detection environment.

[0161] S42. In a testing environment, collect operation data of the electric nail gun under different working conditions.

[0162] It should be explained that the first step of integrating the model into the detection environment usually involves a technical assessment to determine the compatibility of the model with existing systems (such as the control system of the production line, tool monitoring equipment, etc.); the collaboration of software engineers and system administrators is required to ensure that all interfaces can run smoothly; model deployment also includes the adjustment of environmental parameters (such as the influence of environmental factors such as temperature and humidity) to adapt to the actual working environment; adjust the input parameters of the model according to the actual conditions, such as adjusting the normalization and standardization process of data input; before the model is officially put into use, a series of tests are required, such as simulated fault detection, to verify the accuracy and response speed of the model; testing helps identify Identify possible problems and make adjustments before actual use; determine which key parts of the electric nail gun to install sensors, such as the motor, trigger mechanism and battery interface; the sensors can be vibration sensors, temperature sensors, current and voltage sensors, etc., depending on the expected fault type and detection requirements; set an appropriate sampling rate and resolution to ensure that the collected data can reflect the fault characteristics without excessively occupying storage space. For example, vibration data requires a higher sampling rate to capture rapidly changing dynamics. Ensure that the transmission of data from the sensor to the processing unit is secure and reliable, and use appropriate data encryption and transmission protocols to avoid data loss or tampering during transmission.

[0163] S43, inputting the collected operation data into a fault type identification model, and using the fault type identification model to identify known fault type features, and outputting a specific fault category.

[0164] It should be explained that the collected data is cleaned to remove invalid or erroneous data points; then, the data is formatted into a format acceptable to the model, such as converting the timestamp into a unified format, normalizing all sensor data to the same dimension, and extracting features that are helpful for fault diagnosis from the raw data according to the needs of the model, such as the peak value of the vibration frequency, the duration of the current anomaly, etc. The features are normalized or standardized to eliminate the influence between different dimensions, so that the model can more easily identify patterns in the data; the preprocessed data is input into the fault type identification model; it is implemented through a programming interface (API) or a direct database connection, the model receives the input data, analyzes and processes the data according to the patterns learned in its training, identifies matching fault features, and based on the identification results, the model outputs a specific fault category, including the severity of the fault, confidence and other information, performs a preliminary verification of the fault category output by the model, compares the model prediction with known fault cases, and records the diagnosis results.

[0165] S44. Based on the output fault categories, use data visualization tools to display the distribution of different fault types and write a detailed diagnostic report.

[0166] It should be explained that suitable data visualization tools should be selected, such as Tableau, Power BI, or Matplotlib and Seaborn libraries using Python; visualization tools can handle large amounts of data and support multiple chart types, which are suitable for complex data display requirements. Data derived from fault identification models usually need further processing before they can be used for visualization, including data cleaning, classification, and conversion; for example, the fault category, time of occurrence, and frequency are organized into a table format; appropriate chart types are selected according to the characteristics of the data and display requirements; for example, pie charts are used to represent the proportion of different fault types, and trend charts are used to show the time series analysis of fault occurrence; visual elements of charts, such as color, font, and layout, are adjusted to ensure that the information is clear and easy to understand, and interactive design of charts is performed, such as adding filters and exploration functions to enhance the user experience, and the basic structure of the report is determined, including introduction, methodology, result analysis, conclusions, and recommendations; this helps keep the report clear and easy to read and understand, and the process and results of fault detection are described in detail in the report, including the definition of fault types, detection methods, frequency of occurrence, and possible influencing factors.

[0167] It should be explained that test data, including vibration and sound signals, are first collected from the electric nail gun to be tested. Then, the same preprocessing steps as the training data are applied to the test data, including feature extraction and feature optimization, to ensure that the data format and feature type are consistent with the data used when training the model. Next, the processed test data is classified using the trained fault type recognition model to output the fault type prediction results for each test sample. Fault type distribution analysis: Count the predicted frequencies of each fault type to understand which fault types are most common and which are rare.

[0168] Fault diagnosis accuracy analysis: Evaluate the accuracy of the model on each fault type and determine the strengths and weaknesses of the model in identifying specific fault types.

[0169] Confusion Matrix Analysis: Construct a confusion matrix to visualize the model's classification performance for each fault type and identify whether the model has a tendency to misclassify one fault type as another.

[0170] Feature Importance Analysis: Analyze the key features that determine the fault type classification and understand the criteria for distinguishing different fault types.

[0171] Implementation steps: Based on the analysis results, identify potential opportunities for improving model performance. For example, if you find that the model is not accurate for certain fault types, you need to return to the feature engineering stage to find more discriminating features or adjust model parameters to improve performance. For fault types that are confused by the model, you can deeply analyze the commonalities and differences between these types, and adjust or add features to improve the model's ability to distinguish.

[0172] S5. Obtain the fault type of the electric nail gun based on the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

[0173] Preferably, obtaining the fault type of the electric nail gun according to the analysis result, providing decision-making basis and explanation information, taking maintenance measures according to the fault situation, and classifying and managing the fault type of the electric nail gun include the following steps:

[0174] S51, using the fault type recognition model trained in step 1 to detect the electric nail gun and identify the fault type (using the previously trained model to predict the fault type of the test data of the electric nail gun. This requires collecting the operation data of the electric nail gun; such as vibration and sound data, and passing it through the same preprocessing and feature extraction process, and then using the model for classification);

[0175] S52, different fault types, analyze the differences and characteristics between each fault (analyze and summarize the typical characteristics of each fault type and its differences from other fault types, by looking at the key features identified by the model in the feature importance assessment, and by analyzing actual fault cases to understand the manifestations and causes of specific fault types);

[0176] S53. Provide maintenance suggestions based on the fault type and the analysis results of different fault types (based on the characteristic analysis of each fault type, provide targeted maintenance suggestions. For example, if a certain fault type is related to poor battery contact, the maintenance suggestions include checking and cleaning the battery interface. The maintenance suggestions should be clear and specific so that maintenance personnel can understand and implement them easily);

[0177] S54. Implement corresponding maintenance measures according to the maintenance suggestions provided (implement corresponding maintenance measures according to the maintenance suggestions provided, including replacement of damaged parts, adjustment of equipment configuration, cleaning and lubrication. During the maintenance process, maintenance personnel shall record the details of the maintenance activities, including the parts replaced, maintenance time and other information);

[0178] S55. Classify and record the identified and processed fault types, and feed back the feedback information after maintenance and the fault processing results to the fault type identification model, and classify and manage the fault types of the electric nail gun.

[0179] It should be explained that after the maintenance is completed, the handled fault types are classified and recorded, and a fault type database is established. This not only helps future fault diagnosis and maintenance decisions, but also provides feedback information for the model, which can be used to deeply optimize and train the fault identification model. At the same time, by classifying and managing the fault types of electric nail guns, we can better understand the trends and patterns of fault occurrence and provide data support for preventive maintenance.

[0180] The following is an additional example: A production plant has a key electric nail gun. Recently, it was found that the electric nail gun occasionally vibrates abnormally during operation. It is suspected that some internal components are faulty. In order to accurately diagnose the problem and perform repairs in time to avoid long-term downtime of the production line, the engineer decided to use the above technical path for fault detection;

[0181] First, sensors, including vibration sensors and temperature sensors, are installed on the electric nail gun to continuously record operating data for one week, including data during normal operation and data during abnormal operation.

[0182] Engineers conducted time domain analysis on the collected vibration data and calculated various statistics (mean, standard deviation, etc.) and waveform indicators (waveform factor, pulse factor, etc.). Through analysis, they found that under certain abnormal conditions, the pulse factor of the vibration data was significantly higher than the normal level, indicating that the device was impacted or had loose components during the period.

[0183] By converting the time domain data into the frequency domain using Fast Fourier Transform (FFT), and analyzing the spectral characteristics of the vibration data, it was found that under abnormal conditions, the vibration amplitude of a certain frequency increased significantly, which was due to the resonance caused by the damage of a specific component.

[0184] Combining the results of time domain and frequency domain analysis, an initial feature set consisting of statistics, waveform indicators, and spectral features is constructed.

[0185] The feature set is evaluated and optimized through rough set theory and heuristic function evaluation method, and the features that contribute most to the fault type classification are screened out.

[0186] Rough principal component analysis is applied to process the optimized feature set to identify the most important components, which are then used to build a fault diagnosis model.

[0187] The waveform indicators are combined to form a final feature set, which is used to train a machine learning model for fault classification.

[0188] Using the collected data and the above methods, engineers successfully established a fault diagnosis model that can accurately identify and classify the normal state and multiple abnormal states of the electric nail gun. Through the analysis of this model, it was found that the main cause of the abnormal vibration was the damage of a certain bearing. Subsequently, the engineers replaced the bearing and the device resumed normal operation, avoiding production losses.

[0189] According to another embodiment of the present invention, Figure 2 As shown, a data-driven electric tool fault type identification system is also provided, the system comprising: a data collection module 1, a feature extraction module 2, a model building module 3, a fault detection and identification module 4 and a maintenance management module 5;

[0190] The data collection module 1 is connected to the model building module 3 via the feature extraction module 2, and the model building module 3 is connected to the maintenance management module 5 via the fault detection and identification module 4;

[0191] A data collection module 1, used to collect operation data of the electric nail gun under different environmental conditions and operating states;

[0192] Feature extraction module 2, used to perform time domain analysis and frequency domain analysis on the operation data, and extract a feature set for distinguishing fault types from the analysis results using waveform indicators and rough principal component analysis;

[0193] Model building module 3, used to build a fault type recognition model using a K-nearest neighbor support vector machine combination according to the extracted feature set and preset task requirements, and train the fault type recognition model using a data set of known fault types;

[0194] The fault detection and identification module 4 is used to detect and identify the fault type of the electric nail gun using the trained fault type identification model, and to compare and analyze the differences between different fault types based on the identification results;

[0195] The maintenance management module 5 is used to obtain the fault type of the electric nail gun according to the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

[0196] In summary, with the help of the above technical solutions of the present invention, the present invention comprehensively understands the working status and potential faults of the equipment by paying attention to the time domain and frequency domain characteristics of the operation data at the same time. The time domain analysis reveals the fluctuation pattern and statistical properties of the signal, while the frequency domain analysis provides the frequency components and energy distribution of the signal. The combination of these two aspects of information can more accurately identify and distinguish different fault types; the extraction of waveform indicators and spectrum features enables the model to capture the subtle differences in the fault signal, reflecting the basic form and internal structure of the signal, which is crucial for understanding the normal and abnormal states of the equipment. The rough set theory is used to evaluate and simplify the feature set, which helps to remove redundant and irrelevant features, reduce the computational burden, and improve the efficiency of model training and the accuracy of classification. By calculating the heuristic function value, the contribution of each feature to fault classification is evaluated to ensure that the model focuses on the most informative features. The rough principal component analysis method deeply processes the feature set, reduces the feature dimension by extracting the components that best represent the data variation, and retains the most critical information for fault identification. By synthesizing and optimizing the feature set, not only the performance of the fault identification model is improved, but also the interpretability of the model is enhanced. The model can make judgments based on features with rich physical meanings, so that the fault The fault diagnosis process is more transparent and credible; the K-nearest neighbor algorithm of the present invention performs well in processing local data structures and can identify training samples that are most similar to test samples, while the support vector machine has an advantage in finding the optimal separation boundary between different categories. By combining the two, the model can not only effectively process local data structures, but also make accurate classification decisions on a global scale. By integrating the output of the K-nearest neighbor into the decision of the support vector machine, the decision boundary can be adjusted when necessary to make it closer to the distribution of actual data, thereby improving the accuracy of classification. By carefully selecting the K value and the regularization parameter of the support vector machine, the model can be effectively prevented from The overfitting problem caused by excessive complexity is solved. The performance of the model on unseen data is enhanced. The optimal K value and support vector machine parameters are determined through cross-validation to ensure that the model can maintain high performance on different data sets and enhance the generalization ability of the model. According to different task requirements and data characteristics, the parameters of K-nearest neighbor and support vector machine (such as K value, kernel function type and regularization parameter) are flexibly selected and adjusted to enable the model to better adapt to different fault diagnosis scenarios. The weighted voting mechanism enables the model to consider both local similarity information and global classification boundaries when making the final decision, thus enhancing the model's ability to handle complex data.The present invention uses the trained model to detect and identify the fault type to quickly and accurately determine the fault type of the electric nail gun, significantly improving the efficiency and accuracy of fault diagnosis. By comparing the differences between different fault types, the causes and characteristics of the faults can be more deeply understood, providing a scientific basis for formulating maintenance strategies. According to the specific analysis results of the fault type, more accurate maintenance suggestions can be provided to ensure the effectiveness and pertinence of maintenance measures. The feedback information after maintenance and the fault handling results are fed back to the fault type identification model for continuous learning and optimization of the model to enhance its diagnostic performance. Through systematic fault classification and recording, a complete fault database is established to facilitate tracking, analysis and management of fault information. ;

[0197] Although the present invention has been disclosed as above with preferred embodiments, the embodiments are merely examples for the convenience of description and are not intended to limit the present invention. Those skilled in the art may make several changes and modifications without departing from the spirit and scope of the present invention. The scope of protection claimed by the present invention shall be based on the claims.

Claims

1. A data-driven method for identifying fault types of electric tools, characterized in that: The electric tool fault type identification method comprises the following steps: S1. Collect the operation data of the electric nail gun under different environmental conditions and operating states; S2. Performing time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis and processing results using waveform indicators and rough principal component analysis; S3. Based on the extracted feature set and the preset task requirements, a fault type recognition model is constructed using a K-nearest neighbor support vector machine combination, and the fault type recognition model is trained using a data set of known fault types; S4. Use the trained fault type recognition model to detect and identify the fault type of the electric nail gun, and compare and analyze the differences between different fault types based on the recognition results; S5. Obtain the fault type of the electric nail gun based on the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

2. The method for identifying the fault type of an electric tool based on data drive according to claim 1, characterized in that: The processing of performing time domain analysis and frequency domain analysis on the operation data respectively, and extracting a feature set for distinguishing fault types from the analysis and processing results by using waveform indicators and rough principal component analysis method comprises the following steps: S21, performing time domain analysis on the operation data, calculating statistics of the operation data, and extracting waveform indicators; S22, converting the operation data from the time domain to the frequency domain using fast Fourier transform, and extracting spectrum features; S23, constructing an initial feature set using the waveform index and the spectrum feature; S24, using the conditional reduction attribute in the rough set theory, evaluating the feature set to obtain the feature set of the attribute; S25, calculating a heuristic function value for each feature in the attribute feature set, and evaluating the contribution to the fault type classification according to the heuristic function value; S26. Use the rough principal component analysis method to process the attribute feature set and combine it with the waveform index to form the final feature set.

3. The method for identifying the fault type of an electric tool based on data drive according to claim 2, characterized in that: The method of converting the operation data from the time domain to the frequency domain by using the fast Fourier transform and extracting the spectrum features comprises the following steps: S221, setting the data window and sampling frequency of the operation data, and performing time domain preprocessing; S222, performing fast Fourier transform on the pre-processed operation data to convert the time domain signal of the operation data into a frequency domain signal; S223, analyzing the spectrum obtained by fast Fourier transform to identify the frequency components in the signal; S224, calculating the amplitude of the frequency component to determine the relative strength of different frequency components in the signal; S225. Extracting spectrum features from the amplitude of each frequency component.

4. The method for identifying the fault type of an electric tool based on data drive according to claim 3, characterized in that: The method of evaluating the feature set by using the conditional reduction attribute in the rough set theory and obtaining the feature set of the attribute comprises the following steps: S241, constructing a decision table according to the initial feature set, wherein the frequency domain features in the decision table are used as condition attributes and the corresponding fault types are used as decision attributes; S242. Use the attribute importance measurement method in rough set theory to evaluate the contribution of each condition attribute in the decision table to the decision attribute; S243, identifying and removing redundant attributes whose contribution to the decision category in the contribution degree is less than a preset threshold, to obtain a simplified attribute subset; S244, construct a new decision table using the simplified attribute subset, sort the condition attributes by importance, and obtain a feature set of the attributes; The method of using the attribute importance measurement method in the rough set theory to evaluate the contribution of each condition attribute in the decision table to the decision attribute includes the following steps: S2421. Calculate the dependency of the decision table; S2422. For each condition attribute in the decision table, determine a set of data objects classified according to the attribute; S2423, calculating the change in the dependency after removing a conditional attribute in the data object set, and calculating the importance of each conditional attribute; S2424, performing normalization processing on different condition attributes according to the calculation result of the importance of each condition attribute; Among them, the dependence formula for calculating the decision table is: In the formula, U represents the set of all objects in the decision table; C represents the conditional attribute set; D represents the decision attribute set; U / C represents the equivalent partitioning of the set U of all objects in the decision table according to the conditional attribute set C; POS C (X) represents the positive domain of the decision attribute set D relative to the condition attribute set C; |POS C (X)| represents the number of objects in the positive domain; |U| represents the total number of objects in the universe U; YC(D) represents the dependency of the decision table.

5. The method for identifying the fault type of an electric tool based on data drive according to claim 4, characterized in that: The step of calculating the heuristic function value for each feature in the attribute feature set and evaluating the contribution to the fault type classification according to the heuristic function value comprises the following steps: S251, defining a heuristic evaluation function according to the attribute feature set; S252. For each feature in the attribute feature set, calculate the heuristic function value of each feature; S253, evaluating the contribution of each feature to the fault type classification according to the heuristic function value of each feature; S254, a preset contribution threshold is set, and when the contribution of the screening feature to the fault type classification exceeds the preset contribution threshold, an optimized feature set is formed.

6. The method for identifying the fault type of an electric tool based on data drive according to claim 5, characterized in that: The method of processing the attribute feature set by using the rough principal component analysis method and combining the waveform index to form the final feature set includes the following steps: S261, performing a rough principal component analysis method on the optimized feature set to identify component information of the optimized feature set; S262, selecting component information representing data variation according to the result of the rough principal component analysis method; S263, constructing a new feature set using the selected component information; S264. Combine the waveform indicator with the new feature set, and set the new feature set combined with the waveform indicator as the final feature set.

7. The method for identifying power tool fault types based on data drive according to claim 1, characterized in that: The method of constructing a fault type recognition model by using a K-nearest neighbor support vector machine combination according to the extracted feature set and the preset task requirements, and training the fault type recognition model by using a data set of known fault types includes the following steps: S31, obtaining a final feature set, and combining the feature set with a known fault type label to form a complete data set; S32, determining the K value in the K nearest neighbor algorithm and the kernel function and regularization parameter of the support vector machine; S33, using support vector machine to build a fault type recognition model, and integrating the output of K nearest neighbor algorithm into the decision of support vector machine; S34, using the similarity weight coefficient calculated by the K nearest neighbor algorithm to optimize and adjust the prediction output of the fault type recognition model, and using the weighted voting mechanism to ensure that the final recognition result reflects both local information and global decision-making; S35. Evaluate the performance of the fault type identification model using the test data set; S36. According to the results of the performance evaluation, adjust the K value of the fault type identification model and the kernel function and regularization parameter of the support vector machine.

8. The method for identifying the fault type of an electric tool based on data drive according to claim 1, characterized in that: The method of using the trained fault type recognition model to detect and identify the fault type of the electric nail gun and performing comparative analysis on the differences between different fault types according to the recognition results includes the following steps: S41, deploying the trained fault type recognition model to the actual detection environment; S42. In a testing environment, collect operation data of the electric nail gun under different working conditions; S43, inputting the collected operation data into the fault type identification model, and using the fault type identification model to identify known fault type features, and outputting a specific fault category; S44. Based on the output fault categories, use data visualization tools to display the distribution of different fault types and write a detailed diagnostic report.

9. The method for identifying the fault type of an electric tool based on data drive according to claim 8, characterized in that: The method of obtaining the fault type of the electric nail gun according to the analysis result, providing a decision basis and explanation information, taking maintenance measures according to the fault situation, and classifying and managing the fault type of the electric nail gun includes the following steps: S51, using the fault type recognition model trained in step 1 to detect the electric nail gun and identify the fault type; S52, different fault types, analyze the differences and characteristics between each fault; S53. Provide maintenance suggestions based on the fault type and analysis results of different fault types; S54. Implement corresponding maintenance measures according to the maintenance suggestions provided; S55. Classify and record the identified and processed fault types, and feed back the feedback information after maintenance and the fault processing results to the fault type identification model, and classify and manage the fault types of the electric nail gun.

10. A data-driven electric tool fault type identification system, used to implement the data-driven electric tool fault type identification method according to any one of claims 1 to 9, characterized in that: The system includes: data collection module, feature extraction module, model building module, fault detection and identification module and maintenance management module; The data collection module is connected to the model building module through the feature extraction module, and the model building module is connected to the maintenance management module through the fault detection and identification module; The data collection module is used to collect the operation data of the electric nail gun under different environmental conditions and operating states; The feature extraction module is used to perform processing including time domain analysis and frequency domain analysis on the operation data, and extract a feature set for distinguishing fault types from the analysis and processing results using waveform indicators and rough principal component analysis; The model building module is used to build a fault type recognition model using a K-nearest neighbor support vector machine combination according to the extracted feature set and the preset task requirements, and train the fault type recognition model using a data set of known fault types; The fault detection and identification module is used to detect and identify the fault type of the electric nail gun using the trained fault type identification model, and to compare and analyze the differences between different fault types based on the identification results; The maintenance management module is used to obtain the fault type of the electric nail gun according to the analysis results, provide decision-making basis and explanation information, take maintenance measures according to the fault situation, and classify and manage the fault type of the electric nail gun.

Images