CN115828164A - Electric nail gun fault type identification method based on data driving - Google Patents

Abstract

The invention discloses a data-driven electric nail gun fault type identification method, which comprises the following steps: analyzing common fault types of the nail gun, and supposing the influence of the faults on the current and the voltage of the motor; collecting data of the nail gun during working through a sensor, collecting current and voltage data values under different working conditions, calculating corresponding energy consumption values, and classifying the data under normal working conditions and fault working conditions; extracting the characteristics of current, voltage and energy consumption values in time domains under fault and normal working conditions, and classifying the fault type identification method according to different characteristic rules expressed by different fault data in the time domains; by applying the method, the test data of the electric nail gun can be analyzed to obtain an automatic fault type identification model, so that the common fault types of the electric nail gun can be identified efficiently and accurately, the automation degree of a nail gun detection platform is improved, and the requirement of the automatic detection platform in the field of electric tools on automatic identification of the fault types is met.

Description

Electric nail gun fault type identification method based on data driving

Technical Field

The invention relates to the technical field of fault identification, in particular to a data-driven electric nail gun fault type identification method.

Background

The electric nail gun tool plays an important role in a plurality of fields such as decoration industry, building industry and the like, and is one of essential basic tools in the construction process. With the rapid development of science and technology, product detection technologies are continuously updated, a large amount of manpower and material resources are required to be invested in a traditional manual detection test, the duration and accuracy of the test are often determined by the state of test operators, an automatic control detection platform gradually replaces the traditional manual detection test due to the characteristics of high efficiency, low cost and high precision, the detection technology gradually changes from manual to automatic and intelligent, and a series of problems are exposed while technical means change.

The invention patent of patent No. CN105004497A discloses a method for sign recognition of a fault of an electric tool, which carries out fault diagnosis on noise by collecting noise when the electric tool is in fault compared with an audio signal, and has the characteristics of low cost, high stability and the like. The invention patent with the patent number of CN108921303A discloses a fault diagnosis and predictive maintenance method for an industrial motor, which detects and predicts the fault problem of the motor by combining data collected by a plurality of sensors with a BP neural network model, and has the characteristics of low cost, predictability and the like.

Disclosure of Invention

In order to overcome the defects of the technology, the fault type identification method of the electric nail gun is provided by combining the characteristics of the fault working condition of the electric nail gun under the condition of considering the data values of the current, the voltage and the like of the motor, analyzing the characteristics of different fault type data on a time domain and a frequency domain by introducing an energy consumption angle and establishing a fault identification model by combining a support vector machine algorithm, and has the characteristics of high efficiency and accuracy.

In order to achieve the purpose, the invention adopts the following scheme:

a data-driven electric nail gun fault type identification method comprises the following steps:

step 1, analyzing and finishing different fault types and reasons of the nail gun, and further estimating the influence trend of the fault on the current and the voltage of the motor theoretically;

step 2, collecting and recording data generated when the nail gun works through a current sensor and a voltage sensor, collecting current data values and voltage data values under different working conditions and calculating corresponding energy consumption values, and manually classifying different data under a normal working condition and a fault working condition, wherein the normal working condition refers to a working environment of the nail gun in a working state, and the fault working condition refers to a working environment in which the nail gun cannot continuously work, such as broken needles, staple bolts and the like;

step 3, extracting the characteristics of current, voltage and energy consumption values in a time domain under the fault working condition and the normal working condition, and classifying the fault type identification method according to the inconsistency of characteristic rules expressed by different fault type data in the time domain;

step 4, performing frequency domain conversion on the digital signals acquired in the time domain, extracting the characteristics of current, voltage and energy consumption in the frequency domain, and combining the time domain characteristics obtained in the step 3 with the frequency domain characteristics obtained in the current step to construct an initial characteristic vector;

step 5, performing dimensionality reduction on the initial feature vector by adopting a random forest algorithm;

and 6, putting the characteristic vectors processed in the step 5 into a support vector machine, diagnosing faults and optimizing parameters to finally obtain a fault identification model, thereby completing identification of common fault types of the nail gun.

Further, the step 1 is specifically as follows:

according to fault records frequently occurring in manual nail gun tests, different nail gun faults are classified according to the fault areas of the nail gun, including broken nails, staple bolts, spring pad abrasion, spring faults and firing pin faults, wherein the broken nails refer to the fact that the nails are broken in the nail gun, and the staple bolts refer to the fact that the nails are clamped in the nail gun, so that a motor can be locked, and therefore current is obviously changed; spring pad wear, spring failure, and striker failure can affect the load on the motor, causing different levels of current and voltage variation depending on the severity of the failure.

Further, the step 2 is specifically as follows:

carry out several times automatic nailing tests to certain model nail rifle through the automatic nailing platform of nail rifle, according to each moment electric current of nail rifle motor, the voltage numerical value under voltage, the current sensor record different work condition to the mark is by the moment information that the nail rifle that the platform detected broke down, the energy consumption value of motor single excitation process under the artificial calculation under the different work condition under the normal condition with the trouble time, and according to operating mode 1: normal working condition, fault 1, fault 2 \8230; working condition 2: the normal working condition, the fault 1 and the fault 2 \8230areclassified.

Further, the step 3 is specifically as follows:

extracting characteristics of current, voltage and energy consumption values in a time domain under the fault working condition and the normal working condition, including a mean value, a standard deviation, a kurtosis, a peak-to-peak value, a skewness, a maximum value and a minimum value, and judging whether at least one characteristic P contains an obvious variation relation, wherein the obvious variation relation refers to the following steps: under two different working conditions of a fault working condition and a normal working condition, the difference of the values of the characteristic P acquired at the same time is large; or under one kind of working condition, the numerical value of the characteristic P is linearly changed at continuous time, and under the other kind of working condition, the numerical value of the characteristic P is nonlinearly changed; or in one of the operating modes, the value of the characteristic P fluctuates little at successive moments, while in the other operating mode the value of the characteristic P fluctuates greatly.

For the fault type with the characteristic P, the fault type is called A-type fault, only time domain characteristics are used for fault identification, and for the fault type without the characteristic P, the fault type is called B-type fault, and the next step is needed.

Further, the step 4 is specifically as follows:

for B-type faults and normal working conditions, performing frequency domain conversion on digital signals acquired under the time domain, extracting characteristics of current, voltage and energy consumption on a frequency domain graph, including barycentric frequency, mean square frequency, root mean square frequency, frequency variance and frequency standard deviation, and constructing an initial characteristic vector set by combining the time domain characteristics obtained in the step 3 with the frequency domain characteristics obtained in the current step

Further, the step 5 is specifically as follows:

step 5.1, the initial feature vector set

Taking the feature labels of different dimensions of each feature vector and corresponding data values as a data set X, and assigning the fault conditions corresponding to the feature vectors as 0, 1,0 is normal and 1 is fault according to whether the fault conditions belong to fault working conditions; and establishing a data set y, for the data set X, different fault types aiming at the B-type faults can be subdivided into a plurality of sub data sets X1, X2 and X3.. Xn, the data under the normal working condition is divided into sub data sets X0, and each sub data set Xn comprises a plurality of sub data sets X0

The distinguishing degrees of the characteristic vectors are reflected in different corresponding working conditions, the specific number of the sub data sets is determined by the type number of the B-type faults, the corresponding data set y is correspondingly divided into a plurality of sub data sets y0, y1, y2 and y3Forming;

step 5.2, cutting the data sets X and y by using a cutting function, wherein 75% of data is used as a training set, and 25% of data is used as a testing set;

step 5.3, random forest modeling is carried out by utilizing the training set data, as the training set is divided into a plurality of sub-training sets, a plurality of sub-models are correspondingly established in the models, and each sub-model is formed by training one Xn and the yn data set corresponding to the Xn;

step 5.4, constructing an importance index of the variable, and calculating the accumulated mean value and standard deviation of the decrease of the degree of mixing of different feature labels under each tree;

step 5.5, visualizing the importance degree, quantizing the importance of the features by using the average kini coefficient obtained by the decision tree, wherein the higher the average kini coefficient is, the more important the features are;

step 5.6, aiming at the subdata sets segmented by each type B fault, selecting variables with the top three importance ranks to form new feature vectors, and integrating all the new feature vectors according to different fault types to serve as feature vector sets after dimension reduction

And (6) outputting.

Further, the step 6 is specifically as follows:

will be provided with

The characteristic vector is put into a Support Vector Machine (SVM), a radial basis kernel function is selected as a kernel function of the SVM, and kernel parameters and regularization parameters of the SVM are optimized by combining an intelligent optimization algorithm, so that accurate diagnosis of B-type faults is finally realized; and observing the diagnosis precision of the test set under the three angles of current, voltage and energy consumption, and selecting the angle corresponding to the optimal result as the recognition angle of the fault, thereby establishing a fault recognition model based on the SVM.

The beneficial effects of the invention are as follows:

1) The invention considers the actual working condition of the electric nail gun, introduces the energy consumption analysis angle for the first time, analyzes different characteristics of the electric nail gun under different faults from multiple angles by combining the current and the voltage of the motor, provides a specific and reliable automatic fault type identification method of the electric nail gun, improves the automation degree of a nail gun test platform, and reduces the labor cost of nail gun nailing test.

2) The invention classifies the identification difficulty of different fault types of the nail gun, identifies the fault type with obvious change of the characteristic value of the fault working condition and the normal working condition in the time domain, utilizes FFT to carry out spectrum analysis without obvious change, and combines the support vector machine algorithm to identify the series of fault types, thereby efficiently and accurately identifying different fault types.

Drawings

FIG. 1 is a schematic flow diagram of the process of the present invention;

FIG. 2 is a schematic flow diagram of the embodiment of FIG. 1;

FIG. 3 is a schematic diagram of a dimension reduction process of a random forest algorithm.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The detailed description of the drawings 1 and 2 is as follows:

step 1, classifying different nail gun faults according to fault records frequently occurring in manual nail gun tests and according to the fault areas of the nail guns, wherein the faults include broken nails, staple bolts, spring pad abrasion, spring faults, firing pin faults and the like, the broken nails refer to the fact that the nails are broken inside the nail guns, the staple bolts refer to the fact that the nails are clamped inside the nail guns, and theoretically, a motor can be locked, so that obvious changes of current are caused; the abrasion of a spring pad, the failure of a spring and the failure of a firing pin theoretically influence the load of a motor, and the current and the voltage are changed in different degrees according to the severity of the failure;

step 2, carrying out automatic nailing test for a certain type of nail gun through the automatic nailing platform of the nail gun, recording current and voltage values of each moment of the nail gun motor under different working conditions according to voltage and current sensors, marking the moment information of the nail gun fault detected by the platform, manually calculating the energy consumption value of the motor single excitation process under the normal condition and the fault moment under different working conditions, and according to' working condition 1: normal case, fault 1, fault 2 \8230; working condition 2: normal, fault 1, fault 2 \8230;

step 3, extracting the characteristics of the current, voltage and energy consumption values in the time domain under the fault working condition and the normal working condition, including a mean value, a standard deviation, a kurtosis, a peak-peak value, a skewness, a maximum value and a minimum value, judging whether at least one characteristic P contains an obvious variation relation, identifying the fault type (called A-type fault) with the characteristic P by using the time domain characteristic only, and entering the next step for the fault type (called B-type fault) without the characteristic P;

and 4, for the B-type faults and the normal working conditions, performing frequency domain conversion on the digital signals acquired under the time domain, extracting the characteristics of current, voltage and energy consumption on a frequency domain graph, including center-of-gravity frequency, mean square frequency, root-mean-square frequency, frequency variance and frequency standard deviation, and constructing an initial characteristic vector set by combining the time domain characteristics obtained in the step 3 with the frequency domain characteristics obtained in the current step

Step 5, vector set of initial features

The dimension of the medium feature vector is higher, so that the calculation process is complex and the calculation period is long, and therefore, a random forest algorithm is adopted to carry out on the feature vector set

Performing dimension reduction processing to obtain a feature vector set after dimension reduction

Step 6, mixing

The characteristic vectors are put into a Support Vector Machine (SVM), a radial basis kernel function is selected as a kernel function of the SVM, and kernel parameters and regularization parameters of the SVM are optimized by combining an intelligent optimization algorithm (such as a Particle Swarm Optimization (PSO) algorithm, a Genetic Algorithm (GA) algorithm and the like), so that accurate diagnosis of the B-type faults is finally realized. And observing the diagnosis precision of the test set under the three angles of current, voltage and energy consumption, and selecting the angle corresponding to the optimal result as the recognition angle of the fault, thereby establishing a fault recognition model based on the SVM.

The detailed description with respect to fig. 3 is as follows:

step 1, an initial feature vector set is obtained

Taking feature labels (such as mean value, standard deviation, center of gravity frequency and the like) of different dimensions of each feature vector and corresponding data values as a data set X, assigning the fault conditions corresponding to each feature vector to be 0 and 1 (0 is normal and 1 is fault) according to whether the fault conditions belong to the fault working conditions or not, establishing a data set y, subdividing the data set X into a plurality of subdata sets such as X1, X2 and X3 aiming at different fault types of B-type faults, dividing the data under the normal working conditions into subdata sets X0, wherein each subdata set Xn comprises a plurality of subdata sets X

The distinguishing degrees of the characteristic vectors are mainly reflected in different corresponding working conditions, the specific number of the sub data sets is determined by the type number of the B-type faults, the corresponding data set y is correspondingly divided into a plurality of sub data sets such as y0, y1, y2, y3 and the like, and each sub data set yn consists of a plurality of vectors with the values of 0 and 1;

step 2, cutting the data sets X and y by using a cutting function, wherein 75% of data is used as a training set, and 25% of data is used as a testing set;

step 3, random forest modeling is carried out by utilizing training set data, as the training set is divided into a plurality of sub-training sets, a plurality of sub-models are correspondingly established in the models, and each sub-model is formed by training one Xn and the yn data set corresponding to the Xn;

step 4, constructing importance indexes of variables, and calculating the accumulated mean value and standard deviation of the decrease of the degree of mixing of different feature labels under each tree;

step 5, visualizing the importance degree, and quantifying the importance of the features by using the average 'purity' (kini coefficient) obtained by the decision tree, wherein the higher the average 'purity', the more important the features are;

step 6, aiming at the subdata sets segmented by each type B fault, selecting variables with the top three importance ranks to form new feature vectors, and integrating all the new feature vectors according to different fault types to serve as feature vector sets after dimension reduction

And (6) outputting.

The above embodiments are only preferred embodiments of the present invention, and are not intended to limit the technical solutions of the present invention, so long as the technical solutions can be realized on the basis of the above embodiments without creative efforts, which should be considered to fall within the protection scope of the patent of the present invention.

Claims (7)

1. A fault type identification method of an electric nail gun based on data driving is characterized by comprising the following steps:

step 1, analyzing and finishing different fault types and reasons of the nail gun, and further estimating the influence trend of the fault on the current and the voltage of the motor theoretically;

step 2, collecting and recording data generated when the nail gun works through a current sensor and a voltage sensor, collecting current data values and voltage data values under different working conditions and calculating corresponding energy consumption values, and manually classifying different data under a normal working condition and a fault working condition, wherein the normal working condition refers to a working environment of the nail gun in a working state, and the fault working condition refers to a working environment of the nail gun in which the nail gun cannot work continuously;

step 3, extracting the characteristics of current, voltage and energy consumption values in the time domain under the fault working condition and the normal working condition, and classifying the fault type identification method according to the inconsistency of the characteristic rules expressed by different fault type data in the time domain;

step 4, performing frequency domain conversion on the digital signals acquired in the time domain, extracting the characteristics of current, voltage and energy consumption in the frequency domain, and combining the time domain characteristics obtained in the step 3 with the frequency domain characteristics obtained in the current step to construct an initial characteristic vector;

step 5, performing dimensionality reduction on the initial feature vector by adopting a random forest algorithm;

and 6, putting the characteristic vectors processed in the step 5 into a support vector machine, diagnosing faults and optimizing parameters to finally obtain a fault identification model, thereby completing identification of common fault types of the nail gun.

2. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 1 is as follows:

according to fault records frequently occurring in manual nail gun tests, different nail gun faults are classified according to the fault areas of the nail gun, including broken nails, staple bolts, spring pad abrasion, spring faults and firing pin faults, wherein the broken nails refer to the fact that the nails are broken in the nail gun, and the staple bolts refer to the fact that the nails are clamped in the nail gun, so that a motor can be locked, and therefore current is obviously changed; spring pad wear, spring failure and striker failure can affect the load on the motor, causing different levels of current and voltage variation depending on the severity of the failure.

3. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 2 is as follows:

carry out several times automatic nailing tests to certain model nail rifle through the automatic nailing platform of nail rifle, according to each moment electric current of nail rifle motor, the voltage numerical value under voltage, the current sensor record different work condition to the mark is by the moment information that the nail rifle that the platform detected broke down, the energy consumption value of motor single excitation process under the artificial calculation under the different work condition under the normal condition with the trouble time, and according to operating mode 1: normal working condition, fault 1, fault 2 \8230; working condition 2: the normal working condition, the fault 1 and the fault 2 \8230areclassified.

4. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 3 is as follows:

extracting the characteristics of current, voltage and energy consumption values in a time domain under the fault working condition and the normal working condition, including a mean value, a standard deviation, a kurtosis, a peak-to-peak value, a skewness, a maximum value and a minimum value, and judging whether at least one characteristic P contains an obvious variation relation, wherein the obvious variation relation refers to the following steps: under the fault working condition and the normal working condition, and under two different working conditions, the feature P acquired at the same time has difference in numerical value; or under one kind of working condition, the numerical value of the characteristic P changes linearly at continuous time, and under the other kind of working condition, the numerical value changes nonlinearly; or under two different working conditions, the numerical fluctuation of the characteristic P at continuous time is different;

for the fault type with the characteristic P, the fault type is called A-type fault, only time domain characteristics are used for fault identification, and for the fault type without the characteristic P, the fault type is called B-type fault, and the next step is needed.

5. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 4 is as follows:

for B-type faults and normal working conditions, performing frequency domain conversion on digital signals acquired under the time domain, extracting characteristics of current, voltage and energy consumption on a frequency domain graph, including barycentric frequency, mean square frequency, root mean square frequency, frequency variance and frequency standard deviation, and constructing an initial characteristic vector set by combining the time domain characteristics obtained in the step 3 with the frequency domain characteristics obtained in the current step

6. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 5 is as follows:

step 5.1, the initial feature vector set

Taking the feature labels of different dimensions of each feature vector and corresponding data values as a data set X, and assigning the fault conditions corresponding to the feature vectors as 0, 1,0 is normal and 1 is fault according to whether the fault conditions belong to fault working conditions; and establishing a data set y, for the data set X, different fault types aiming at the B-type faults can be subdivided into a plurality of sub data sets X1, X2 and X3.. Xn, the data under the normal working condition is divided into sub data sets X0, and each sub data set Xn comprises a plurality of sub data sets X0

The distinguishing degrees of the characteristic vectors are reflected in different corresponding working conditions, the specific number of the sub data sets is determined by the type number of the B-type faults, the corresponding data set y is correspondingly divided into a plurality of sub data sets y0, y1, y2 and y3.. Yn, and each sub data set yn consists of a plurality of vectors with the values of 0 and 1;

step 5.2, cutting the data sets X and y by using a cutting function, wherein 75% of data is used as a training set, and 25% of data is used as a testing set;

step 5.3, random forest modeling is carried out by utilizing the training set data, as the training set is divided into a plurality of sub-training sets, a plurality of sub-models are correspondingly established in the models, and each sub-model is formed by training one Xn and the yn data set corresponding to the Xn;

step 5.4, constructing an importance index of the variable, and calculating the accumulated mean value and standard deviation of the decrease of the promiscuity of different feature labels under each tree;

step 5.5, visualizing the importance degree, quantizing the importance of the features by using the average kini coefficient obtained by the decision tree, wherein the higher the average kini coefficient is, the more important the features are;

step 5.6, aiming at the subdata sets segmented by each type B fault, selecting variables with the top three importance ranks to form new feature vectors, and integrating all the new feature vectors according to different fault types to serve as feature vector sets after dimension reduction

And (6) outputting.

7. The method for identifying the fault type of the electric nail gun based on the data driving as claimed in claim 1, wherein the step 6 is as follows:

will be provided with

The characteristic vector is put into a Support Vector Machine (SVM), a radial basis kernel function is selected as a kernel function of the SVM, and kernel parameters and regularization parameters of the SVM are optimized by combining an intelligent optimization algorithm, so that accurate diagnosis of B-type faults is finally realized; and observing the diagnosis precision of the test set under the three angles of current, voltage and energy consumption, and selecting the angle corresponding to the optimal result as the recognition angle of the fault, thereby establishing a fault recognition model based on the SVM.

Images