CN114509971B - An automatic stopping method for hydraulic ratchet wrench based on SVM oil pressure status recognition - Google Patents

Abstract

The invention belongs to the technical field of automatic control, and discloses an automatic judging and stopping method of a hydraulic ratchet wrench based on SVM oil pressure state identification, which comprises judging whether to start judging and stopping and judging whether to stop driving the wrench operation; in the judgment of whether to start judgment stop, the real-time oil pressure state label of each pressurizing period is identified through an SVM model, and when the oil pressure state label is identified as 1, the judgment stop is started; in the judgment of stopping driving the spanner operation, the real-time oil pressure state label of each pressurizing period is continuously identified through the SVM model, if the real-time oil pressure state label is identified to be 1 and is kept until the real-time oil pressure reaches the bolt pre-tightening target oil pressure P2, the real-time oil pressure reaches P2, and then the pump station is judged to stop driving the hydraulic ratchet spanner operation. The method effectively avoids the stage of doing idle work in the pressurizing period, and further can reduce the use times of the wrench under high pressure and prolong the service life of the wrench.

Description

An automatic stopping method for hydraulic ratchet wrench based on SVM oil pressure status recognition

Technical field

The invention belongs to the field of automation control technology, relates to hydraulic ratchet wrench control technology, and specifically relates to a hydraulic ratchet wrench judgment and stop method.

Background technique

The hydraulic ratchet wrench tightens the bolts by cyclically tightening the bolts. The hydraulic pump station serves as the driving device of the hydraulic ratchet wrench and provides power for the ratchet wrench by outputting special hydraulic oil with high impact pressure. Therefore, the pump station plays a key role in the control of the wrench, including the start and stop control of the wrench's work and the control of the driving force of the wrench.

At present, hydraulic pump stations on the market often work in conjunction with torque sensors when driving wrenches for automated operations. The torque sensors are used to measure the torque on the bolts in real time, and the torque values are converted into wireless signals through wireless transmission and transmitted to the pump station control CPU. , when the torque on the bolt reaches the preset target torque, the solenoid valve of the pump station closes and the hydraulic ratchet wrench stops working.

This method of judging and stopping a hydraulic wrench based on torque detection mainly has the following shortcomings: 1. The torque sensor is mostly equipped in a torque sleeve. The torque sleeve not only plays the role of transmitting torque during the wrench tightening operation, but it can also detect the torque in real time. The torque exerted on the bolts is measured, but the sleeve is relatively bulky. When the operator needs to tighten multiple bolts at different positions, he also needs to continue to change the position of the sleeve multiple times, which is cumbersome and greatly affects work efficiency. 2. The torque sleeve transmits the torque data to the pump station control CPU wirelessly. This process is easily affected by interference from the external environment, causing transmission delays. It is also limited by the distance between the pump station and the sleeve, which will affect the pump. Station-to-wrench automation accuracy.

Compared with the above-mentioned wrench stopping method based on torque detection, the hydraulic wrench stopping method based on oil pressure waveform similarity analysis disclosed in patent document No. CN112388552A is more progressive. This method is based on the analysis of oil pressure signals to achieve Determine whether to stop wrench work. During the bolt tightening process, there are three main stages of movement of the wrench piston in a certain pumping station pressure cycle. The first stage is when the piston remains stationary at the starting point of the forward stroke (this process overcomes the torque required for the bolt to start rotating. , the pump station has output but does no work, the bolt torque remains unchanged and is greater than the output torque of the wrench), the second stage is when the piston starts the positive stroke until the piston moves to the end of the stroke (in this process, the wrench drive shaft tightens the bolt under the push of the piston , the pump station does useful work, the output torque of the wrench is in the opposite direction to the bolt torque), the third stage is when the piston remains stationary at the end of the stroke (the next cycle will be started after the oil pressure of the pump station is increased to the target value, and the pump station has output But no work is done, and the bolt torque remains unchanged). This method is to work in the mode of controlling the constant output oil pressure of the pump station. It cannot immediately start the next oil pressure cycle when it recognizes that the piston has moved to the end of the stroke, that is, the third piston movement state. This stage is not avoided, so there is still a lot of wasted work during the working process of the pumping station. This is the main reason for the low working efficiency of the pumping station. It can also be found that most of the wasted work done by the pumping station is the high-pressure output stage, which causes relatively large losses to the wrench. Seriously detrimental to extending the service life of the wrench.

The working process of a hydraulic ratchet wrench is a process of repeated pressurization and oil return by the pump station. Therefore, when the hydraulic pump station is used to drive the hydraulic ratchet wrench, the waveform of its output oil pressure is a pulse waveform, and each pressurization cycle forms a pulse.

The entire process of bolt strengthening can be divided into two stages. The first stage is from starting to rotate the nut to the nut starting to enter the tightening state. This stage is called the no-load stage; the second stage is from starting to enter the tightening state to reaching Target pre-tightening force, this stage is called the tightening stage.

Contents of the invention

The object of the present invention is to provide a method to determine whether the pump station should stop driving by identifying the output oil pressure state of the pump station in view of the above-mentioned shortcomings of the existing method of determining whether the wrench should stop by monitoring the torque of the wrench and other methods of determining whether the wrench should stop. Wrench control method.

In order to achieve the above object, the present invention adopts the following technical solution: an automatic stopping method of a hydraulic ratchet wrench based on SVM (Support Vector Machine) oil pressure state recognition. This method is used to determine when the hydraulic pump station should stop operating the hydraulic ratchet wrench. The drive includes the judgment of whether to start the stop and whether to stop the wrench operation; for the no-load stage of the bolt, the no-load working mode of the hydraulic pump station is set. In the no-load working mode of the hydraulic pump station, the maximum output oil pressure is set The no-load driving oil pressure of the wrench is P1. For the tightening stage of the bolt, the tightening working mode of the hydraulic pump station is set. In the tightening working mode of the hydraulic pump station, the maximum output oil pressure is set to the bolt pre-tightening target oil pressure. P2,

Determination of whether to enable judgment and stop: During the no-load stage of the bolt, the control pump station works in the no-load working mode, and the real-time oil pressure status label of each pressurization cycle is identified through the SVM model. When the oil pressure status is identified When the label is 1, it means that the bolt enters the tightening stage, then the working mode of the pump station is changed to the tightening working mode, and the judgment of whether to stop driving the wrench operation is enabled; the judgment of whether to stop driving the wrench operation: after the bolt enters the tightening stage , continue to identify the real-time oil pressure status label of each pressurization cycle through the SVM model. If the oil pressure status label is identified as 1 and remains until the real-time oil pressure reaches P2, then the pumping station will stop when the real-time oil pressure reaches P2. Drive hydraulic ratchet wrench operation. Specifically, it includes the following steps:

Step S0: Train the SVM model;

Step S1: Start the hydraulic pump station to drive the hydraulic ratchet wrench operation. The pump station repeatedly performs the pressurization process starting from 0, and at the same time monitors the output oil pressure waveform of the hydraulic pump station in real time;

Step S2: Starting from the first pressurization, collect n consecutive oil pressure data in polling (the number of single collected data can be adjusted, preferably 10), and calculate their characteristic values, and input the real-time calculated characteristic values into Classify the trained SVM model to obtain the real-time oil pressure status label;

Step S3: If the oil pressure status label is always -1 in a pressurization cycle, start the next pressurization cycle when the oil pressure reaches the wrench no-load driving oil pressure P1; if the oil pressure reaches P1 before When it is recognized that the oil pressure status label changes from -1 to 1, step S4 begins, and the bolt enters the tightening stage;

Step S4: Continue to pressurize based on the last pressurization cycle of step S3. If the oil pressure status label changes from 1 to -1 before the oil pressure reaches P2, start the next new pressurization cycle. If the oil pressure is The oil pressure status label remains at 1 before reaching P2, and the pump station stops working when the oil pressure reaches P2; in the newly opened pressurization cycle, if the oil pressure status label changes from -1 to 1 before the oil pressure reaches P2, then to -1, then start the next pressurization cycle. If the oil pressure status label changes from -1 to 1 before the oil pressure reaches P2 and then remains 1, the pump station will stop working when the oil pressure reaches P2.

Further, in step S1, a sensor is used to collect the output oil pressure data of the pumping station in real time, and the data is transmitted to the pumping station CPU.

Further, step S2 includes:

Step S2-1: Starting from the first pressurization, the pump station CPU samples the real-time collected output oil pressure data of the pump station according to the set sampling frequency t1, and the sampled array (the array adopts polling assignment method to ensure The latest 10 oil pressure data points collected) are stored in the array;

Step S2-2: For the 10 oil pressure data points collected in real time, calculate their mean (μ), standard deviation (σ), skewness (SK), coefficient of variation (cv) and average slope (kp) respectively;

Step S2-3: Import the latest μ, σ, SK, cv, kp into the trained SVM model in real time for classification, and obtain the real-time oil pressure status label.

Furthermore, the mean calculation formula is as follows (1):

(1)

where is the oil pressure data point, is the mean value of the oil pressure data, and is the number of data points;

The standard deviation calculation formula is as follows (2):

(2)

where is the standard deviation of oil pressure data;

The skewness calculation formula is as follows (3):

(3)

Among them, SK is the oil pressure data skewness;

The calculation formula of the coefficient of variation is as follows: (4):

(4)

where cv is the coefficient of variation;

Average slope calculation formula (5):

(5)

where kp is the average slope.

In the present invention, although five eigenvalues of mean μ, standard deviation σ, skewness SK, coefficient of variation cv, and average slope kp are selected to represent the attributes of the sample, in fact, the remaining possible data eigenvalues are used in this method. It is also within the scope of patent protection, and the application of the feature value selection method in this method is also within the scope of patent protection.

The hydraulic ratchet wrench judgment method of the present invention includes real-time collection of the output oil pressure data of the pump station and real-time feature value extraction, and indirectly completes the identification of the oil pressure status label through SVM (support vector machine) to determine whether the bolt reaches the Judgment of pre-tightening status enables automatic judgment and stop control of the hydraulic ratchet wrench. In the whole method, when starting work, because the bolt is in a very loose state, the wrench first enters the no-load working stage. The maximum output oil pressure of the pump station in this stage is the wrench no-load driving oil pressure P1. During the 0~P1 cyclic pressurization cycle of the pump station , if the identified oil pressure status label is always -1, the next pressurization cycle will be started when the real-time oil pressure reaches the wrench no-load driving oil pressure P1 (the piston must complete the entire positive stroke); if the oil pressure status label is identified by -1 changes to 1, which means that the wrench enters the tightening stage from the no-load stage, thus starting the judgment of whether to stop the wrench operation. During the bolt tightening stage, when it is recognized that the oil pressure status label is always 1 and the real-time oil pressure reaches the bolt pretightening target oil pressure P2, the wrench stops working. When the oil pressure status label changes from -1 to 1 and then to -1 , indicating that the piston has completed the forward stroke and has started to remain stationary and the pump station is doing useless work. At this time, the next pressurization cycle will be started in time. Through the wrench judgment and stop method based on SVM (Support Vector Machine) oil pressure state recognition, during the no-load stage, the pump station uses the wrench no-load driving oil pressure P1 as the target value to drive the wrench so that the bolt quickly enters the tightening stage. The SVM (support vector machine) model recognizes and detects in real time whether the oil pressure status label changes from 1 to -1 (the wrench piston moves to the end of the stroke) and adaptively starts the next pressurization cycle, minus the pump station driver There is a lot of wasted work in the process of tightening bolts with hydraulic tools. Compared with the constant target oil pressure working mode of the oil pressure waveform similarity judgment and stopping algorithm, the working efficiency of the pump station is greatly improved; by monitoring the process when the oil pressure status label is always 1 Whether the real-time oil pressure reaches the target oil pressure to achieve accurate stopping of the wrench. Since this algorithm effectively avoids the useless high-pressure section in the oil pressure waveform of the pressurization cycle, it can reduce the number of times the wrench is used under high pressure and extend the service life of the wrench; at the same time, this method only needs to identify the bolt pre-tightening state by The pumping station can output oil pressure data for analysis, which avoids the requirement that a torque sensor be equipped during automated wrench operations, avoids the impact of interference on the pumping station’s accurate stopping due to interference in the wireless transmission process of torque data, and reduces the need for wrench tightening operations. workload of operators.

Description of the drawings

Figure 1 shows the output oil pressure curve of the pump station and the real-time torque curve of the bolt.

Figure 2 is the wrench piston displacement curve.

Figure 3 shows the specific parameters of SVMmodel.

Figure 4 is the actual test pump station working oil pressure curve.

Figure 5 is the wrench piston displacement curve.

Figure 6 is the sampling sample identification result.

Detailed ways

Referring to Figure 1-2, in the first two pressurization cycles, the bolt is in a loose state and has not entered the tightening stage. The torque is always 0. The wrench is equivalent to no-load driving. During this process, the output oil pressure waveform of the pump station is relatively stable. No significant changes have occurred, and the pumping station does no work;

The bolts enter the tightening stage at the beginning of the third cycle and continue until the bolts reach the pre-tightening torque in the fifth pressurization cycle and the tightening is completed. When pressurization begins in the third and fourth pressurization cycles, the oil pressure rises quickly and continues until At t1 and t3, the output torque of the wrench corresponding to the output oil pressure of the pump station in this process is always smaller than the torque when the bolt starts to rotate. The wrench piston remains stationary at the starting point of the stroke, and the pump station has output but does no work; the third and fourth additions In the pressure cycle, the output oil pressure of the pump station rises significantly after moments t1 and t3, and the slope decreases significantly until moments t2 and t4 respectively. In this process, the wrench piston begins its positive stroke and moves to the end of the stroke to complete one tightening of the bolt. operation, the pump station does useful work; in the third and fourth pressurization cycles, after t2 and t4, the oil pressure of the pump station quickly rises to the target oil pressure. During this process, the wrench piston remains stationary at the end of the stroke, and the pump station has output and does no work. You can It was found that the output hydraulic pressure waveform of the pump station in the third and fourth pressurization cycles showed three distinct segments, among which the hydraulic pump station in the second segment performed useful work;

The fifth pressurization cycle is tightened to complete the cycle. The piston movement state is the same as mentioned above. It remains stationary at the starting point of the stroke and the oil pressure at the pump station rises rapidly. Then the oil pressure at the pump station rises slowly. The piston starts the positive stroke and outputs oil pressure at a certain position in the stroke. When the bolt pre-tightening target oil pressure is reached, the output torque value of the corresponding wrench is equal to the bolt torque value, and the bolt is tightened. It can be found that the output oil pressure curve of the pump station during the last pressurization cycle of the pump station in the tightening stage has two sections with obvious differences, of which the second section does useful work.

In the present invention, the oil pressure characteristics of different oil pressure sections are defined as one oil pressure state, and the oil pressure state label corresponding to the second segment in the oil pressure waveform of a pressurization cycle during the bolt tightening stage is expressed as 1, and the The oil pressure status corresponding to the first and third sections, and the oil pressure status label corresponding to the oil pressure waveform in the no-load stage of the bolt is expressed as -1. The SVM (Support Vector Machine) model is a supervised learning model in the field of machine learning. It is usually used for pattern recognition, classification and regression analysis. When applied to the present invention, the SVM model can be used to analyze a small piece of oil pressure data. Multiple feature values are used to identify the corresponding oil pressure status label. Here, the SVM model is mainly used for two-class recognition, and the stopping algorithm of the present invention is designed based on this.

The judgment and stopping algorithm is mainly: establish an SVM model to identify the oil pressure status label in each pumping station pressurization cycle. When the oil pressure status label is always 1 and the real-time oil pressure of the pumping station reaches the bolt pretightening target oil pressure, the bolt is judged. The pre-tightening state has been reached, and the stopping algorithm is designed according to this relationship.

In the stage of whether to start the judgment and stop judgment, when the SVM does not recognize that the oil pressure status label is 1 (that is, the oil pressure status label is always -1) and the oil pressure reaches the wrench no-load driving oil pressure P1, the next P1 pressurization cycle is started; Whether to stop driving the wrench during a certain pressurization cycle when the SVM (Support Vector Machine) model recognizes that the oil pressure label changes from 1 to -1 to start the next pressurization cycle; if the SVM (Support Vector Machine) model recognizes that the oil pressure label changes from 1 to -1, The oil pressure label is always 1 and remains until the real-time oil pressure is equal to the bolt pre-tightening target oil pressure P2 to stop the pump station from driving the wrench.

Implementing the stopping algorithm of the present invention specifically includes the following steps:

Step S0: SVM model training.

Step S0-1: Conduct multiple experiments to obtain the data set (oil pressure characteristic values and corresponding labels) used for training and testing.

Step S0-2: In the matlab software, the training data set obtained from multiple experiments is used to train the SVM model, and the optimal c (penalty coefficient) and g (RBF function as the kernel function) are found through the parameter optimization algorithm. A parameter that implicitly determines the distribution of data after mapping to a new feature space).

Step S0-3: Use the best c, g and training data to train to obtain the SVM model parameters: model structure (the structure contains some parameters as shown in Figure 3).

Step S1: Start the hydraulic pump station to drive the hydraulic ratchet wrench operation. The pump station repeatedly executes the pressurization process starting from 0. The sensor collects the output oil pressure data of the pump station in real time and transmits the data to the pump station control CPU.

Step S2: Real-time oil pressure status tag monitoring.

Step S2-1: Set the sampling frequency to t1, the pump station CPU samples the real-time oil pressure data, and saves 10 consecutive data as an array.

Step S2-2: Calculate the mean, standard deviation, skewness, coefficient of variation and average slope of the data in the array.

Step S2-3: Import the calculated average value, standard deviation, skewness, coefficient of variation and average slope into the SVM model for classification, and obtain the real-time oil pressure status label;

Step S3: Determine whether to enable judgment and stop

If the oil pressure status label is always -1 in a pressurization cycle, the next pressurization cycle will be started when the oil pressure reaches P1. If it is recognized that the oil pressure status changes from -1 before the oil pressure reaches P1 in a pressurization cycle. If it is 1, then continue to pressurize and start the judgment of whether to stop driving the wrench operation;

Step S4: Determine whether to stop driving the wrench operation

During the pressurization cycle, if the oil pressure status label changes from 1 to -1 before the oil pressure reaches P2, the next pressurization cycle will be started. If the oil pressure reaches the bolt pretightening target oil pressure P2 during a pressurization cycle, the classification result will be If it remains at 1, the pump station will stop working when the oil pressure reaches P2.

In step S0-1, take the model 2XLCT-5 hydraulic wrench, three-stage flow wrench pump, and M30 bolt as an example. Conduct multiple experiments under multiple target oil pressures (each experiment is cyclically pressurized in the working mode of a constant target oil pressure in the pump station, and work until the bolts cannot be tightened). Take all the data from each complete tightening operation. For oil pressure data, the oil pressure is segmented according to the data points where the oil pressure slope suddenly changes during the pressurization cycle (oil pressure inflection point), and multiple characteristic values are calculated for the oil pressure data of each segment and based on the corresponding characteristics of the oil pressure segment. The stage at which the wrench piston is located assigns a label to this set of feature data. The training data set is obtained as shown in the following table:

Among them, the feature values and labels are used as the input of the SVM model, and a set of feature values and corresponding labels is a sample training data. The number of samples here is 36. (It is best to train the models separately for different working conditions before using this algorithm to implement wrench judgment and stop control. Different working conditions have many factors that affect the difficulty of the bolt tightening process).

In step S0-2, import the data set into the matlab software, shuffle the order of the data set, take the first 30 sets of data as the training set, and the last 6 sets of data as the test set, and use the training set data to test the c and g parameters. Search for optimization to train the model, and use the test set to conduct experimental verification of the trained model. Through the cross-validation optimization algorithm, find the optimal c and g parameters, that is, given any value of c and g between -, divide the training data set into 5 equal parts, one part is used as the test set, and four parts It is used as a training set to obtain a classification accuracy, and c and g are used as the optimal parameters when the classification accuracy is the highest. c is 4 and g is 0.4353.

In step S0-3, the existing c, g parameters and training set are used to train the SVM model, and the SVM model model structure is obtained. The specific parameters are shown in Figure 3, and the final decision tree function is as shown in Equation (6 ):

(6)

Where b is -model.rho; n represents the number of support vectors, which is model.totalSV; for each i, wi is the model.sv_coef(i) support vector coefficient, xi is the model.SVs(i,:) support vector, x is the label sample to be predicted, and gamma is the -g parameter. Finally, the SVM model is used to predict the test set data. By comparing the prediction results with the known labels, the accuracy rate is 100%, which can verify that the model is accurate.

In step S1, a high-voltage sensor is installed at the oil outlet of the pump station, and is powered by the pump station control box. After the pump station is powered on, the sensor is on and measures the oil pressure at the oil outlet of the pump station in real time.

In step S2, the CPU sampling frequency is set to 6ms. During each pressurization process of the pump station, the data transmitted in real time by the sensor is sampled at 6ms intervals. In order to ensure the accuracy of the data, an average filtering algorithm is used, and the average value of every 3 points is taken and saved in the array. An array with a capacity of 10 is created to store the latest collected oil pressure data points. Extract the eigenvalues of the 10-capacity array obtained in real time, and use the eigenvalues as the input of the model for real-time classification to obtain the oil pressure status label.

In steps S3 and S4, the oil pressure status label is monitored in real time. If the real-time oil pressure reaches the wrench no-load driving oil pressure P1, the oil pressure status label is always -1. When the oil pressure reaches the wrench no-load driving oil pressure P1, the oil pressure status label is turned on. One P1 pressurization cycle; if the oil pressure status label changes to 1 before the real-time oil pressure reaches P1, it means the bolt starts to be tightened, and the judgment of whether to stop the wrench operation is enabled. If the oil pressure reaches the bolt pre-tightening target oil pressure P2 before the oil pressure If the pressure status label is always 1, then when the oil pressure reaches P2, the bolt tightening can be completed and the wrench operation can be stopped. If it changes from 1 to -1, the next pressurization cycle can be started. The no-load driving oil pressure P1 of the wrench is used to quickly complete the idling process of the wrench. It is a very small pressure and the oil pressure is conditioned to overcome the damping of the wrench itself; the bolt pre-tightening target oil pressure P2 is required for the bolt to reach the target torque. The pump station outputs oil pressure.

The following is an application example of the stop judgment method of the present invention. In this example, the ratchet wrench is 2XLCT-50, the bolt model is M30, the bolt pre-tightening target oil pressure P2 output by the hydraulic pump is set to 18MPa, and the CPU is set The sampling frequency is 6ms, the no-load driving oil pressure P1 of the wrench is 5MPa, and the hydraulic pump is started. The real-time oil pressure curve is shown in Figure 4. The oil pressure status label identification based on the SVM model for the entire tightening process is shown at the top of the figure. In the first and second pressurization cycles, the oil pressure status label was -1 before the oil pressure reached 5MPa; in the third pressurization cycle, the bolts started to be tightened and the oil pressure status label was detected to be 1 and then changed to -1 (not reached). Preset 18MPa) to start the next pressurization cycle; the fourth pressurization cycle was also completed. In the fifth pressurization cycle, it was detected that the oil pressure state was first -1 and then 1 and maintained at 18MPa, and the algorithm successfully judged a stop. In order to better verify the SVM model, eight data samples (each with 10 data points) are extracted from a to b, c to d, e to f, g to h, i to j, k to l, m to n, and o to p. Prediction is made and the classification results are obtained as shown in Figure 6. The accuracy is 100% and the model is reliable.

Claims (5)

1. The automatic judging and stopping method for the hydraulic ratchet wrench based on SVM oil pressure state identification is used for judging when a hydraulic pump station should stop driving the hydraulic ratchet wrench, and comprises judging whether to start judging and stopping and judging whether to stop driving the wrench operation; setting an idle working mode of a hydraulic pump station aiming at an idle stage of a bolt, setting maximum output oil pressure as spanner idle driving oil pressure P1 in the idle working mode of the hydraulic pump station, setting a fastening working mode of the hydraulic pump station aiming at a fastening stage of the bolt, and setting maximum output oil pressure as bolt pre-tightening target oil pressure P2 in the fastening working mode of the hydraulic pump station;

judging whether to start judging and stopping: in the idle stage of the bolt, controlling the pump station to work in an idle working mode, identifying a real-time oil pressure state label of each pressurizing period through an SVM model, and when the oil pressure state label is identified as 1, indicating that the bolt enters a fastening stage, converting the working mode of the pump station into a fastening working mode and starting the judgment of whether to stop driving the operation of the wrench;

judging whether to stop driving the spanner operation: after the bolts enter the fastening stage, identifying real-time oil pressure state labels of each pressurizing period through an SVM model, and if the real-time oil pressure state labels are identified to be 1 and are kept until the real-time oil pressure reaches the bolt pre-tightening target oil pressure P2, stopping driving the hydraulic ratchet wrench by a pump station when the real-time oil pressure reaches P2;

the method specifically comprises the following steps:

step S0: training an SVM model;

step S1: starting a hydraulic pump station to drive a hydraulic ratchet wrench to operate, repeatedly executing a pressurizing process starting from 0, and simultaneously monitoring the waveform of the output oil pressure of the hydraulic pump station in real time;

step S2: from the first pressurization, polling to collect n continuous oil pressure data, calculating characteristic values of the n continuous oil pressure data, inputting the characteristic values calculated in real time into a trained SVM model for classification, and obtaining a real-time oil pressure state label;

step S3: if the oil pressure state label is always-1 in one pressurizing period, starting the next pressurizing period when the oil pressure reaches P1, and if the oil pressure state label is recognized to be changed from-1 to 1 before the oil pressure reaches P1 in one pressurizing period, starting to execute step S4, and enabling the bolt to enter a fastening stage;

step S4: continuing to pressurize on the basis of the last pressurizing period in the step S3, if the oil pressure state label is changed from 1 to-1 before the oil pressure reaches P2, starting the next new pressurizing period, and if the oil pressure state label is kept to be 1 before the oil pressure reaches P2, stopping the pump station when the oil pressure reaches P2;

in the newly opened pressurizing period, if the oil pressure state label is changed from-1 to-1 before the oil pressure reaches P2, starting the next pressurizing period, and if the oil pressure state label is kept to be 1 after the oil pressure state label is changed from-1 to 1 before the oil pressure reaches P2, stopping the pump station when the oil pressure reaches P2;

the step S2 includes:

step S2-1: starting from the first pressurization, the CPU of the pump station sets the sampling frequency t ₁ Sampling the oil pressure data output by the pump station acquired in real time, and storing an array obtained by sampling;

step S2-2: for 10 oil pressure data points collected in real time, respectively calculating the mean value mu, standard deviation sigma, skewness SK, variation coefficient cv and average slope k _p ；

Step S2-3: real-time to update mu, sigma, SK, cv, k _p Leading the real-time oil pressure state label into a trained SVM model for classification;

wherein, the oil pressure state label is 1 and indicates that the hydraulic power unit does useful work, and the oil pressure state label is-1 and indicates that the hydraulic power unit does not do work.

2. The automatic stopping method of the hydraulic ratchet wrench according to claim 1, wherein in step S1, the hydraulic sensor is used to collect the pump station output hydraulic data in real time and transmit the data to the pump station CPU.

3. The automatic stopping method of the hydraulic ratchet wrench according to claim 1, wherein the step S0 includes the steps of:

s0-1, obtaining a data set used for training test through multiple experiments;

step S0-2: in matlab software, a training data set is obtained through multiple tests and is used for training an SVM model, and an optimal penalty coefficient c and an RBF parameter g are found through a parameter optimizing algorithm;

step S0-3: training with optimal c and g and training data sets to obtain SVM model parameters: model structure.

4. The automatic judging and stopping method of the hydraulic ratchet wrench according to claim 3, wherein in the step S0-2, the sequence of the data set imported into matlab software is disordered, the first 30 groups of data are taken as training sets, the last 6 groups of data are taken as test sets, the training set data are utilized to optimize the c and g parameters to train the model, and the test set is utilized to perform experimental verification on the trained model; searching the optimal c, g parameter, i.e. giving c, g in 2 by a cross-validation optimizing algorithm ^-10 -2 ¹⁰ And dividing the training data set into 5 equal parts, wherein one part is used as a test set, and four parts are used as training sets to obtain a classification accuracy, and c and g are taken as optimal parameters under the condition of highest classification accuracy.

5. The method of automatic stopping a hydraulic ratchet wrench of claim 3, wherein the decision tree function of the SVM model is as follows:

wherein b is-model rho, n represents the number of support vectors, namely model total SV; for each i, w _i Support vector coefficients, x for model. Sv_coef (i) _i SVs (i, i) support vector, x is the label sample to be predicted, and gamma is the-g parameter.