Intelligent coal preparation equipment servo motor control system and method based on multi-model adaptation

Abstract

The application relates to the technical field of servo motor control, and discloses an intelligent coal preparation equipment servo motor control system and method based on multi-model self-adaption, which comprises the following steps: acquiring rotation speed information at T moment, and acquiring an abnormality evaluation coefficient at T+N moment; determining whether the encoder is in an abnormal state at T+N moment according to the abnormality evaluation coefficient; recording the measured pulse number at T+N moment, performing numerical correction on the measured pulse number according to pulse measurement characteristic data and a pre-configured pulse measurement correction model, determining the actual output rotation speed of the servo motor based on the numerically corrected measured pulse number; determining whether a rotation speed abnormality event exists, inputting a first target input rotation speed and the actual output rotation speed into a pre-configured rotation speed correction model according to the rotation speed abnormality event, and obtaining a second target input rotation speed; the application is beneficial to improving the rotation speed control precision of the servo motor and shortening the response time length of the servo motor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of servo motor control, more particularly, the present application relates to an intelligent coal preparation equipment servo motor control system and method based on multi-model adaptation. BACKGROUND

[0002] With the continuous development of the coal preparation industry, the operating efficiency and separation quality of the coal preparation production system are gradually improving. The servo motor in the coal preparation equipment is a key component of the core driving control link of the coal preparation production system, and its control precision and dynamic response performance directly affect the operation stability and separation accuracy of the coal preparation equipment, and thus affect the core production indicators such as clean coal qualification rate and medium consumption control. The traditional servo motor adopts a closed-loop control system, which monitors the motion state of the output shaft in real time, and transmits the feedback signal to the control system, thereby achieving precise motion control of the output shaft position, speed or torque, to meet the operation requirements of the coal preparation equipment such as vibration frequency control of the vibrating screen, drum speed regulation of the dense medium separator, and conveying speed stability of the scraper conveyor.

[0003] However, when facing different load conditions, working environments and working conditions in the coal preparation scene, such as sudden changes in servo motor load caused by coal quality fluctuations, long-time operation caused by continuous separation operation of the equipment, local overheating caused by heavy load operation of the motor, and continuous impact of coal dust and humid environment on the equipment, the traditional servo motor control system faces the problems of unstable performance and long response time, which easily causes the accuracy of the servo motor speed control to decrease, and thus may cause production problems such as reduced screening efficiency of the vibrating screen, excessive fluctuation of the dense medium separation density, and coal flow blockage of the scraper conveyor. Therefore, a more flexible and adaptive servo motor control system and method is needed, which can automatically adjust parameters under different working conditions of coal preparation to maintain the efficient performance of the motor operation and ensure the continuous stability of the coal preparation production.

[0004] At present, the existing servo motor control system or method of the coal preparation equipment mainly focuses on the control optimization design of the speed, torque or position of the motor, such as the speed regulation scheme of the servo motor of the coal preparation vibrating screen, the torque control method of the drive motor of the dense medium separator, and the position control technology of the servo motor of the scraper conveyor. Although the above method can realize the speed adjustment of the servo motor, research and actual application of the above method and the existing technology show that the above method and the existing technology at least have the following defects:

[0005] (1) Lack of abnormality detection mechanism, which cannot predict the encoder abnormality in the servo motor in advance, and thus easily causes the servo motor speed control error to gradually increase, and easily causes the system to have too much data processing and feedback judgment, and thus easily increases the response time of the servo motor;

[0006] (2) cannot be measured on the basis of abnormal detection encoder data correction, further, cannot be measured according to the encoder data correction result, the speed of the servo motor is adaptively adjusted;Easy to affect the speed control accuracy of servo motor in complex scene, and then it is difficult to guarantee the control stability of servo motor. SUMMARY

[0007] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide an intelligent coal preparation equipment servo motor control system and method based on multi-model adaptation.

[0008] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme:

[0009] The intelligent coal preparation equipment servo motor control method based on multi-model adaptation, the method comprises:

[0010] Step 1: obtain the speed information of the servo motor at the current T time, and obtain the abnormal evaluation coefficient of the encoder at T+N time after N time according to the preconfigured abnormal detection model, the speed information includes the measured output speed and the first target input speed, T and N are positive integers greater than zero;

[0011] Step 2: determine whether the encoder is in an abnormal state at T+N time according to the abnormal evaluation coefficient, if not, continue to control the servo motor at the first target input speed, and let T=T+N+M, and return to step 1;If it is in an abnormal state, the pulse measurement characteristic data of the encoder at T+N time is obtained;M is a preset time span;

[0012] Step 3: record the measured pulse number at T+N time, and correct the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, and determine the actual output speed of the servo motor based on the corrected measured pulse number;

[0013] Step 4: determine whether there is a speed abnormal event, input the first target input speed and the actual output speed to the preconfigured speed correction model according to the speed abnormal event, obtain the second target input speed, and control the speed of the servo motor according to the second target input speed.

[0014] Further, the abnormal evaluation coefficient of the encoder at T+N time is obtained according to the preconfigured abnormal detection model, comprising:

[0015] Obtain the coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data and voltage difference data of the power supply of the encoder from T-M time to T time, M is a positive integer greater than zero;

[0016] The coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data of the encoder and voltage difference data of the encoder power supply from the T-M moment to the T moment are input into the preconfigured abnormality detection model to obtain an abnormality evaluation coefficient of the encoder at the T+N moment.

[0017] Further, the generation method of the preconfigured abnormality detection model is:

[0018] Obtain historical abnormality evaluation data, the historical abnormality evaluation data including coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data of the encoder, voltage difference data of the encoder power supply and corresponding abnormality evaluation coefficients in multiple time spans;

[0019] Preprocess the historical abnormality evaluation data to obtain a first feature data set, a second feature data set and a third feature data set;

[0020] Respectively extract first training data, second training data and third training data in the first feature data set, the second feature data set and the third feature data set;

[0021] Build a first base learner, a second base learner and a third base learner, and train the first training data, the second training data and the third training data by using the first base learner, the second base learner and the third base learner respectively to obtain a first regression model, a second regression model and a third regression model;

[0022] Input the historical abnormality evaluation data into the first regression model, the second regression model and the third regression model respectively for prediction to obtain first prediction data, second prediction data and third prediction data;

[0023] Build a meta-learner, take the first prediction data, the second prediction data and the third prediction data as abnormality evaluation data, and divide the abnormality evaluation data into an abnormality evaluation training set and an abnormality evaluation test set;

[0024] Input the abnormality evaluation training set into the meta-learner, and train according to an ensemble learning strategy to obtain an ensemble learning model, verify the ensemble learning model by using the abnormality evaluation test set, and output an ensemble learning model meeting a preset prediction error as the preconfigured abnormality detection model.

[0025] Further, the abnormality evaluation coefficient is obtained by:

[0026] According to the current speed of the servo motor, the output waveform of the encoder in a set time span is obtained by an oscilloscope, and according to the current speed, the standard output waveform of the encoder in the set time span is obtained;

[0027] acquire an average amplitude value of the standard output waveform in a set time span, and take the average amplitude value as an amplitude reference value;

[0028] extract an amplitude value of the output waveform at each time point in the set time span, perform a formula calculation on the amplitude value and the amplitude reference value, and acquire an abnormality evaluation coefficient of the encoder.

[0029] Further, determining whether the encoder at the time T+N is in an abnormal state comprises:

[0030] comparing the abnormality evaluation coefficient with a preset abnormality evaluation coefficient threshold value;

[0031] if the abnormality evaluation coefficient is greater than or equal to the preset abnormality evaluation coefficient threshold value, it is determined that the encoder at the time T+N is in an abnormal state;

[0032] if the abnormality evaluation coefficient is less than the preset abnormality evaluation coefficient threshold value, it is determined that the encoder at the time T+N is not in an abnormal state.

[0033] Further, according to the pulse measurement characteristic data and a preconfigured pulse measurement correction model, the measured pulse number is numerically corrected, comprising:

[0034] inputting the pulse measurement characteristic data into the preconfigured pulse measurement correction model to obtain a measured pulse correction number;

[0035] performing an accumulation calculation on the measured pulse correction number and the measured pulse number to obtain a numerically corrected measured pulse number.

[0036] Further, the preconfigured pulse measurement correction model is generated by training according to encoder test data; the encoder test data at least includes a relationship between the measured pulse correction number and each Celsius temperature and a relationship between the measured pulse correction number and each decibel electromagnetic noise.

[0037] Further, the acquisition method of the encoder test data comprises:

[0038] S311: placing a test encoder in a temperature change test environment and placing a standard encoder in a set standard constant temperature test environment;

[0039] S312: in the temperature change test environment, acquiring a first test pulse number of the servo motor per rotation measured by the test encoder at the i-th Celsius degree, i being a positive integer greater than zero;

[0040] S313: in the set standard constant temperature test environment, acquiring a first standard pulse data of the servo motor per rotation measured by the standard encoder;

[0041] S314: taking the difference between the first test pulse number and the first standard pulse data as a first pulse data difference, comparing the first pulse data difference with a preset first pulse data difference interval, if the first pulse data difference belongs to the preset first pulse data difference interval, setting i = i + 1 and returning to step S312; if the first pulse data difference does not belong to the preset first pulse data difference interval, taking the first pulse data difference as a measurement pulse correction number, and binding the i Celsius degree with the measurement pulse correction number to obtain the relationship between the measurement pulse correction number and the i Celsius degree;

[0042] S315: repeating steps S312-S314 until i equals a set temperature Q, and ending the cycle to obtain the relationship between the measurement pulse correction number and each Celsius temperature, Q being a positive integer greater than zero.

[0043] Further, the method for obtaining the encoder test data further comprises:

[0044] S321: placing the test encoder in an electromagnetic change test environment, and placing the standard encoder in a set standard invariable electromagnetic noise test environment;

[0045] S322: obtaining a second test pulse number of the test encoder measured under a j decibel electromagnetic noise in each rotation of the servo motor in the electromagnetic change test environment, j being a positive integer greater than zero;

[0046] S323: obtaining a second standard pulse data of the standard encoder measured in each rotation of the servo motor in the set standard invariable electromagnetic noise test environment;

[0047] S324: taking the difference between the second test pulse number and the second standard pulse data as a second pulse data difference, comparing the second pulse data difference with a preset second pulse data difference interval, if the second pulse data difference belongs to the preset second pulse data difference interval, setting j = j + 1 and returning to step S322; if the second pulse data difference does not belong to the preset second pulse data difference interval, taking the second pulse data difference as a measurement pulse correction number, and binding the j decibel electromagnetic noise with the measurement pulse correction number to obtain the relationship between the measurement pulse correction number and the j decibel electromagnetic noise;

[0048] S325: repeating steps S322-S324 until j equals a set temperature E, and ending the cycle to obtain the relationship between the measurement pulse correction number and each decibel electromagnetic noise, E being a positive integer greater than zero.

[0049] Further, the method for generating the preconfigured pulse measurement correction model is:

[0050] dividing the encoder test data into a pulse measurement correction training set and a pulse measurement correction test set;

[0051] The regression network is constructed, the pulse measurement correction training set of the temperature and the electromagnetic noise is taken as the input data of the regression network, the pulse measurement correction training set of the measurement pulse correction number is taken as the output data of the regression network, the regression network is trained, and an initial pulse measurement correction network is obtained.

[0052] The initial pulse measurement correction network is subjected to model verification by using the pulse measurement correction test set, and an initial pulse measurement correction network meeting a preset prediction error is output as a preconfigured pulse measurement correction model.

[0053] Further, the second target input rotating speed is generated based on rotating speed test data; and the rotating speed test data is obtained by the following method:

[0054] S411: An rth rotating speed abnormal event is obtained, where r is a positive integer greater than zero;

[0055] S412: A first target input rotating speed v1 of the servo motor is obtained according to the rth rotating speed abnormal event, and an actual output rotating speed v2 of the servo motor under the control of the first target input rotating speed v1 is obtained;

[0056] S413: A rotating speed error of the first target input rotating speed v1 and the actual output rotating speed v2 is obtained;

[0057] S414: Whether the rotating speed error is equal to zero is judged, if the rotating speed error is not equal to zero, whether the rotating speed error is greater than zero or less than zero is further judged, if the rotating speed error is greater than zero, the first target input rotating speed v1 is incremented, v1=v1+U, and the step S412 is returned; if the rotating speed error is less than zero, the first target input rotating speed v1 is decremented, v1=v1-U, and the step S412 is returned; if the rotating speed error is equal to zero, the first target input rotating speed v1 after the rotating speed adjustment is taken as a second target input rotating speed v3, and the first target input rotating speed v1 and the actual output rotating speed v2 are associated with the second target input rotating speed v3, to obtain a group of relations between the first target input rotating speed v1 and the actual output rotating speed v2 and the second target input rotating speed v3, and U is a preset rotating speed adjustment step;

[0058] S415: The steps S414-S414 are repeated until the r=R, the loop is ended, a group F of relations between the first target input rotating speed v1 and the actual output rotating speed v2 and the second target input rotating speed v3 is obtained, and the group F of relations between the first target input rotating speed v1 and the actual output rotating speed v2 and the second target input rotating speed v3 is taken as the rotating speed test data, R is a preset total rotating speed abnormal time quantity, and F is a rotating speed test data group number.

[0059] Further, the preconfigured rotating speed correction model is generated by the following method:

[0060] The rotating speed test data is divided into a rotating speed training set and a rotating speed test set.

[0061] The learning network is constructed, the first target input rotating speed and the actual output rotating speed in the rotating speed training set are taken as the learning network input data, the second target input rotating speed in the rotating speed training set is taken as the learning network output data, the learning network is trained, and an initial rotating speed correction network is obtained;

[0062] The initial rotating speed correction network is verified by using the rotating speed test set, and the initial rotating speed correction network meeting the preset prediction error is output as a preconfigured rotating speed correction model.

[0063] The intelligent coal preparation equipment servo motor control system based on multi-model adaptation is used for realizing the intelligent coal preparation equipment servo motor control method based on multi-model adaptation, and comprises:

[0064] The data acquisition module is used for acquiring the rotating speed information of the servo motor at the current T time, and acquiring the abnormal evaluation coefficient of the encoder at T+N time after N times according to the preconfigured abnormal detection model, wherein the rotating speed information comprises a measured output rotating speed and a first target input rotating speed, T and N are positive integers greater than zero;

[0065] The abnormality judgment module is used for determining whether the encoder is in an abnormal state at T+N time according to the abnormal evaluation coefficient, if not, continuing to control the servo motor at the first target input rotating speed, and setting T=T+N+M, and triggering the data acquisition module, if yes, acquiring the pulse measurement characteristic data of the encoder at T+N time, and M is a preset time span;

[0066] The measurement correction module is used for recording the measurement pulse number at T+N time, correcting the measurement pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, and determining the actual output rotating speed of the servo motor based on the measurement pulse number after numerical correction;

[0067] The adjustment control module is used for determining whether there is a rotating speed abnormal event, inputting the first target input rotating speed and the actual output rotating speed into the preconfigured rotating speed correction model according to the rotating speed abnormal event, obtaining the second target input rotating speed, and adjusting and controlling the rotating speed of the servo motor according to the second target input rotating speed.

[0068] Compared with the prior art, the present application has the following advantages:

[0069] 1.The application discloses a multi-model adaptive intelligent coal preparation equipment servo motor control system and method, which obtains the speed information at time T and the abnormal evaluation coefficient at time T+N; determines whether the encoder is in an abnormal state at time T+N according to the abnormal evaluation coefficient; records the measured pulse number at time T+N, performs numerical correction on the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, determines the actual output speed of the servo motor based on the numerically corrected measured pulse number; determines whether there is a speed abnormal event, inputs the first target input speed and the actual output speed into the preconfigured speed correction model according to the speed abnormal event, obtains the second target input speed, and performs speed regulation according to the second target input speed; based on the above steps, the application is beneficial to correcting the encoder measurement data based on abnormal detection, further, can adaptively adjust the speed of the servo motor according to the encoder measurement data correction result; thereby, it is beneficial to improve the speed control accuracy of the servo motor in complex scenes, and further, it is beneficial to ensure the control stability of the servo motor.

[0070] 2.The application discloses a multi-model adaptive intelligent coal preparation equipment servo motor control system and method, which can predict the encoder abnormality in the servo motor in advance by introducing an abnormal detection mechanism, thereby avoiding the gradual increase of the servo motor speed control error, and reducing the data processing and feedback judgment times of the servo motor, thereby improving the response speed of the servo motor. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1 The flowchart of the multi-model adaptive intelligent coal preparation equipment servo motor control method of the embodiment of the application;

[0072] Figure 2 The schematic diagram of the multi-model adaptive intelligent coal preparation equipment servo motor control system of embodiment 1 of the application;

[0073] Figure 3 The schematic diagram of the multi-model adaptive intelligent coal preparation equipment servo motor control system of embodiment 2 of the application. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0075] Embodiment 1

[0076] Please refer toFigure 2 The embodiment shown provides a multi-model adaptive-based intelligent coal preparation equipment servo motor control system, which comprises:

[0077] The data acquisition module 210 is configured to acquire the speed information of the servo motor at the current time T, and acquire the abnormal evaluation coefficient of the encoder at T+N time after N times according to the preconfigured abnormal detection model, wherein the speed information includes the measured output speed and the first target input speed, T and N are positive integers greater than zero.

[0078] It should be understood that the speed control of the servo motor can generally be divided into three stages: starting stage, stable stage and stopping stage. In the starting stage, the motor starts rotating from a stationary state. In this stage, sufficient current and torque are usually required to overcome the inertia and friction of the motor to ensure that the motor can start smoothly, so the motor speed is gradually rising. In the stable stage, the motor is successfully started and reaches a certain speed, and the system will enter the stable stage. In this stage, the control system adjusts the current and voltage of the motor to maintain the desired speed, so the motor speed is in stable operation. In the stopping stage, when the motor needs to be stopped, the control system gradually reduces the current of the motor to reduce the torque until the motor stops rotating, so the motor speed gradually decreases.

[0079] It should be noted that the present application is designed for speed control of the servo motor in the stable stage. When the servo motor is in the stable stage, it usually means that the motor is working at a relatively stable speed for a long time. As the working time of the motor increases, the servo motor will face complex working condition changes. At this time, the servo motor is prone to speed mismatch. For high-precision fields such as robots, small speed loss of control may cause huge production accidents. Therefore, the present application is designed around this situation.

[0080] In implementation, the abnormal evaluation coefficient of the encoder at T+N time is acquired according to the preconfigured abnormal detection model, comprising:

[0081] The coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data and voltage difference data of the encoder power supply within T-M time to T time are acquired, M is a preset time span, and M is a positive integer greater than zero.

[0082] It should be noted that: the specific value of M is determined according to the pre-set time span, and the exemplary explanation is that assuming that T time is 9:05, and the set time span is 5 minutes, the value range of M is 5, that is, the encoder coal dust concentration difference data, raw coal load impact difference data, vibration difference data, encoder temperature difference data and encoder power voltage difference data within 9:00 to 9:05 are obtained.

[0083] It should also be noted that: the vibration difference data includes a plurality of vibration difference values, each of which is obtained by calculating the difference between two adjacent vibration values in time; The exemplary explanation is that, based on the above assumption, if the set time span is assumed to be 5 minutes, the vibration difference data obtained is within 9:00 to 9:05, and the time interval between two adjacent times is 10 seconds, then the vibration value at 9:00:10 is subtracted from the vibration value at 9:00:00, the vibration value at 9:00:20 is subtracted from the vibration value at 9:00:10, the vibration value at 9:00:30 is subtracted from the vibration value at 9:00:20, and so on, until the minuend is 9:05, to obtain all vibration difference values. Similarly, the calculation methods of coal dust concentration difference data, raw coal load impact difference data, temperature difference data and voltage difference data are similar to those of vibration difference data, which will not be described here.

[0084] The encoder coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data and encoder power voltage difference data within T-M time to T time are input into the pre-configured abnormality detection model to obtain the abnormality evaluation coefficient of the encoder at T+N time;

[0085] Specifically, the generation method of the pre-configured abnormality detection model is:

[0086] Obtain historical abnormality evaluation data, which includes a plurality of sets of time span encoder coal dust concentration difference data, raw coal load impact difference data, vibration difference data, encoder temperature difference data, encoder power voltage difference data and corresponding abnormality evaluation coefficients.

[0087] It should be noted that: the specific length of each time span is determined according to the pre-set set time span.

[0088] The abnormality evaluation coefficient is obtained by:

[0089] According to the current speed of the servo motor and the output waveform of the encoder within the set time span obtained by the oscilloscope; and according to the current speed to obtain the standard output waveform of the encoder within the set time span;

[0090] Obtain the average amplitude value of the standard output waveform within the set time span, and take the average amplitude value as the amplitude reference value;

[0091] extracting the amplitude value of the output waveform at each time point in the set time span, and performing formula calculation on the amplitude value and the amplitude reference value to obtain an abnormality evaluation coefficient of the encoder; the formula is: ; in the formula: represents the abnormality evaluation coefficient, represents the amplitude value at the gth time point; represents the amplitude reference value; and G is the total number of time points contained in the set time span.

[0092] The historical abnormality evaluation data is preprocessed, and the preprocessing process includes data cleaning, noise removal, missing value processing and error data repair. After the preprocessing is completed, the historical abnormality evaluation data is divided based on different characteristic dimensions or data attributes to obtain a first characteristic data set, a second characteristic data set and a third characteristic data set. The historical abnormality evaluation data includes multiple sets of time span coal dust concentration difference data of the encoder, raw coal load impact difference data of the encoder, vibration difference data, temperature difference data of the encoder, voltage difference data of the encoder power supply and corresponding abnormality evaluation coefficients.

[0093] When the historical abnormality evaluation data is divided, the above characteristic data can be split by type or use. For example, the first characteristic data set can focus on containing multiple sets of time span coal dust concentration difference data of the encoder and corresponding abnormality evaluation coefficients, the second characteristic data set can focus on containing multiple sets of time span raw coal load impact difference data of the encoder, vibration difference data and corresponding abnormality evaluation coefficients, and the third characteristic data set can focus on containing multiple sets of time span temperature difference data of the encoder, voltage difference data of the encoder power supply and corresponding abnormality evaluation coefficients. The first characteristic data set, the second characteristic data set and the third characteristic data set are all derived from the preprocessed historical abnormality evaluation data, and together constitute the basic data required for training the first base learner, the second base learner and the third base learner, and all contain the association between the corresponding characteristic data and the abnormality evaluation coefficient, serving the goal of obtaining the first regression model, the second regression model and the third regression model. The first characteristic data set, the second characteristic data set and the third characteristic data set each contain different specific characteristic data types, corresponding to different characteristic combinations. This division method can focus each data set on a specific characteristic dimension, so that the base learners trained based on each data set can learn the association rules between different characteristics and abnormality evaluation coefficients.

[0094] It should be noted that the historical abnormal evaluation data is obtained by monitoring various sensors and pre-stored in the system database, the various sensors include but are not limited to dust concentration sensor, pressure sensor, temperature sensor, vibration sensor and voltage sensor, and the preprocessing includes but is not limited to data cleaning, noise removal, missing value processing or error data repair.

[0095] Respectively extract the first training data, the second training data and the third training data in the first feature data set, the second feature data set and the third feature data set;

[0096] Construct the first base learner, the second base learner and the third base learner, and train the first training data, the second training data and the third training data by using the first base learner, the second base learner and the third base learner respectively to obtain the first regression model, the second regression model and the third regression model;

[0097] Input the historical abnormal evaluation data into the first regression model, the second regression model and the third regression model respectively for prediction to obtain the first prediction data, the second prediction data and the third prediction data;

[0098] Construct the meta-learner, take the first prediction data, the second prediction data and the third prediction data as the abnormal evaluation data, and divide the abnormal evaluation data into an abnormal evaluation training set and an abnormal evaluation test set;

[0099] Input the abnormal evaluation training set into the meta-learner, and train according to the ensemble learning strategy to obtain an ensemble learning model, use the abnormal evaluation test set to verify the ensemble learning model, and output the ensemble learning model meeting the preset prediction error as the preconfigured abnormal detection model.

[0100] It should be noted that the first prediction data includes the predicted value of the abnormal evaluation coefficient predicted by the first regression model, and also includes the true value of the abnormal evaluation coefficient; similarly, the second prediction data and the third prediction data are also the same; in addition, the meta-learner takes the predicted value of the abnormal evaluation coefficient as the model input data, and takes the true value of the abnormal evaluation coefficient as the model output data, and obtains the ensemble learning model through training.

[0101] It should also be noted that the first base learner, the second base learner and the third base learner can be homogeneous base learners or heterogeneous base learners; the first base learner, the second base learner and the third base learner are specifically one or more of linear regression, decision tree regression, random forest regression or support vector regression prediction model; the ensemble learner is specifically a neural network regression model, and the ensemble learning strategy is specifically a weighted average strategy.

[0102] The abnormality judgment module 220 is configured to determine whether the encoder at the time of T+N is in an abnormal state according to the abnormality evaluation coefficient. If the encoder at the time of T+N is not in an abnormal state, the servo motor is controlled at the first target input rotating speed, T is set to T+N+M, and the data acquisition module 210 is triggered. If the encoder at the time of T+N is in an abnormal state, the pulse measurement characteristic data of the encoder at the time of T+N is acquired. M is a preset time span. The pulse measurement characteristic data specifically includes temperature data of an environment in which the encoder is located at the time of T+N, electromagnetic noise decibel data of the environment in which the encoder is located at the time of T+N, and related attribute data of the measured pulse number directly measured by the encoder at the time of T+N. The related attribute data of the measured pulse number can include a fluctuation amplitude of the measured pulse number at the time, a stability parameter of the measured pulse number in a continuous collection process, and the like. These data can fully reflect the influence of environmental factors such as temperature and electromagnetic noise and the state of the pulse measurement process itself on the accuracy of the measured pulse number, and provide a necessary data basis for subsequent accurate numerical correction of the measured pulse number in combination with the preconfigured pulse measurement correction model.

[0103] In implementation, the determination of whether the encoder at the time of T+N is in an abnormal state includes:

[0104] The abnormality evaluation coefficient is compared with a preset abnormality evaluation coefficient threshold value.

[0105] If the abnormality evaluation coefficient is greater than or equal to the preset abnormality evaluation coefficient threshold value, it is determined that the encoder at the time of T+N is in an abnormal state.

[0106] If the abnormality evaluation coefficient is less than the preset abnormality evaluation coefficient threshold value, it is determined that the encoder at the time of T+N is not in an abnormal state.

[0107] The measurement correction module 230 is configured to record the measured pulse number at the time of T+N, perform numerical correction on the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, and determine the actual output rotating speed of the servo motor based on the numerically corrected measured pulse number.

[0108] In implementation, the numerical correction on the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model includes:

[0109] The pulse measurement characteristic data is input into the preconfigured pulse measurement correction model to obtain a measured pulse correction number.

[0110] The measured pulse correction number and the measured pulse number are accumulated to obtain a numerically corrected measured pulse number.

[0111] Specifically, the pre-configured pulse measurement correction model is generated according to encoder test data; and the encoder test data at least includes a relationship between a measurement pulse correction number and each Celsius temperature and a relationship between the measurement pulse correction number and each decibel electromagnetic noise.

[0112] In one embodiment, the method for obtaining the encoder test data comprises:

[0113] S311: placing the test encoder in a temperature change test environment and placing the standard encoder in a set standard constant temperature test environment;

[0114] S312: obtaining a first test pulse number of the test encoder measured at the i-th Celsius degree under the temperature change test environment, i being a positive integer greater than zero;

[0115] S313: obtaining a first standard pulse data of the standard encoder measured under the set standard constant temperature test environment;

[0116] S314: taking a difference between the first test pulse number and the first standard pulse data as a first pulse data difference, comparing the first pulse data difference with a preset first pulse data difference interval, if the first pulse data difference belongs to the preset first pulse data difference interval, then i = i + 1 and returning to step S312; if the first pulse data difference does not belong to the preset first pulse data difference interval, then taking the first pulse data difference as a measurement pulse correction number and binding and associating the i-th Celsius degree with the measurement pulse correction number to obtain a relationship between the measurement pulse correction number and the i-th Celsius degree;

[0117] S315: repeating the steps S312-S314 until i equals a set temperature Q, thereby ending the cycle and obtaining a relationship between the measurement pulse correction number and each Celsius temperature, Q being a positive integer greater than zero.

[0118] In another embodiment, the method for obtaining the encoder test data further comprises:

[0119] S321: placing the test encoder in an electromagnetic change test environment and placing the standard encoder in a set standard invariable electromagnetic noise test environment;

[0120] S322: obtaining a second test pulse number of the test encoder measured at the j-th decibel electromagnetic noise under the electromagnetic change test environment, j being a positive integer greater than zero;

[0121] S323: obtaining a second standard pulse data of the standard encoder measured under the set standard invariable electromagnetic noise test environment;

[0122] S324: taking the difference value between the second test pulse number and the second standard pulse data as a second pulse data difference, comparing the second pulse data difference with a preset second pulse data difference interval, if the second pulse data difference belongs to the preset second pulse data difference interval, then j = j + 1, and returning to step S322; if the second pulse data difference does not belong to the preset second pulse data difference interval, then taking the second pulse data difference as a measurement pulse correction number, and binding the jth decibel electromagnetic noise with the measurement pulse correction number to obtain the relationship between the measurement pulse correction number and the jth decibel electromagnetic noise;

[0123] S325: repeating the above steps S322-S324 until j equals the set temperature E, and the cycle is ended, to obtain the relationship between the measurement pulse correction number and each decibel electromagnetic noise, E is a positive integer greater than zero.

[0124] In implementation, the generation method of the preconfigured pulse measurement correction model is:

[0125] dividing the encoder test data into a pulse measurement correction training set and a pulse measurement correction test set;

[0126] constructing a regression network, taking the temperature and electromagnetic noise in the pulse measurement correction training set as the regression network input data, taking the measurement pulse correction number in the pulse measurement correction training set as the regression network output data, training the regression network to obtain an initial pulse measurement correction network;

[0127] using the pulse measurement correction test set to verify the model of the initial pulse measurement correction network, and outputting the initial pulse measurement correction network satisfying a preset prediction error as the preconfigured pulse measurement correction model.

[0128] It should be noted that: the regression network is one of a convolutional neural network, a recurrent neural network, or a long short-term memory network.

[0129] It should be understood that: the actual output speed of the servo motor is determined based on the measurement pulse number after numerical correction, and is realized according to an existing calculation formula, the calculation formula is: ; wherein, the resolution of the encoder refers to the number of pulses or encoding values that the encoder can generate when rotating one circle or moving a certain distance, and the resolution is usually represented by the number of pulses or lines; for example, an encoder with a resolution of 1000 lines will generate 1000 pulses when the motor rotates one circle; the time interval represents the time interval between two adjacent pulses; it should also be understood that: the resolution and the time interval of the encoder are determined according to the specific model and specifications of the encoder, and will not be described here.

[0130] The adjustment control module 240 is configured to determine whether a speed abnormal event exists, input the first target input speed and the actual output speed into a preconfigured speed correction model according to the speed abnormal event, obtain a second target input speed, and control the speed of the servo motor according to the second target input speed.

[0131] In implementation, the determination of whether the speed abnormal event exists includes:

[0132] calculating a speed difference between the first target input speed and the actual output speed v2;

[0133] comparing the speed difference with a speed difference threshold value;

[0134] if the speed difference is greater than or equal to the speed difference threshold value, it is determined that the speed abnormal event exists;

[0135] if the speed difference is less than the speed difference threshold value, it is determined that the speed abnormal event does not exist.

[0136] Specifically, the second target input speed is generated based on speed test data; and the speed test data is obtained by:

[0137] S411: obtaining an rth speed abnormal event, r is a positive integer greater than zero;

[0138] S412: obtaining a first target input speed v1 of the servo motor according to the rth speed abnormal event, and obtaining an actual output speed v2 of the servo motor under the control of the first target input speed v1;

[0139] S413: obtaining a speed error of the first target input speed v1 and the actual output speed v2;

[0140] S414: determining whether the speed error is equal to zero, if the speed error is not equal to zero, further determining whether the speed error is greater than zero or less than zero, if the speed error is greater than zero, incrementing the first target input speed v1, setting v1=v1+U, and returning to step S412; if the speed error is less than zero, decrementing the first target input speed v1, setting v1=v1-U, and returning to step S412; if the speed error is equal to zero, taking the first target input speed v1 after speed adjustment as a second target input speed v3, and associating the first target input speed v1 and the actual output speed v2 with the second target input speed v3 to obtain a set of relationships between the first target input speed v1, the actual output speed v2 and the second target input speed v3, U being a preset speed adjustment step, and U being a positive integer greater than zero;

[0141] S415: Repeat the above steps S414-S414 until r=R, and end the loop to obtain the relationship between the F sets of first target input speed v1 and actual output speed v2 and the second target input speed v3, and take the relationship between the F sets of first target input speed v1 and actual output speed v2 and the second target input speed v3 as the speed test data, R is the total number of preset speed abnormal time, F is the number of speed test data sets, and R and F are positive integers greater than zero.

[0142] U is set by combining the speed response characteristics of the servo motor and the control accuracy requirements. If the step size is too large, overshoot will occur during adjustment. If the step size is too small, it will prolong the time to adjust to the target speed. For example, in a certain coal preparation equipment servo system, in order to balance the adjustment speed and accuracy, U can be set to 5 revolutions per minute. When R is set, the number of speed abnormal scenarios that may occur in actual application and the sample size required for model training are considered to ensure that enough typical abnormal situations are covered to improve the generalization ability of the model. For example, to make the preconfigured speed correction model adapt to various working conditions, R can be set to 100. The setting of F is determined by R. Because each speed abnormal event generates a set of valid data after adjustment, the value of F is consistent with R. When R is set to 100, F is also set to 100 accordingly, to ensure that there are enough samples to support the training of the preconfigured speed correction model.

[0143] In implementation, the generation method of the preconfigured speed correction model is:

[0144] Divide the speed test data into a speed training set and a speed test set;

[0145] Construct a learning network, take the first target input speed and the actual output speed in the speed training set as the input data of the learning network, take the second target input speed in the speed training set as the output data of the learning network, train the learning network, and obtain an initial speed correction network;

[0146] Use the speed test set to verify the initial speed correction network, and output the initial speed correction network that meets the preset prediction error as the preconfigured speed correction model.

[0147] It should be noted that: as with the above regression network, the learning network is one of a convolutional neural network, a recurrent neural network, or a long short-term memory network.

[0148] It should be understood that: the first target input speed is the expected speed artificially input according to the scene, and the second target input speed is not artificially input, but the adjustment value made by the system based on the first target input speed according to the abnormal situation, so that the speed of the servo motor is regulated according to the second target input speed, and the actual output speed of the servo motor reaches the expected speed artificially input, i.e. the first target input speed.

[0149] Embodiment 2

[0150] Referring to Figure 3 As shown in the figure, the embodiment discloses an intelligent coal preparation equipment servo motor control system based on multi-model adaptation, which further comprises a cooperative correction module 250 for constructing a cooperative correction model to adapt to the cooperative control requirements of multiple servo motors in the coal preparation equipment, solve the multi-motor cooperation problem in the scenarios such as vibration screen double-motor synchronization, speed matching of the sorting machine and the scraper machine, etc. during the operation of the coal preparation equipment, make up for the limitations of single servo motor control, realize cooperative adjustment through the speed error of multiple motors, and ensure the stability and separation precision of the overall operation of the coal preparation equipment. The cooperative correction model needs to work cooperatively with the preconfigured abnormality detection model and the preconfigured pulse measurement correction model to ensure the accuracy and reliability of multi-motor control.

[0151] The method for constructing the cooperative correction model is as follows:

[0152] Cooperative training data of multiple servo motors of the coal preparation equipment in the historical operation process are acquired, the cooperative training data comprising multiple sets of first target input speed, actual output speed, speed error of each servo motor within a set time span, and operation correlation parameters between each servo motor. The actual output speed of each servo motor is calculated based on the measured pulse number corrected by the preconfigured pulse measurement correction model; the operation correlation parameters are determined according to the specific type of the coal preparation equipment, for example, the operation correlation parameters of the vibration screen double-motor include the speed synchronization error threshold of the double-motor and the load balancing coefficient of the double-motor, and the operation correlation parameters of the sorting machine and the scraper machine include the speed matching coefficient of the two and the matching threshold of the coal flow conveying rate and the separation rate. The specific length of the set time span is determined according to the operation period of the coal preparation equipment to ensure that the acquired data can cover typical working conditions such as coal quality fluctuation and load change.

[0153] The cooperative training data are preprocessed, and the preprocessing process includes data cleaning, noise removal, missing value processing and abnormal value repair. In the data cleaning process, invalid speed data caused by sensor failure are removed; in the noise removal process, a filtering algorithm is used to eliminate the influence of coal dust and electromagnetic interference on the speed data; in the missing value processing process, the linear interpolation method or the average value of the speed data based on the adjacent time points is used to fill in the missing speed information; in the abnormal value repair process, the abnormal data exceeding the preset speed range are replaced by the average value of the normal speed data before and after the corresponding time point, so that the preprocessed cooperative training data can accurately reflect the actual operation state of the multiple motors.

[0154] According to the operation requirements of the coal preparation equipment and the operation logic of the multi-motor, a multi-motor speed correlation matrix is constructed for quantifying the speed correlation degree between each servo motor. First, the correlation weight of each servo motor is determined, and the value of the correlation weight is calculated based on the influence degree of the speed deviation of each motor on the coal preparation production index. For example, the correlation weight of the double motor of the vibrating screen is set to be equal, because the speed synchronization of the two motors directly affects the screening efficiency, and the speed deviation of any motor will cause the screening effect to decrease. The correlation weight of the separator and the scraper is determined according to the coal flow balance requirement. If the influence degree of the conveying rate of the scraper on the separation efficiency of the separator is higher, the correlation weight of the scraper is greater than that of the separator. The correlation weight is calculated by using the analytic hierarchy process, and the final correlation weight value is determined by constructing a judgment matrix and consistency check. The correlation weight of each motor is arranged in rows and columns to form a multi-motor speed correlation matrix.

[0155] A regression network is constructed as the basic network structure of the collaborative correction model. The regression network uses a long short-term memory network, which can effectively handle the time sequence correlation of multi-motor speed data and adapt to the dynamic characteristics of the working condition changes of the coal preparation equipment. The preprocessed collaborative training data is divided into a collaborative training set and a collaborative test set, and the division ratio is determined according to the data size. Usually, the proportion of the collaborative training set is not less than 70%. The first target input speed, the actual output speed, the speed error and the multi-motor speed correlation matrix of each motor in the collaborative training set are used as the input data of the regression network, and the speed collaborative correction value of each motor is used as the output data of the regression network. The regression network is trained to obtain an initial multi-motor speed collaborative correction network. The collaborative test set is used to verify the model of the initial multi-motor speed collaborative correction network, calculate the error between the speed collaborative correction value output by the collaborative correction model and the actual demand correction value, and if the error is less than the preset prediction error, the initial multi-motor speed collaborative correction network is used as the final collaborative correction model; if the error is greater than or equal to the preset prediction error, the parameters of the regression network are adjusted and retrained until the preset prediction error requirement is met.

[0156] In the actual operation process of the coal preparation equipment, the application of the collaborative correction model follows the following process:

[0157] The running data of each servo motor is acquired in real time, including the first target input speed of each motor, the original pulse number measured by the encoder, the load rate and temperature data of the motor. The original pulse number measured by the encoder is input into the preconfigured pulse measurement correction model to obtain the corrected measurement pulse number, and then the actual output speed of each motor is calculated based on the corrected measurement pulse number to ensure that the collected actual output speed data is accurate and reliable. The running time stamp of each motor is recorded to ensure the time synchronization of the multi-motor speed data and avoid collaborative calculation errors caused by data time deviation.

[0158] Firstly, the single motor speed error of each servo motor is calculated, which is the difference between the first target input speed and the actual output speed of the motor. Then, the multi-motor collaborative error is calculated by combining the correlation weight in the multi-motor speed correlation matrix, and the calculation formula of the multi-motor collaborative error is:

[0159] ;

[0160] In the formula, represents the multi-motor collaborative error; K represents the total number of servo motors participating in collaborative control; L represents the number of servo motors associated with the kth motor; represents the single motor speed error of the kth motor; represents the single motor speed error of the lth motor; represents the correlation weight of the kth motor and the lth motor in the multi-motor speed correlation matrix.

[0161] The first target input speed, actual output speed, single motor speed error and calculated multi-motor collaborative error of each motor are input into the collaborative correction model, and the collaborative correction model outputs the speed collaborative correction value of each motor according to the input data. The output logic of the speed collaborative correction value needs to match the working condition of the coal preparation equipment. For example, in the double-motor control scene of the vibrating screen, if the multi-motor collaborative error is greater than the preset synchronization error threshold, and the single motor speed error of the first motor is positive and the single motor speed error of the second motor is negative, the collaborative correction model outputs a negative speed collaborative correction value of the first motor and a positive speed collaborative correction value of the second motor, so that the speed of the double motor tends to be synchronized; in the control scene of the sorting machine and the scraper machine, if the multi-motor collaborative error is greater than the preset matching error threshold, and the single motor speed error of the scraper machine is negative, i.e. the actual speed is lower than the target speed, the collaborative correction model outputs a positive speed collaborative correction value of the scraper machine and a small positive speed collaborative correction value of the sorting machine, to ensure that the coal flow conveying rate and the sorting rate match.

[0162] According to the speed collaborative correction value of each motor output by the collaborative correction model, the second target input speed of the corresponding servo motor is adjusted, and the adjustment formula is:

[0163] ;

[0164] In the formula, represents the adjusted second target input speed of the kth motor; represents the first target input speed of the kth motor; The speed correction value of the kth motor is represented. The servo motor is controlled in speed according to the adjusted second target input speed. After the control is completed, the current actual output speed of each motor is collected in real time, the single motor speed error and the multi-motor coordination error are recalculated, and if the multi-motor coordination error is less than the preset coordination error threshold, the current second target input speed is maintained; if the multi-motor coordination error is greater than or equal to the preset coordination error threshold, the above steps are repeated until the multi-motor coordination error meets the requirements.

[0165] It should be noted that in the implementation process, the multi-motor speed correlation matrix needs to be updated regularly, and the update period is determined according to the wear condition of the coal preparation equipment and the frequency of coal quality changes, usually once every 3 months. When updating, the recent multi-motor operation data is collected again, the new correlation weight is calculated using the original correlation matrix construction method, and it is ensured that the correlation matrix can adapt to the changes of the working condition of the coal preparation equipment.

[0166] It should also be noted that the model needs to work with the pre-configured abnormality detection model. Before applying the coordination correction model, it is determined whether the encoder of each motor is in an abnormal state through the pre-configured abnormality detection model. If there is an abnormal encoder, the encoder measurement data is corrected or switched to redundant encoder data, and then the multi-motor coordination calculation is performed, to avoid the failure of coordinated control caused by abnormal data.

[0167] Through the coordination correction model of the present embodiment, precise coordinated control of multiple servo motors in the coal preparation equipment can be achieved, for example, the speed synchronization error of the double motor of the vibrating screen can be controlled within the preset range, the speed matching degree of the sorting machine and the scraper machine is significantly improved, effectively reducing the problems of coal flow blockage and insufficient separation caused by insufficient multi-motor coordination, and ensuring the continuous stability of coal preparation production and the improvement of clean coal qualification rate.

[0168] Embodiment 3

[0169] Please refer to Figure 1 The present embodiment discloses an intelligent coal preparation equipment servo motor control method based on multi-model self-adaption, which comprises:

[0170] Step 1: Obtain the speed information of the servo motor at the current T time, and obtain the abnormality evaluation coefficient of the encoder at T+N time after N times according to the pre-configured abnormality detection model. The speed information includes the measured output speed and the first target input speed, and T and N are positive integers greater than zero;

[0171] Step 2: Determine whether the encoder is in an abnormal state at T+N time according to the abnormality evaluation coefficient, if not, continue to control the servo motor at the first target input speed, and let T=T+N+M, and return to step 1; if it is in an abnormal state, obtain the pulse measurement feature data of the encoder at T+N time; M is a preset time span;

[0172] Step 3: record the measured pulse number at T+N moment, perform numerical correction on the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, and determine the actual output rotating speed of the servo motor based on the numerically corrected measured pulse number;

[0173] Step 4: determine whether there is a rotating speed abnormal event, input the first target input rotating speed and the actual output rotating speed into the preconfigured rotating speed correction model according to the rotating speed abnormal event, obtain the second target input rotating speed, and perform rotating speed regulation and control on the servo motor according to the second target input rotating speed.

[0174] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0175] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0176] The above is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0177] Finally, the above is merely a preferred embodiment of the present application, and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be covered in the protection scope of the present application.

Claims

1. A method for controlling a servo motor of an intelligent coal preparation equipment based on multi-model self-adaption, characterized in that, The method comprises: Step 1: obtaining the speed information of the servo motor at the current T time, and obtaining the abnormal evaluation coefficient of the encoder at T+N time after N times according to the preconfigured abnormal detection model, wherein the speed information comprises the measured output speed and the first target input speed; Step 2: determining whether the encoder at T+N time is in an abnormal state according to the abnormal evaluation coefficient, if not, continuing to control the servo motor at the first target input speed, and setting T=T+N+M, and returning to step 1; if it is in an abnormal state, obtaining the pulse measurement characteristic data of the encoder at T+N time; M is a preset time span; Step 3: recording the measured pulse number at T+N time, and numerically correcting the measured pulse number according to the pulse measurement characteristic data and the preconfigured pulse measurement correction model, comprising: inputting the pulse measurement characteristic data into the preconfigured pulse measurement correction model to obtain the measured pulse correction number; accumulating the measured pulse correction number and the measured pulse number to obtain the numerically corrected measured pulse number; determining the actual output speed of the servo motor based on the numerically corrected measured pulse number; Step 4: determining whether there is a speed abnormal event, inputting the first target input speed and the actual output speed into the preconfigured speed correction model according to the speed abnormal event to obtain the second target input speed, and controlling the speed of the servo motor according to the second target input speed; The preconfigured pulse measurement correction model is trained and generated according to the encoder test data; the encoder test data at least comprises the relationship between the measured pulse correction number and each Celsius temperature and the relationship between the measured pulse correction number and each decibel electromagnetic noise; The method for obtaining the encoder test data comprises: S311: placing the test encoder in a temperature change test environment, and placing the standard encoder in a set standard constant temperature test environment; S312: obtaining the first test pulse number of the servo motor per revolution of the test encoder at the i Celsius degree in the temperature change test environment, i is a positive integer greater than zero; S313: obtaining the first standard pulse data of the servo motor per revolution of the standard encoder in the set standard constant temperature test environment; S314: taking the difference between the first test pulse number and the first standard pulse data as the first pulse data difference, comparing the first pulse data difference with the preset first pulse data difference interval, if the first pulse data difference belongs to the preset first pulse data difference interval, setting i=i+1, and returning to step S312; if the first pulse data difference does not belong to the preset first pulse data difference interval, taking the first pulse data difference as the measured pulse correction number, and binding the i Celsius degree with the measured pulse correction number to obtain the relationship between the measured pulse correction number and the i Celsius degree; S315: repeating steps S312-S314 until i equals the set temperature Q to end the cycle, obtaining the relationship between the measured pulse correction number and each Celsius temperature, Q is a positive integer greater than zero.

2. The intelligent coal preparation equipment servo motor control method based on multi-model self-adaption according to claim 1, characterized in that, According to the preconfigured abnormal detection model, the abnormal evaluation coefficient of the encoder at T+N time is obtained, comprising: Acquire the coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data of the encoder and the voltage difference data of the encoder power supply from the T-M moment to the T moment, M is a positive integer greater than zero; Input the coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data of the encoder and the voltage difference data of the encoder power supply from the T-M moment to the T moment into the preconfigured abnormality detection model to obtain the abnormality evaluation coefficient of the encoder at T+N moment.

3. The intelligent coal preparation equipment servo motor control method based on multi-model self-adaption according to claim 2, characterized in that, The generation method of the preconfigured abnormality detection model is: Acquire historical abnormality evaluation data, which includes multiple sets of time span coal dust concentration difference data, raw coal load impact difference data, vibration difference data, temperature difference data of the encoder, voltage difference data of the encoder power supply and corresponding abnormality evaluation coefficients; Preprocess the historical abnormality evaluation data to obtain a first feature data set, a second feature data set and a third feature data set; Extract first training data, second training data and third training data in the first feature data set, the second feature data set and the third feature data set respectively; Build a first base learner, a second base learner and a third base learner, and train the first training data, the second training data and the third training data using the first base learner, the second base learner and the third base learner respectively to obtain a first regression model, a second regression model and a third regression model; Input the historical abnormality evaluation data into the first regression model, the second regression model and the third regression model for prediction to obtain first prediction data, second prediction data and third prediction data; Build a meta-learner, take the first prediction data, the second prediction data and the third prediction data as abnormality evaluation data, and divide the abnormality evaluation data into an abnormality evaluation training set and an abnormality evaluation test set; Input the abnormality evaluation training set into the meta-learner, and train according to an ensemble learning strategy to obtain an ensemble learning model, verify the ensemble learning model using the abnormality evaluation test set, and output the ensemble learning model meeting a preset prediction error as the preconfigured abnormality detection model.

4. The intelligent coal preparation equipment servo motor control method based on multi-model self-adaption according to claim 3, characterized in that, The acquisition method of the abnormality evaluation coefficient is: Obtain the output waveform of the encoder within a set time span according to the current speed of the servo motor and through an oscilloscope, and obtain the standard output waveform of the encoder within the set time span according to the current speed; Obtain the average amplitude value of the standard output waveform within the set time span, and take the average amplitude value as an amplitude reference value; Extract the amplitude value of the output waveform at each time point within the set time span, and perform formula calculation on the amplitude value and the amplitude reference value to obtain the abnormality evaluation coefficient of the encoder.

5. The intelligent coal preparation equipment servo motor control method based on multi-model self-adaption according to claim 4, characterized in that, Determine whether the encoder at T+N moment is in an abnormal state, including: Compare the abnormality evaluation coefficient with a preset abnormality evaluation coefficient threshold; If the abnormality evaluation coefficient is greater than or equal to the preset abnormality evaluation coefficient threshold, it is determined that the encoder at T+N moment is in an abnormal state; If the abnormality evaluation coefficient is less than the preset abnormality evaluation coefficient threshold, it is determined that the encoder at T+N moment is not in an abnormal state.

6. The intelligent coal preparation equipment servo motor control method based on multi-model self-adaption according to claim 5, characterized in that, The acquisition method of the encoder test data further includes: S321: Place the test encoder in an electromagnetic variation test environment, and place the standard encoder in a standard invariable electromagnetic noise test environment; S322: In the electromagnetic variation test environment, obtain the second test pulse number of the servo motor per rotation measured by the test encoder under the j decibel electromagnetic noise, j is a positive integer greater than zero; S323: In the standard invariable electromagnetic noise test environment, obtain the second standard pulse data of the servo motor per rotation measured by the standard encoder; S324: Take the difference between the second test pulse number and the second standard pulse data as the second pulse data difference, compare the second pulse data difference with the preset second pulse data difference interval, if the second pulse data difference belongs to the preset second pulse data difference interval, then j = j + 1, and return to step S322; if the second pulse data difference does not belong to the preset second pulse data difference interval, then take the second pulse data difference as the measurement pulse correction number, and bind the j decibel electromagnetic noise and the measurement pulse correction number, to obtain the relationship between the measurement pulse correction number and the j decibel electromagnetic noise; S325: Repeat steps S322-S324 until j equals the set temperature E, and the cycle ends, to obtain the relationship between the measurement pulse correction number and each decibel electromagnetic noise, E is a positive integer greater than zero.

7. The multi-model adaptive based intelligent coal preparation plant servo motor control method according to claim 6, characterized in that, The generation method of the preconfigured pulse measurement correction model is: Divide the encoder test data into a pulse measurement correction training set and a pulse measurement correction test set; Construct a regression network, take the temperature and electromagnetic noise in the pulse measurement correction training set as the regression network input data, take the measurement pulse correction number in the pulse measurement correction training set as the regression network output data, train the regression network, and obtain an initial pulse measurement correction network; Use the pulse measurement correction test set to verify the model of the initial pulse measurement correction network, output the initial pulse measurement correction network that meets the preset prediction error as the preconfigured pulse measurement correction model.

8. The multi-model adaptive based intelligent coal preparation plant servo motor control method according to claim 7, characterized in that, The second target input speed is generated based on speed test data; the acquisition method of the speed test data is: S411: Obtain the rth speed abnormal event, r is a positive integer greater than zero; S412: Obtain the first target input speed v1 of the servo motor according to the rth speed abnormal event, and obtain the actual output speed v2 of the servo motor under the control of the first target input speed v1; S413: Obtain the speed error of the first target input speed v1 and the actual output speed v2; S414: Determine whether the speed error is equal to zero, if the speed error is not equal to zero, further determine whether the speed error is greater than zero or less than zero, if greater than zero, then increment the first target input speed v1, v1 = v1 + U, and return to step S412; If less than zero, then decrement the first target input speed v1, v1 = v1 - U, and return to step S412; If the rotation speed error is equal to zero, the first target input rotation speed v1 after rotation speed adjustment is taken as the second target input rotation speed v3, and the first target input rotation speed v1 and the actual output rotation speed v2 are associated with the second target input rotation speed v3 to obtain a set of relationships between the first target input rotation speed v1, the actual output rotation speed v2 and the second target input rotation speed v3, U is a preset rotation speed adjustment step; S415: repeating steps S414-S414 until r=R, obtaining F sets of relationships between the first target input rotation speed v1, the actual output rotation speed v2 and the second target input rotation speed v3, taking the F sets of relationships between the first target input rotation speed v1, the actual output rotation speed v2 and the second target input rotation speed v3 as rotation speed test data, R is a preset total rotation speed abnormal time quantity, and F is a rotation speed test data group number.

9. The multi-model adaptive based intelligent coal preparation plant servo motor control method according to claim 8, characterized in that, The generation method of the preconfigured rotation speed correction model is: Divide the rotation speed test data into a rotation speed training set and a rotation speed test set; Construct a learning network, take the first target input rotation speed and the actual output rotation speed in the rotation speed training set as learning network input data, take the second target input rotation speed in the rotation speed training set as learning network output data, train the learning network, and obtain an initial rotation speed correction network; Use the rotation speed test set to verify the initial rotation speed correction network, and output the initial rotation speed correction network meeting a preset prediction error as the preconfigured rotation speed correction model.

10. The intelligent coal preparation equipment servo motor control system based on multi-model adaptation, which is used for realizing the intelligent coal preparation equipment servo motor control method based on multi-model adaptation in any one of claims 1-9, characterized in that, Comprise: A data acquisition module is configured to acquire rotation speed information of a servo motor at a current time T, and acquire an abnormality evaluation coefficient of an encoder at a time T+N after N times according to a preconfigured abnormality detection model, wherein the rotation speed information comprises a measured output rotation speed and a first target input rotation speed; An abnormality judgment module is configured to determine whether the encoder is in an abnormal state at the time T+N according to the abnormality evaluation coefficient, if not, continue to control the servo motor at the first target input rotation speed, and set T=T+N+M, and trigger the data acquisition module; If in the abnormal state, the pulse measurement characteristic data of the encoder at the time T+N is acquired; M is a preset time span; A measurement correction module is configured to record the measurement pulse number at the time T+N, correct the measurement pulse number according to the pulse measurement characteristic data and a preconfigured pulse measurement correction model, and determine the actual output rotation speed of the servo motor based on the measurement pulse number after numerical correction; An adjustment control module is configured to determine whether there is a rotation speed abnormal event, input the first target input rotation speed and the actual output rotation speed into the preconfigured rotation speed correction model according to the rotation speed abnormal event, obtain a second target input rotation speed, and control the rotation speed of the servo motor according to the second target input rotation speed.

Images