阅读说明：本技术 一种数据驱动的多驱输送机转矩控制方法及装置 (Data-driven multi-drive conveyor torque control method and device ) 是由 尹小明 岑梁 何海国 王伟 季国良 范津津 林瑞学 汪剑荣 王佳峰 王晟 倪浅雨 于 2021-06-21 设计创作，主要内容包括：本发明公开了一种数据驱动的多驱输送机转矩控制方法及装置,控制方法包括以下步骤：将输送带承载段分成定长的n等份,结合在线图像装载量测量手段,获得每等份输送带上的物料量。并以带速、分布向量、主驱动单元转矩、中部驱动单元转矩、尾部驱动单元转矩为输入量,建立主驱动与中部驱动转矩偏差以及中部驱动与尾部驱动转矩偏差的数据驱动学习模型,本发明结合装载量在线测量手段,采用数据驱动方法获得转矩偏差预测模型,以实现主从驱动单元之间的协调控制。提高卸载式多点驱动输送机的控制性能,提高输送机的安全性。采用LSSVM来完成建模训练,存储空间及在线算力的要求低,适合PLC等工业控制器,解决卸载式多点驱动带式输送机卸载点胶带张力问题。(The invention discloses a data-driven multi-drive conveyor torque control method and a device, wherein the control method comprises the following steps: the bearing section of the conveying belt is divided into n equal parts with fixed length, and the material quantity on each equal part of the conveying belt is obtained by combining an online image loading quantity measuring means. And establishing a data driving learning model of the deviation of the main driving torque and the middle driving torque and the deviation of the middle driving torque and the tail driving torque by taking the belt speed, the distribution vector, the main driving unit torque, the middle driving unit torque and the tail driving unit torque as input quantities. The control performance of the unloading type multi-point driving conveyor is improved, and the safety of the conveyor is improved. The LSSVM is adopted to complete modeling training, the requirements on storage space and on-line force calculation are low, the LSSVM is suitable for industrial controllers such as PLC and the like, and the problem of the tension of the rubber belt at the unloading point of the unloading type multi-point driving belt conveyor is solved.)

1. A data-driven multi-drive conveyor torque control method is characterized by comprising a main drive unit, a middle drive unit and a tail drive unit, and the control method comprises the following steps:

step S1: the head driving unit is set to be in a speed driving mode, and the middle driving unit and the tail driving unit are both set to be in a torque control mode;

step S2: setting a main sampling period T_SCollecting material distribution vector x^kMain drive unit torque T_ETMiddle drive unit torque T_EZTail drive unit torque T_EW；

Step S3: will T_ET、T_EWV addition of x^kIn forming the input vector of the middle drive unit

Will T_EZ、T_ETV distribution vector x of added material^kIn forming the input vector of the tail drive unit

Step S4: calling the off-line built and trained prediction model, reading the pre-stored model parameters, and inputting the vector of the middle driving unitAnd input vector of tail drive unitInput to predictionCalculating in the model to obtain the output value of the model, namely the middle torque deviation delta T_E(TZ)And tail torque deviation Δ T_E(ZW)；

Step S6: according to the mid-section torque deviation Delta T_E(TZ)And tail torque deviation Δ T_E(ZW)And setting a torque adjustment value of the middle driving unit and a torque adjustment value of the tail driving unit.

2. The method as claimed in claim 1, wherein said step of obtaining the material distribution vector in step S1 comprises the steps of:

step S11: dividing the conveyor belt bearing section into n equal conveyor belt sections with fixed length to obtain the instantaneous loading q of the conveyor belt;

step S12: the material quantity of the first section of the conveying belt is obtained as follows:

wherein, N ═ t₁/t_s1 is the length of the conveyor belt, v is the speed of the conveyor belt, t_l＝l/v，t_sIs a sampling period;

step S13: throughObtaining a material distribution vector on the conveying belt after time:

x＝(x₁，x₂...x_m，x_m+1…x_n)，

wherein s is the belt length between the middle driving winding-out point and the blanking point, and v is the belt speed of the conveying belt.

3. The method as claimed in claim 2, wherein the step of building and training the predictive model of step S4 comprises the steps of:

s41: setting the head driving unit, the middle driving unit and the tail driving unit to be in a speed control mode;

s42: setting on-line load measurement period t_sStarting on-line measurement of the loading capacity; setting variable sampling period T_SObtaining v,T_ET、T_EZ、T_EW，T_s＞t_s；

S43: will T_ET、T_EWV is added to X to form the input vector of the central drive unit, X_Z＝(x₁，x₂...x_m，x_m+1...x_n，v，T_EW，T_ET) (ii) a Will T_EZ，T_ETV addition ofIn, form the tail drive unit input vector, X_w＝(x₁，x₂...x_m，x_m+1…x_n，v，T_EZ，T_ET)；

S44: the method comprises the steps of collecting data of multiple working conditions of a conveyor, and processing the data to obtain a middle driving unit learning sampleAnd tail drive unit learning samplesDeposit into a sample setAndwhere k is the kth T_SPeriod, N^ZRepresenting the number of samples of the middle drive unit, N^WRepresenting the number of tail drive unit samples;

s45: modeling and training a sample set through an LSSVM algorithm;

s46: and after the training is finished, storing the LSSVM model parameters into the PLC for online use.

4. The method as claimed in claim 3, wherein the step S44 further comprises the step of

Step S441: establishing a head drive unit to mid drive unit torque differential Δ T_E(TZ)Prediction model f₁Torque difference Δ T between the middle drive unit and the tail drive unit_E(ZW)Prediction model f₂：

ΔT_E(TZ)＝f₁(X_z)，

ΔT_E(ZW)＝f₂(X_W)；

Step S442: by T_SFor periodic acquisition of T_ET、T_EZ、T_EWAnd v, forming a sample through steps S21 and S22Adding model f₁Collecting the sample to form a sampleAdding f₂Collecting samples, wherein K is the Kth collection period;

step S443: obtaining a sample setAnd

5. the data-driven multi-drive conveyor torque control method as claimed in claim 4, wherein the step S45 of performing modeling training on the sample set through the LSSVM algorithm comprises: establishing a torque deviation decision function of the middle driving unit:

tail drive unit torque deviation decision function:

wherein alpha is a vector of support values, b is an offset, sigma is a kernel parameter, c is a normalization parameter,andis the kernel function of the LSSVM algorithm.

6. The method of claim 5, wherein the LSSVM model parameters of step S46 include: the vector of support values α, offset b, kernel parameter σ, is the normalization parameter c.

7. The method as claimed in claim 6, wherein the kernel function is a radial basis kernel function (RBF).

8. The method as claimed in claim 7, wherein the step S6 further includes a torque deviation Δ T at the center portion_E(TZ)Upper superimposed artificial correction value delta T_ZThen forming the torque adjustment value of the middle driving unit and the torque deviation delta T at the tail part_E(ZW)Upper superimposed artificial correction value delta T_WAnd then forming a tail drive unit torque adjustment value.

9. The method as claimed in claim 8, wherein the step S6 further comprises a limiting module for determining whether the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit exceed the limiting limit, respectively, if so, limiting the torque adjustment value to an upper limit value, and reducing the torque change rate, otherwise, outputting the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit.

10. A data-driven multi-drive conveyor torque control apparatus, characterized by adopting the data-driven multi-drive conveyor torque control method according to any one of claims 1 to 9.

Technical Field

The invention relates to the field of conveying devices, in particular to a torque control method and device for a data-driven multi-drive conveyor.

Background

The belt conveyor system is used as a continuous bulk material conveying device, is widely applied to various industries of national economy such as steel, coal, ports and docks, electric power, building materials and the like, and has large use amount and wide application range. The belt conveyor is continuously developed towards large-scale, long-distance, high-speed, large-capacity and intelligentized direction. When designing and developing a long-distance belt conveyor system, because a single drive motor can provide limited driving force and the maximum tension force that a conveying belt can bear is limited, a large belt conveyor usually adopts a mode of driving by a plurality of motors, so that the capacity of the single motor can be reduced, and the tension force of the belt can be reduced. The multipoint driving not only reduces the requirement on the strength of the adhesive tape, but also facilitates the type selection of equipment and realizes the miniaturization of the equipment. Essentially multi-point driving is one way to improve cost performance. The multi-point driving mode is many, including linear friction type, wire rope traction type and middle roller unloading type. The roller unloading type multi-point driving is to add one or more groups of driving devices in the middle of a common belt conveyor to distribute the driving force normally arranged at the head part to several parts. The unloading type intermediate driving mode has a simple structure, is convenient to arrange, can effectively reduce the cost of manufacturing, processing, conveying, mounting, maintenance, management and the like, and is more adopted in a multi-drive conveyor system. However, the middle transfer point causes the tension distribution of the whole adhesive tape to be completely different from the centralized driving arrangement form, the tension of the adhesive tape at the winding-out side of the middle driving wheel is obviously reduced, and the adhesive tape can slip if the control is not proper; the middle transfer point also causes the materials to fall for multiple times, the materials are not uniformly distributed on the rubber belt, and the driving system of the middle roller unloading type multi-drive conveyor is difficult to control.

For example, a novel belt conveyor disclosed in chinese patent literature, No. CN103662715, includes a body, an unloading unit, a nose, a belt storage and tensioning unit, a belt retracting and releasing device, a driving device, a transition deviation-preventing front-mounted carrier roller set, a belt, a self-moving tail, a belt protection and belt conveyor control system, and a video monitoring system. But above-mentioned scheme passes through tail cylinder and drives the belt operation, through converter control drive drum work, only afterbody cylinder drive, and the goods on the sticky tape distributes when uneven in long distance transportation and can lead to the distribution of whole sticky tape tension unbalanced, even adjust the afterbody cylinder also can't make whole sticky tape tension balanced, if control improper existence can cause the sticky tape to skid, the problem that middle reprint point still can make the material whereabouts.

Disclosure of Invention

The invention aims to solve the problem that the unloading type multi-point driving conveyor in the prior art slips due to the fact that the tension of a rubber belt wound out of a driving roller is reduced under the condition of working condition change, and provides a torque deviation prediction model between a main driving unit and a driven driving unit by adopting a data driving method, so that the coordination control between the main driving unit and the driven driving unit is realized, and the control performance of the unloading type multi-point driving conveyor and the safety of the conveyor are improved.

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

a data-driven multi-drive conveyor torque control method comprises a main drive unit, a middle drive unit and a tail drive unit, and comprises the following steps:

step S1: the head driving unit is set to be in a speed driving mode, and the middle driving unit and the tail driving unit are both set to be in a torque control mode;

step S2: setting a main sampling period T_SCollecting material distribution vector x^kMain drive unit torque T_ETMiddle drive unit torque T_EZTail drive unit torque T_EW；

Step S3: will T_ET、T_EWV addition of x^kIn forming the input vector of the middle drive unit

Will T_EZ、T_ETV distribution vector x of added material^kIn forming the input vector of the tail drive unit

Step S4: calling the off-line built and trained prediction model, reading the pre-stored model parameters, and inputting the vector of the middle driving unitAnd input vector of tail drive unitInputting the data into a prediction model for calculation to obtain a model output value, namely a middle torque deviation delta T_E(TZ)And tail torque deviation Δ T_E(ZW)；

Step S6: according to the mid-section torque deviation Delta T_E(TZ)And tail torque deviation Δ T_E(ZW)And setting a torque adjustment value of the middle driving unit and a torque adjustment value of the tail driving unit.

Preferably, the obtaining of the material distribution vector in step S1 includes the following steps:

step S11: dividing the conveyor belt bearing section into n equal conveyor belt sections with fixed length to obtain the instantaneous loading q of the conveyor belt;

step S12: the material quantity of the first section of the conveying belt is obtained as follows:

wherein, N ═ t₁/t_sL is the length of the conveyor belt, v is the speed of the conveyor belt, t_l＝l/v，t_sIs a sampling period;

step S13: throughObtain the output after the timeMaterial distribution vector on belt:

x＝(x₁，x₂...x_m，x_m+1...x_m)，

wherein s is the belt length between the middle driving winding-out point and the blanking point, and v is the belt speed of the conveying belt.

Along with the time, the material will pass backward in proper order on the conveyer belt, forms the linked list: x is the number of₁→x₂→……→x_m→x_m+1→……→x_n. When the conveyor run time is greater thanThereafter, the loading of each segment will become a known amount. Vector x ═ x₁，x₂…x_m，x_m+1…x_n) The material distribution on the conveyor belt is described, in this specification referred to as material distribution vector. And periodically acquiring images of the tail rubber belt, and calculating the instantaneous loading capacity of the belt conveyor. The trend of the loading capacity is obtained by data-driven modeling, and the absolute value of the trend is not needed, so that a measurement system does not need to be calibrated, and the method is more convenient to realize.

Preferably, the building and training of the prediction model in step S4 includes the following steps:

s41: setting the head driving unit, the middle driving unit and the tail driving unit to be in a speed control mode;

s42: setting on-line load measurement period t_sStarting on-line measurement of the loading capacity; setting variable sampling period T_SObtaining v, x, T_ET、T_EZ、T_EW，T_s＞t_s；

S43: will T_ET、T_EWV is added to X to form the input vector of the central drive unit, X_Z＝(x₁，x₂...xm，x_m+1…x_n，v，T_EW，T_ET) (ii) a Will T_EZ，T_ETV addition ofIn (1), a tail part is formedInput vector, X, of the drive unit_W＝(x₁，x₂...x_m，x_{m+ 1}...x_n，v，T_EZ，T_ET)；

S44: the method comprises the steps of collecting data of multiple working conditions of a conveyor, and processing the data to obtain a middle driving unit learning sampleAnd tail drive unit learning samplesDeposit into a sample setWhere k is the kth T_SPeriod, N^ZRepresenting the number of samples of the middle drive unit, N^WRepresenting the number of tail drive unit samples;

s45: modeling and training a sample set through an LSSVM algorithm;

s46: and after the training is finished, storing the LSSVM model parameters into the PLC for online use.

Wherein N is^εFor sample capacity, since the learning samples need to be removed by adopting a sparsification algorithm in the LSSVM training process, N is^εLess than the number N of original samples^Z。

Conveyor drive unit motor torque T_EDriving circumferential force F with driving wheel_uIn a proportional relationship, i.e. T_E＝k·F_uAnd the driving circumferential force is equal to the difference between the tension at the winding point and the tension at the separation point. The conveyer belt is annular, the tension of the winding point of the main driving wheel, the middle driving wheel and the tail driving wheel is coupled with the tension of the separation point, and the resistance of the adhesive tape is also influenced by the running state, so that the accurate analytic relation of the torques of the multiple driving units cannot be obtained. However, the torque deviation model can be obtained by adopting a learning modeling method because a definite relation exists among the torque of the main driving unit, the torque of the auxiliary driving unit and the real-time running state of the conveyor.

If the main driving unit adopts speed control, the middle slave driving unit and the tail slave driving unit also adopt speed control, and the speed set value tracks the main driving frequency converter. In this case, there is no coupling relationship between the torques of the drive units, and the difference depends mainly on the operating parameters of the conveyor, such as the operating speed and the material distribution. The torque prediction model can be established by utilizing the operation data obtained in the working mode.

Preferably, the step S44 further includes the following steps

Step S441: establishing a head drive unit to mid drive unit torque differential Δ T_E(TZ)Prediction model f₁Torque difference Δ T between the middle drive unit and the tail drive unit_E(ZW)Prediction model f₂：

ΔT_E(TZ)＝f₁(X_z)，

ΔT_E(ZW)＝f₂(X_W)；

Step S442: by T_STo periodically collect TET, TEZ, TEW, and v, samples are formed through steps S21 and S22Adding model f₁Collecting the sample to form a sampleAdding f₂Collecting samples, wherein K is the Kth collection period;

step S443: obtaining a sample setAnd

setting the main drive unit as speed control, the middle slave drive unit and the tail slave drive unit as speed control, setting the speed set value to track the main drive controller, and setting the data sampling period as T_sAnd T is_S＞t_sAccording to t_sThe on-line measurement of the load capacity is periodically executed, and the material distribution phasor X is periodically updated.

Preferably, the step S45 of performing modeling training on the sample set through the LSSVM algorithm includes: establishing a torque deviation decision function of the middle driving unit:

tail drive unit torque deviation decision function:

wherein alpha is a vector of support values, b is an offset, sigma is a kernel parameter, c is a normalization parameter,andis the kernel function of the LSSVM algorithm. After the modeling training is completed, the support vectors in the learning sample set are required to be collectedAnd a support value vector alpha, an offset b, a kernel parameter sigma, a normalized parameter c and the like are stored in the PLC. Approximating a non-linear model f by training using a learning sample set₁、f₂The LSSVM model has low requirement on storage space and low requirement on calculated amount, and is suitable for industrial controllers such as PLC.

Preferably, the LSSVM model parameters in step S46 include: the vector of support values α, offset b, kernel parameter σ, is the normalization parameter c. And in the online application stage, the extracted parameters are subjected to prediction calculation. In actual use, a kernel function K can be called for calculation, and accumulated summation is carried out to obtain a model prediction value.

Preferably, the kernel function is a radial basis kernel function RBF.

Wherein alpha is_kB is the offset, which is the support value for the corresponding sample. Model parameter alpha_kAnd b can be solved by the following system of equations:

wherein the content of the first and second substances,

U＝(K+c^-1I)^-1；

in the U expression, I is a unit matrix, sigma is a nuclear parameter, and c is a normalized parameter, namely a penalty coefficient;

preferably, the step S6 further includes a torque deviation Δ T at the middle part_E(TZ)Upper superimposed artificial correction value delta T_ZThen forming the torque adjustment value of the middle driving unit and the torque deviation delta T at the tail part_E(ZW)Upper superimposed artificial correction value delta T_WAnd then forming a tail drive unit torque adjustment value. Ideally,. DELTA.T_Z＝0，ΔT_WWhen the model is not matched, the output value of the model is predicted, the output value of the model is adjusted, and the output value of the model is adjusted.

Preferably, step S6 further includes a limiting module, which determines whether the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit exceed the limiting limits, respectively, if so, the torque adjustment value is limited to an upper limit value, and the torque change rate is appropriately reduced, otherwise, the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit are output. The upper limit of the amplitude limit is delta increase on the basis of the current torque, and the lower limit of the amplitude limit is delta decrease on the basis of the current actual torque. If the difference between the torque set value and the current torque exceeds delta, the change rate of the adjustment value is properly reduced through torque amplitude limiting, and the torque adjustment of the middle driving unit and the tail driving unit is gradually completed.

A data-driven multi-drive conveyor torque control device adopts the data-driven multi-drive conveyor torque control method. The head of the conveyor is provided with a main driving unit, the middle of the conveyor is provided with a middle driving unit, the tail of the conveyor is provided with a tail driving unit, each driving unit comprises two driving motors which are coaxially arranged, a motor control system controller adopts a PLC and adopts a frequency converter for driving, and the PLC and the frequency converter communicate by adopting a field bus protocol so as to realize the coordination control among the driving units. The bus is made of optical fiber materials, so that electromagnetic interference is prevented, and lightning stroke is prevented.

Therefore, the invention has the following beneficial effects: (1) the invention combines the on-line measuring means of the loading capacity and adopts a data driving method to obtain the torque deviation prediction models of the main driving unit, the middle driving unit and the tail driving unit so as to realize the coordination control between the main driving unit and the slave driving unit. (2) The data driving modeling means obtains the set torque value of the slave driving set, improves the control performance of the unloading type multipoint driving conveyor and improves the safety of the conveyor. (3) The LSSVM is adopted to complete modeling training, the LSSVM has low requirements on storage space and on-line force calculation, is suitable for industrial controllers such as PLC and the like, and solves the problem of the tension of the rubber belt at the unloading point of the unloading type multi-point drive belt conveyor.

Drawings

Fig. 1 is a schematic view of a multi-drive conveyor model according to an embodiment of the invention.

Fig. 2 is a schematic tension diagram of a multi-drive conveyor according to an embodiment of the invention.

FIG. 3 is a block diagram of a multi-drive conveyor control according to an embodiment of the invention.

Fig. 4 is a schematic diagram of material distribution phasors for a multi-drive conveyor according to an embodiment of the invention.

Fig. 5 is a schematic view of an input vector of a multi-drive conveyor according to an embodiment of the invention.

FIG. 6 is a schematic diagram of the control system mechanism of the multi-drive conveyor according to one embodiment of the invention.

Fig. 7 is a flow chart for modeling and using the multi-drive conveyor torque control method according to an embodiment of the invention.

In the figure: 1. main drive unit 2, middle part drive unit 3, afterbody drive unit 4, camera and light source 5, material.

Detailed Description

The invention is further described with reference to the following detailed description and accompanying drawings.

Example (b):

as shown in fig. 1 to 7, 3 driving units are configured at the head, the middle and the tail of the conveyer, and each driving unit comprises two coaxially mounted driving motors. The motor control system controller adopts a PLC and adopts a frequency converter for driving, and the PLC and the frequency converter communicate by adopting a field bus protocol so as to realize the coordination control among the driving units. The bus is made of optical fiber materials, so that electromagnetic interference is prevented, and lightning stroke is prevented.

Two motors in each group of driving units are coaxially arranged, belong to rigid body connection and are controlled by a master motor and a slave motor, and a slave motor frequency converter works in a direct torque control mode and receives a set torque value sent by a master motor. In the following description, the drive unit consisting of two coaxially mounted electric motors is described as a whole.

The torque of each drive unit is related to the distribution of material on the belt, the speed of the belt, etc. while the conveyor is running. The embodiment adopts an image processing means to realize the real-time measurement of the instantaneous conveying capacity, as shown in figure 1, an industrial camera uses t_sPeriodically acquiring the adhesive tape image at the head driving unit, and calculating the instantaneous loading capacity of the belt conveyor, wherein the method comprises the following steps: as shown in fig. 4, the tape carrier segment is divided into n equal parts of fixed length, each equal part has a length of 1, in the inclined carrier segment, the length is calculated by the actual tape length, not by the horizontal projection, and the rear end of the middle unloading station takes the blanking point as the starting point of calculation. The method for calculating the instantaneous loading of the belt conveyor comprises the following steps: the industrial camera periodically acquires the tail rubber belt image and calculates the instantaneous loading capacity of the belt conveyor. The trend of the loading capacity is obtained by data-driven modeling, and the absolute value of the trend is not needed, so that the trend of the loading capacity does not need to be correctedThe measurement system is calibrated, and the realization is more convenient.

Step S11: dividing the conveyor belt bearing section into n equal conveyor belt sections with fixed length to obtain the instantaneous loading q of the conveyor belt; step S12: the material quantity of the first section of the conveying belt is obtained as follows:

wherein, N ═ t₁/t_sL is the length of the conveyor belt, v is the speed of the conveyor belt, t_l＝l/v，t_sIs a sampling period;

step S13: throughObtaining a material distribution vector on the conveying belt after time:

x＝(x₁，x₂...x_m，x_m+1...x_n)，

wherein s is the belt length between the middle driving winding-out point and the blanking point, and v is the belt speed of the conveying belt.

Along with the time, the material will pass backward in proper order on the conveyer belt, forms the linked list: x is the number of₁→x₂→……→x_m→x_m+1→……→x_n. When the conveyor run time is greater thanThereafter, the loading of each segment will become a known amount. Vector x ═ x₁，x₂…x_m，x_m+1...x_n) The material distribution on the conveyor belt is described, in this specification referred to as material distribution vector.

The off-line establishment of the trained torque deviation prediction model comprises the following steps:

s41: setting the head driving unit, the middle driving unit and the tail driving unit to be in a speed control mode;

s42: setting on-line load measurement period t_sStart of on-line measurement of load(ii) a Setting variable sampling period T_SObtaining v, x, T_ET、T_EZ、T_EW，T_s＞t_s；

S43: will T_ET、T_EWV is added to X to form the input vector of the central drive unit, X_z＝(x₁，x₂...x_m，x_{m+ 1}...x_n，v，T_EW，T_ET) (ii) a Will T_EZ，T_ETV addition ofIn, form the tail drive unit input vector, X_W＝(x₁，x₂...x_m，x_m+1...x_n，v，T_EZ，T_ET)；

S44: the method comprises the steps of collecting data of multiple working conditions of a conveyor, and processing the data to obtain a middle driving unit learning sampleAnd tail drive unit learning samplesDeposit into a sample setAndwhere k is the kth T_SPeriod, N^ZRepresenting the number of samples of the middle drive unit, N^WRepresenting the number of tail drive unit samples;

step S441: establishing a head drive unit to mid drive unit torque differential Δ T_E(TZ)Prediction model f₁Torque difference Δ T between the middle drive unit and the tail drive unit_E(ZW)Prediction model f₂：

ΔT_E(TZ)＝f₁(X_z)，

ΔT_E(ZW)＝f₂(X_W)；

Step S442: by T_SFor periodic acquisition of T_ET、T_EZ、T_EWAnd v, forming a sample through steps S21 and S22Adding model f₁Collecting the sample to form a sampleAdding f₂Collecting samples, wherein K is the Kth collection period;

step S443: obtaining a sample setAnd

setting the main drive unit as speed control, the middle slave drive unit and the tail slave drive unit as speed control, setting the speed set value to track the main drive controller, and setting the data sampling period as T_sAnd T is_S＞t_sAccording to t_sThe on-line measurement of the load capacity is periodically executed, and the material distribution phasor X is periodically updated.

S45: modeling and training a sample set through an LSSVM algorithm;

establishing a torque deviation decision function of the middle driving unit:

tail drive unit torque deviation decision function:

wherein alpha is a vector of support values, b is an offset, sigma is a kernel parameter, c is a normalization parameter,andis the kernel function of the LSSVM algorithm. After the modeling training is completed, the support vectors in the learning sample set are required to be collectedAnd a support value vector alpha, an offset b, a kernel parameter sigma, a normalized parameter c and the like are stored in the PLC. Approximating a non-linear model f by training using a learning sample set₁、f₂The LSSVM model has low requirement on storage space and low requirement on calculated amount, and is suitable for industrial controllers such as PLC.

The kernel function is a radial basis kernel function RBF.

Wherein alpha is_kB is the offset, which is the support value for the corresponding sample. Model parameter alpha_kAnd b can be solved by the following system of equations:

wherein the content of the first and second substances,

U＝(K+c^-1I)^-1；

in the U expression, I is a unit matrix, sigma is a nuclear parameter, and c is a normalized parameter, namely a penalty coefficient;

s46: and after the training is finished, storing the support value model parameters into the PLC for online use.

Wherein N is^εFor the sample capacity, the LSSVM training process needs to adopt a sparsification algorithmEliminate the learning samples, so N^εLess than the number N of original samples^Z。

Conveyor drive unit motor torque T_EDriving circumferential force F with driving wheel_uIn a proportional relationship, i.e. T_E＝k·F_uAnd the driving circumferential force is equal to the difference between the tension at the winding point and the tension at the separation point. The conveyer belt is annular, the tension of the winding point of the main driving wheel, the middle driving wheel and the tail driving wheel is coupled with the tension of the separation point, and the resistance of the adhesive tape is also influenced by the running state, so that the accurate analytic relation of the torques of the multiple driving units cannot be obtained. However, the torque deviation model can be obtained by adopting a learning modeling method because a definite relation exists among the torque of the main driving unit, the torque of the auxiliary driving unit and the real-time running state of the conveyor.

If the main driving unit adopts speed control, the middle slave driving unit and the tail slave driving unit also adopt speed control, and the speed set value tracks the main driving frequency converter. In this case, there is no coupling relationship between the torques of the drive units, and the difference depends mainly on the operating parameters of the conveyor, such as the operating speed and the material distribution. The torque prediction model can be established by utilizing the operation data obtained in the working mode.

The online torque control of the multi-drive conveyor comprises the following steps:

step S1: the head driving unit is set to be in a speed driving mode, and the middle driving unit and the tail driving unit are both set to be in a torque control mode;

step S2: setting a main sampling period T_SCollecting material distribution vector x^kMain drive unit torque T_ETMiddle drive unit torque T_EZTail drive unit torque T_EW；

Step S3: will T_ET、T_EWV addition of x^kIn forming the input vector of the middle drive unit

Will T_EZ、T_ETV distribution vector x of added material^kIn (1), a tail part is formedInput vector of drive unit

Step S4: calling the off-line built and trained prediction model, reading the pre-stored model parameters, and inputting the vector of the middle driving unitAnd input vector of tail drive unitInputting the data into a prediction model for calculation to obtain a model output value, namely a middle torque deviation delta T_E(TZ)And tail torque deviation delta_TE(ZW)；

Step S6: according to the mid-section torque deviation Delta T_E(TZ)And tail torque deviation Δ T_E(ZW)And setting a torque adjustment value of the middle driving unit and a torque adjustment value of the tail driving unit. At mid-section torque deviation Δ T_E(TZ)Upper superimposed artificial correction value delta T_ZThen forming the torque adjustment value of the middle driving unit and the torque deviation delta T at the tail part_E(ZW)Upper superimposed artificial correction value delta T_WAnd then forming a tail drive unit torque adjustment value. And respectively judging whether the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit exceed the amplitude limit, if so, limiting the torque adjustment value to an upper limit value, properly reducing the torque change rate, and if not, outputting the torque adjustment value of the middle driving unit and the torque adjustment value of the tail driving unit.

The control performance of the unloading type multi-point driving conveyor is improved, and the safety of the conveyor is improved.

The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.

Although the terms torque adjustment, predictive model, material distribution phasor, on-time loading, torque bias, etc. are used more herein, the possibility of using other terms is not excluded. These terms are used merely to more conveniently describe and explain the nature of the present invention; they are to be construed as being without limitation to any additional limitations that may be imposed by the spirit of the present invention.