Intelligent detection and control method and device for winding mechanism

Abstract

The invention discloses an intelligent detection and control method and device for a winding mechanism, and the device comprises a winding control device, a gateway node, an on-site monitoring end, a cloud platform, a remote monitoring end, and a mobile phone App, and the winding control device transmits the detected operation parameters to the on-site monitoring end and the cloud platform through the gateway node. Two-way transmission of operation parameters and control information of the winding control device is achieved among the winding control device, the gateway node, the cloud platform, the field monitoring terminal, the remote monitoring terminal and the mobile phone App, and intelligent monitoring of the operation parameters of the winding control device is achieved through the field monitoring terminal, the remote monitoring terminal and the mobile phone App. In order to improve the speed and tension stability, the intelligent detection and control method and device of the winding mechanism are designed, the intelligent detection and stable control method of the speed and tension is obtained, and it can be known through speed and tension control results that the method can obtain the rapid and stable state of an unwinding device and a winding device; and the speed and the tension can be continuously kept stable even if being interfered.

Description

Technical Field

[0001] The present invention relates to an automated equipment technology for winding mechanism detection and intelligent winding mechanism control, and in particular to an intelligent winding mechanism detection and control method and device. Background Art

[0002] The stable operation of the unwinding and winding devices during product processing mainly depends on the control of speed and tension. If the product quality is to be stable, the fluctuation of speed and tension should not be too large. Excessive speed and tension will directly reduce the service life of the product, and too small speed and tension will affect the product quality. The continuous change of the outer diameter of the unwinding and winding devices and the instability of the unwinding and winding devices are important factors that cause the speed and tension to change. In order to improve the stable control of speed and tension during product processing, an intelligent detection and control method and device for winding mechanism are designed. On the basis of the Internet of Things, cloud platform and intelligent control, the traditional speed and tension stability control method is optimized to ensure that the overshoot of its stable control can basically be within the allowable error range, which can reduce the stabilization time and improve the stability of speed and tension control in the product processing process. The adaptive adjustment of the speed and tension control of the unwinding and winding devices is realized, and a new method for intelligent detection and stable control of speed and tension is provided. Summary of the invention

[0003] In order to improve the speed and tension stability, the present invention designs an intelligent detection and control method and device for a winding mechanism, and obtains an intelligent detection and stable control method for speed and tension. From the speed and tension control results, it can be seen that this method can obtain a fast and stable state of the unwinding and winding devices, and can continue to maintain speed and tension stability even if disturbed.

[0004] To solve the above problems, the present invention adopts the following technical solutions:

[0005] 1. Intelligent detection and control method of winding mechanism:

[0006] 1. Construct a parameter detection module, which includes a TDL beat delay device, an AANN autoassociative neural network model-metabolism gray GM (1,1) model, a PSO BiLSTM neural network model, and a PSO DRNN neural network model. The structure and function of the parameter detection module are shown in Figure 1 shown.

[0007] The parameter sensor output is used as the input of TDL beat delay device A, the output of the temperature sensor and TDL beat delay device A are used as the corresponding inputs of the AANN self-associative neural network model-metabolism grey GM (1,1) model respectively, the time series difference between the parameter sensor output and the output of the AANN self-associative neural network model-metabolism grey GM (1,1) model is used as the input of TDL beat delay device B and the corresponding input of the DRNN neural network model of PSO respectively, the output of the AANN self-associative neural network model-metabolism grey GM (1,1) model is used as the input of TDL beat delay device C, the output of TDL beat delay device C and the output of TDL beat delay device B are used as the corresponding inputs of the BiLSTM neural network model of PSO respectively, the output of the BiLSTM neural network model of PSO is used as the corresponding input of the DRNN neural network model of PSO, and the DRNN neural network model of PSO outputs the parameter sensor output prediction value;

[0008] 2. Decoupling subsystem for winding mechanism detection and intelligent control, the system includes TDL beat delay device, TS neural network tension and speed decoupling setting model, PSO's NARX neural network tension dynamic setting model, PI controller-PSO's BiLSTM neural network tension controller, PSO's NARX neural network speed dynamic setting model, PI controller-PSO's BiLSTM neural network speed controller, PSO's DRNN neural network unwinding controller, PSO's NARX neural network speed dynamic tracking model, PI controller-PSO's BiLSTM neural network speed tracking controller and parameter detection module; the structure and function of the decoupling subsystem for winding mechanism detection and intelligent control can be seen in Figure 2 shown.

[0009] The tension set value is used as the input of TDL beat delay device A and the corresponding input of the NARX neural network tension dynamic setting model of PSO, the speed set value is used as the input of TDL beat delay device B and the corresponding input of the NARX neural network speed dynamic setting model of PSO, the output of TDL beat delay device A and TDL beat delay device B are used as the corresponding input of the TS neural network tension and speed decoupling setting model, and the output tension and speed decoupling setting values ​​of the TS neural network tension and speed decoupling setting model are used as the NARX neural network tension dynamic setting model of PSO and the NARX neural network speed dynamic setting model of PSO. The corresponding input of the setting model, the error and error change rate of the output of the NARX neural network tension dynamic setting model of PSO and the output of the parameter detection module 4 are respectively used as the corresponding inputs of the NARX neural network tension dynamic setting model of PSO and the PI controller-PSO BiLSTM neural network tension controller, the output of the tension sensor and the temperature sensor measuring the winding object are used as the corresponding input of the parameter detection module 4, the output of the speed sensor 3 and the temperature sensor measuring the winding object are used as the corresponding input of the parameter detection module 3, and the output of the parameter detection module 3 is used as the corresponding input of the TS neural network tension and speed decoupling setting model;

[0010] The error and error change rate of the output of the NARX neural network speed dynamic setting model of PSO and the output of the parameter detection module 1 are respectively used as the corresponding inputs of the NARX neural network speed dynamic setting model of PSO and the BiLSTM neural network speed controller of PI controller-PSO, the cumulative sum of the output of the BiLSTM neural network speed controller of PI controller-PSO and the output of the BiLSTM neural network tension controller of PI controller-PSO is used as the corresponding input of the DRNN neural network unwinding controller of PSO, the output of the DRNN neural network unwinding controller of PSO is respectively used as the input of the TDL beat delay device C and the unwinding device, the output of the TDL beat delay device C is used as the corresponding input of the DRNN neural network unwinding controller of PSO, the output of the speed sensor 1 and the temperature sensor of the unwinding device is used as the corresponding input of the parameter detection module 1, and the output of the parameter detection module 1 is respectively used as the TS neural network tension and speed decoupling device. The corresponding input of the fixed model and the corresponding input of the NARX neural network speed dynamic tracking model of PSO, the output of the NARX neural network speed dynamic tracking model of PSO is used as the input of the TDL beat delay device D, the output of the TDL beat delay device D is used as the corresponding input of the NARX neural network speed dynamic tracking model of PSO, the error and error change rate of the output of the NARX neural network speed dynamic tracking model of PSO and the output of the parameter detection module 2 are used as the corresponding input of the NARX neural network speed dynamic tracking model of PSO and the input of the PI controller-PSO BiLSTM neural network speed tracking controller, the output of the PI controller-PSO BiLSTM neural network speed tracking controller is used as the input of the winding device, the output of the speed sensor 2 and the temperature sensor of the winding device is used as the corresponding input of the parameter detection module 2, and the output of the parameter detection module 2 is used as the corresponding input of the TS neural network tension and speed decoupling setting model.

[0011] 2. Intelligent detection and control device of winding mechanism:

[0012] An intelligent detection and control device for a winding mechanism realizes intelligent detection and control of speed and tension parameters. The system includes a winding control device, a gateway node, an on-site monitoring terminal, a cloud platform, a remote monitoring terminal and a mobile phone App. The winding control device is constructed into a wireless communication network to realize a wireless communication network between winding control devices; the winding control device sends the detected operating parameters to the on-site monitoring terminal and the cloud platform through the gateway node, and the winding control device, the gateway node, the cloud platform, the on-site monitoring terminal, the remote monitoring terminal and the mobile phone App realize two-way transmission of the operating parameters and control information of the winding control device. The on-site monitoring terminal, the remote monitoring terminal and the mobile phone App realize intelligent monitoring of the operating parameters of the winding control device. The structure and function of an intelligent detection and control device for a winding mechanism can be seen in Figure 3 shown.

[0013] Compared with the prior art, the present invention has the following obvious advantages:

[0014] 1. The present invention adopts the AANN auto-associative neural network model - metabolic gray GM (1,1) model as the output of the AANN auto-associative neural network model as the input of the metabolic gray GM (1,1) model. The purpose of data compression and decompression of the AANN auto-associative neural network model is to capture the non-random information of the sensor input signal, thereby filtering out the random noise in the raw data output by the parameter measurement sensor. The output of the AANN auto-associative neural network model is used as the input of the metabolic gray GM (1,1) model. The metabolic gray GM (1,1) model can predict the parameter sensor output prediction value at a future moment according to the output of the AANN auto-associative neural network model. After the parameter sensor prediction value is predicted by the above method, the parameter sensor output prediction value is added to the original series of the parameter sensor output prediction value, and a data model is correspondingly removed at the beginning of the series, and then the parameter sensor output future value prediction is predicted. And so on, the parameter sensor output value is predicted. This method is called the metabolic gray GM (1,1) model to achieve longer-term predictions. The AANN autoassociative neural network model-metabolic gray GM (1,1) model can more accurately grasp the changing trend of the parameter sensor output value and improve the accuracy of predicting the parameter sensor output value.

[0015] 2. The DRNN neural network model of PSO and the DRNN neural network unwinding controller of PSO of the present invention are diagonal recurrent neural networks, and each layer of neurons only accepts feedback of its own output. However, compared with the traditional static multi-layer feedforward network, the dynamic nonlinear mapping of the DRNN neural network model and the DRNN neural network unwinding controller can still be realized, and have dynamic memory function. The DRNN neural network model and the DRNN neural network unwinding controller are a dynamic recurrent neural network with feedback and the ability to adapt to time-varying characteristics. The DRNN neural network model can more directly and vividly reflect the dynamic change performance of the input information and can more accurately predict the input information parameters.

[0016] 3. The present invention decomposes the historical data sequence output by the original AANN auto-associative neural network model-metabolism gray GM (1,1) model into components of different frequency bands through the EMD empirical mode decomposition model. Each component shows different characteristic information implicit in the original sequence to reduce the non-stationarity of the sequence. The high-frequency part of the data has a weak correlation and a relatively high frequency, representing the fluctuation component of the original sequence, and has a certain periodicity and randomness, which is consistent with the periodic changes output by the AANN auto-associative neural network model-metabolism gray GM (1,1) model; the low-frequency component represents the change trend of the original sequence output by the AANN auto-associative neural network model-metabolism gray GM (1,1) model. The EMD empirical mode decomposition model can decompose the fluctuation component, periodic component and trend component output by the AANN auto-associative neural network model-metabolism gray GM (1,1) model step by step. Each decomposed component contains the same deformation information, which reduces the mutual interference between different characteristic information to a certain extent, and the change curves of each decomposed component are smoother than the deformation sequence curve output by the original EMD empirical mode decomposition model. The EMD empirical mode decomposition model can effectively analyze the output deformation data of the AANN autoassociative neural network model-metabolic grey GM (1,1) model under the joint action of multiple factors. The decomposed components are conducive to better prediction of the BiLSTM neural network model of PSO. The output of the BiLSTM neural network model of PSO has higher prediction accuracy.

[0017] Fourth, the BiLSTM neural network model of the PSO of the present invention adds a backward LSTM neural network model on the basis of the forward LSTM neural network model, and the forward LSTM neural network model learns the time series data of the input information; while the backward LSTM neural network model is used to learn the time series data of the reverse input information. After the forward LSTM neural network model and the backward LSTM neural network model have completed data learning, the learning results of the forward and backward LSTM neural network models are superimposed to obtain the output of the BiLSTM neural network model of the PSO. The output of the BiLSTM neural network model of the PSO contains the feature information of the order of the input information of the time series and the reverse data. By using the bidirectional learning feature of the BiLSTM neural network model of the PSO, the connection between the sequential input and reverse input data is learned to improve the accuracy of the BiLSTM neural network model of the PSO in predicting the input information.

[0018] 5. The input of the PSO NARX neural network tension dynamic setting model, the PSO NARX neural network speed dynamic setting model and the PSO NARX neural network speed dynamic tracking model adopted by the present invention includes the input and output history feedback of the input information for a period of time. This part of the feedback input can be considered to include the setting model output history information for a period of time to participate in the setting model prediction of the input information, and for a suitable feedback time length, the accuracy of the setting model prediction is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A parameter detection module of the present invention;

[0020] Figure 2 It is a decoupling subsystem for winding mechanism detection and intelligent control of the present invention;

[0021] Figure 3 It is an intelligent detection and control device for the winding mechanism of the present invention;

[0022] Figure 4 A winding control device of the present invention;

[0023] Figure 5 A gateway node of the present invention;

[0024] Figure 6 This is a functional diagram of the on-site monitoring terminal software of the present invention. DETAILED DESCRIPTION

[0025] In order to better explain the present invention, so as to facilitate understanding, the following Figure 1-6 The technical scheme of the present invention is described in detail. The following examples are provided to explain the present invention, but the present invention is not limited to the following examples.

[0026] 1. Parameter detection module design

[0027] The parameter detection module includes TDL beat delay device, AANN autoassociative neural network model-metabolism grey GM (1,1) model, PSO BiLSTM neural network model and PSO DRNN neural network model;

[0028] 1. AANN autoassociative neural network model - metabolic grey GM (1,1) model design

[0029] The AANN auto-associative neural network model-metabolism grey GM (1,1) model is a series connection of the AANN auto-associative neural network model and the metabolism grey GM (1,1) model. The output of the AANN auto-associative neural network model is used as the input of the metabolism grey GM (1,1) model. The AANN auto-associative neural network model is a feedforward neural network. Its idea comes from principal component analysis and nonlinear principal component analysis. It is used to map the data in high-dimensional space to low-dimensional feature space under the premise of minimizing the input information loss. The input layer, mapping layer and bottleneck layer of the AANN auto-associative neural network model compress the input raw data into the low-dimensional feature space to complete the compression of the input data. The three-layer structure of the bottleneck layer, demapping layer and output layer demaps the data in the input data feature space to the output layer to complete the reconstruction of the input data. The bottleneck layer is the key to the AANN auto-associative neural network model. Its number of nodes is significantly less than that of other layers, and it completes the internal encoding and information compression. In order to prevent the formation of a simple one-to-one mapping between input data and output data, except for the linear excitation function of the output layer, other layers all use nonlinear excitation functions. During the training process of the AANN auto-associative neural network model, the input data and output data are consistent. The internal correlation of the input data is fitted and the interference of the parameter sensor output data is filtered out through information compression and recovery. The metabolic gray GM (1,1) model can output the predicted values ​​of multiple parameter sensor outputs at future moments according to the AANN auto-associative neural network model. After the multiple predicted values ​​of the sensor output are predicted by the above method, the multiple predicted values ​​of the sensor output are added to the original sequence of the sensor output, and the data modeling is removed at the beginning of the sequence accordingly, and then the prediction of the sensor output is predicted. And so on, the sensor value is predicted. This method is called the metabolic model, which can achieve a longer-term prediction of the sensor output.

[0030] 2. BiLSTM neural network model design of PSO

[0031] The LSTM neural network model predicts the information of the next moment based on the input time series information of the previous moment, and cannot combine the information of the future moment. In the input continuously changing signal predicted by the BiLSTM neural network model of PSO, the output of the BiLSTM neural network model of PSO at the current moment is not only related to the state of the previous input information, but also to the state of the future input information. It is necessary to input the previous input time series information and the future input time series information into the BiLSTM neural network model of PSO to achieve a more accurate prediction of the output of the EMD empirical mode decomposition model. The BiLSTM neural network model of PSO is superior to the LSTM neural network model of PSO in the corresponding time series tasks. The structure of the BiLSTM neural network model is the same as that of the standard LSTM neural network model. The BiLSTM neural network model is also composed of basic cell states, input gates, forget gates and output gates. The basic principle of the BiLSTM neural network model is that it consists of a forward standard LSTM neural network model and a backward standard LSTM neural network model. The BiLSTM neural network structure consists of a bidirectional recurrent neural network and a standard LSTM neural network model. The BiLSTM neural network contains LSTM neural network models in two training directions. The network nodes of each layer of the LSTM neural network model contain the past and future information output by the complete EMD empirical mode decomposition model. The two training directions are finally connected to the output layer. From the initial moment to the final moment, forward calculation is performed to obtain and save the output of the hidden layer at each moment. The output information in the backward layer is obtained by reverse calculation from the final moment to the initial moment and the output of the backward hidden layer at each moment is saved. The forward output at each moment is combined with the backward output and output as o at each moment. t , whose expressions are:

[0032]

[0033] is the output information of the previous moment in the forward layer of the BiLSTM neural network module, is the output information of the backward layer of the BiLSTM neural network model at the next moment; w1-w6 are the weights between the input and output information, and the PSO process of the BiLSTM neural network model of PSO:

[0034] ①. Initialize the particle swarm and build a BiLSTM neural network model with the corresponding parameters of each particle. Train the model through training data and use the root mean square error of the prediction result as the fitness value of each particle. ②. Use the fitness value as the termination condition. If the termination condition is met, return the optimal parameter value, otherwise return to step ①. ③. The particle swarm optimization algorithm updates the particles according to the fitness of the BiLSTM neural network model and determines the global optimal position of the particles, and builds the BiLSTM neural network model of PSO.

[0035] 3. Design of DRNN neural network model of PSO

[0036] The DRNN neural network model of PSO is a dynamic recurrent neural network with feedback and the ability to adapt to time-varying characteristics. The network can more directly and vividly reflect the dynamic change performance of input information and can accurately predict the value of input information. In the DRNN neural network model of PSO of the present invention, let I = [I1(t), I2(t), ..., I n (t)] is the network input vector, where I i (t) is the input of the ith neuron in the input layer of the PSO DRNN neural network model at time t, and the output of the jth neuron in the regression layer is X j (t), S j (t) is the sum of the inputs of the jth regression neuron, f(·) is the function of S, and O(t) is the output of the DRNN neural network model of PSO. Then the output of the DRNN neural network model of PSO is:

[0037]

[0038] The DRNN neural network model of PSO is a feedback network with local feedback characteristics. By storing internal states, it has the function of mapping dynamic features, so that the DRNN neural network model of PSO has the ability to adapt to time changes. The DRNN neural network model of PSO is basically similar to a 4-layer BP network, with an additional structural layer, which feeds back the output of the hidden layer to the input of the hidden layer through a delay link, thereby realizing partial feedback and achieving the effect of memorizing the previous state. This self-connection method of the DRNN neural network model of PSO makes it sensitive to historical state data. The addition of an internal feedback network increases the ability of the DRNN neural network model of PSO to process dynamic information and is conducive to the prediction of dynamic processes, thereby improving the prediction accuracy of input information. PSO process of the DRNN neural network model of PSO:

[0039] ①. Initialize the particle swarm and build a DRNN neural network model with the corresponding parameters of each particle. Train the model through training data and use the root mean square error of the prediction results as the fitness value of each particle. ②. Use the fitness value as the termination condition. If the termination condition is met, return the optimal parameter value, otherwise return to step ①. ③. The particle swarm optimization algorithm updates the particles according to the fitness of the DRNN neural network model and determines the global optimal position of the particles, and builds the DRNN neural network model of PSO.

[0040] 2. Design of decoupling subsystem for winding mechanism detection and intelligent control

[0041] The decoupling subsystem of winding mechanism detection and intelligent control includes TDL beat delay device, TS neural network tension and speed decoupling setting model, PSO's NARX neural network tension dynamic setting model, PI controller-PSO's BiLSTM neural network tension controller, PSO's NARX neural network speed dynamic setting model, PI controller-PSO's BiLSTM neural network speed controller, PSO's DRNN neural network unwinding controller, PSO's NARX neural network speed dynamic tracking model, PI controller-PSO's BiLSTM neural network speed tracking controller and parameter detection module.

[0042] 1. Design of TS neural network tension and speed decoupling setting model

[0043] ①. Antecedent network. The first layer is the input layer, which has n nodes. The second layer is the fuzzification layer, which fuzzifies the input data, and each neuron executes the corresponding membership function. The third layer is the fuzzy rule layer. The number of nodes in the fourth layer is m, and this layer implements normalized calculation. ②, consequent network. The first layer is the input layer, where the input value x0 of the 0th node is 1, and its function is to provide the constant term of the fuzzy rule consequent. The second layer has m nodes, and its function is to calculate the consequent of each rule:

[0044]

[0045] Layer 3 computes system output:

[0046]

[0047] Adjust the central value c of the membership function of the second layer through the network learning algorithm j and width b j and the connection weight p of the subsequent network jk For simplicity, the parameter p jkThe consequent of each rule is fixed, and the subsequent of each rule becomes the connection weight of one layer in the simplified structure. This simplified structure has the same structure as the fuzzy neural network of the conventional model, and the calculation results of the conventional model can be applied.

[0048] The TS neural network tension and speed decoupling setting model of the present invention adopts a parallel connection mode of a neural network and a fuzzy system, which share a common input, so that the two parts of the network do not interfere with each other and can process data in parallel, thereby improving the training speed of the network. The network simultaneously concentrates the advantages of the neural network and the fuzzy reasoning system, thereby enhancing the robustness and fault tolerance of the network, improving the approximation ability of the TS neural network tension and speed decoupling setting model, and greatly enhancing the computing power and generalization ability of the TS neural network tension and speed decoupling setting model, thereby obtaining the decoupling control value of the tension and speed output by the TS neural network tension and speed decoupling setting model.

[0049] 2. Design of PSO’s NARX neural network tension dynamic setting model

[0050] The NARX neural network tension dynamic setting model of PSO is a dynamic feedforward neural network with input nonlinear autoregressive network. It has a dynamic characteristic of a multi-step delay of input information and connects several layers of the closed network through feedback. The NARX neural network tension dynamic setting model of PSO is mainly composed of input layer, hidden layer, output layer and input and output delay. Before application, the delay order of input and output and the number of hidden layer neurons should be determined in advance. The current output of the NARX neural network tension dynamic setting model of PSO depends not only on the past output y(tn), but also on the current input vector X(t) and the delay order of the input vector, etc. The NARX neural network tension dynamic setting model of PSO is transmitted to the hidden layer through the delay layer, and the hidden layer processes the input signal and transmits it to the output layer. The output layer linearly weights the hidden layer output signal to obtain the final output signal of the NARX neural network tension dynamic setting model of PSO. The delay layer delays the output feedback signal of the NARX neural network tension dynamic setting model of PSO and the signal output by the input layer, and then transmits it to the hidden layer. x(t) represents the external input of the PSO NARX neural network tension dynamic setting model, m represents the delay order of the external input; y(t) is the output of the PSO NARX neural network tension dynamic setting model, n is the output delay order; s is the number of neurons in the hidden layer; thus, the h of the jth hidden unit can be obtained. i The output is:

[0051]

[0052] The output of the jth output layer node of the NARX neural network tension dynamic setting model of PSO is o j for:

[0053]

[0054] PSO process of the NARX neural network tension dynamic setting model:

[0055] A. Initialization of particle swarm algorithm parameters. First, determine the initial weights and thresholds of the NARX neural network tension dynamic setting model, and then determine the relevant parameters of the adaptive particle swarm algorithm. B. Calculate the adaptive weights. C. Take the minimum sum of squares of regression errors as the fitness, calculate and compare the fitness. D. Record the best position and global best position of each particle. E. The particle swarm optimization algorithm updates the particles according to the fitness of the NARX neural network tension dynamic setting model and determines the global optimal position of the particles, and constructs the NARX neural network tension dynamic setting model of PSO.

[0056] 3. Design of BiLSTM neural network tension controller based on PI controller-PSO

[0057] The PI controller-PSO BiLSTM neural network tension controller is a PI controller and a PSO BiLSTM neural network tension controller connected in series, the PI controller output is the PSO BiLSTM neural network tension controller input, the error and error change rate of the PSO NARX neural network tension dynamic setting model output and the parameter detection module 4 output are respectively used as the corresponding inputs of the PSO NARX neural network tension dynamic setting model and the PI controller-PSO BiLSTM neural network tension controller, and the design process of the PI controller-PSO BiLSTM neural network tension controller refers to the design method of the PSO BiLSTM neural network model of this patent.

[0058] 4. Design of NARX neural network speed dynamic setting model for PSO

[0059] The decoupling setting values ​​of tension and speed output by the TS neural network tension and speed decoupling setting model are respectively used as the corresponding inputs of the NARX neural network tension dynamic setting model of PSO and the NARX neural network speed dynamic setting model of PSO, the speed given value is used as the corresponding input of the NARX neural network speed dynamic setting model of PSO, the error and error change rate of the output of the NARX neural network speed dynamic setting model of PSO and the output of the parameter detection module 1 are used as the corresponding inputs of the NARX neural network speed dynamic setting model of PSO, and the design process of the NARX neural network speed dynamic setting model of PSO refers to the design process of the NARX neural network tension dynamic setting model of PSO in this patent.

[0060] 5. Design of BiLSTM neural network speed controller based on PI controller-PSO

[0061] The PI controller-PSO BiLSTM neural network speed controller is a PI controller and a PSO BiLSTM neural network speed controller connected in series, and the PI controller output is used as the input of the PSO BiLSTM neural network speed controller; the error and error change rate of the NARX neural network tension dynamic setting model output of the PSO and the parameter detection module 1 output are used as the input of the PI controller-BiLSTM neural network speed controller, and the cumulative sum of the PI controller-BiLSTM neural network speed controller output and the PI controller-BiLSTM neural network tension controller output is used as the corresponding input of the PSO DRNN neural network unwinding controller. The design process of the PI controller-PSO BiLSTM neural network speed controller refers to the PSO BiLSTM neural network model design method of this patent.

[0062] 6. Design of DRNN neural network unwinding controller based on PSO

[0063] The cumulative sum of the output of the BiLSTM neural network speed controller of the PI controller-PSO and the output of the BiLSTM neural network tension controller of the PI controller-PSO is used as the corresponding input of the DRNN neural network unwinding controller of the PSO. The output of the DRNN neural network unwinding controller of the PSO is used as the input of the TDL beat delay device C and the unwinding device respectively. The output of the TDL beat delay device C is used as the corresponding input of the DRNN neural network unwinding controller of the PSO. The design process of the DRNN neural network unwinding controller of the PSO refers to the design method of the DRNN neural network model of the PSO in this patent.

[0064] 7. Design of NARX neural network speed dynamic tracking model based on PSO

[0065] The output of parameter detection module 1 is used as the corresponding input of TS neural network tension and speed decoupling setting model and the corresponding input of NARX neural network speed dynamic tracking model of PSO. The output of NARX neural network speed dynamic tracking model of PSO is used as the input of TDL beat delay device D. The output of TDL beat delay device D is used as the corresponding input of NARX neural network speed dynamic tracking model of SO. The error and error change rate of the output of NARX neural network speed dynamic tracking model of PSO and the output of parameter detection module 2 are used as the corresponding input of NARX neural network speed dynamic tracking model of PSO and the input of PI controller-BiLSTM neural network speed tracking controller of PSO. The design of NARX neural network speed dynamic tracking model of PSO refers to the design method of NARX neural network tension setting model of PSO in this patent.

[0066] 8. Design of BiLSTM neural network speed tracking controller based on PI controller-PSO

[0067] The error and error change rate of the output of the NARX neural network speed dynamic tracking model of PSO and the output of the parameter detection module 2 are respectively used as the corresponding input of the NARX neural network speed dynamic tracking model of PSO and the input of the PI controller-PSO BiLSTM neural network speed tracking controller. The output of the PI controller-PSO BiLSTM neural network speed tracking controller is used as the input of the winding device. The design of the PSO BiLSTM neural network speed tracking controller of the PI controller-PSO BiLSTM neural network speed tracking controller refers to the design method of the PSO BiLSTM neural network model of this patent.

[0068] 2. Design of an intelligent detection and control device for winding mechanism

[0069] An intelligent detection and control device for a winding mechanism realizes intelligent detection and control of speed and tension parameters. The system includes a winding control device, a gateway node, an on-site monitoring terminal, a cloud platform, a remote monitoring terminal and a mobile phone App. The winding control device is constructed into a wireless communication network to realize a wireless communication network between winding control devices; the winding control device sends the detected operating parameters to the on-site monitoring terminal and the cloud platform through the gateway node, and the winding control device, the gateway node, the cloud platform, the on-site monitoring terminal, the remote monitoring terminal and the mobile phone App realize two-way transmission of the operating parameters and control information of the winding control device. The on-site monitoring terminal, the remote monitoring terminal and the mobile phone App realize intelligent monitoring of the winding control device. An intelligent detection and control device for a winding mechanism Figure 3 shown.

[0070] 1. Design of winding control device

[0071] The winding control device uses a 5G communication module to realize information interaction between the winding control device and the gateway node, on-site monitoring terminal, cloud platform, remote monitoring terminal and mobile phone APP. The winding control device includes an STM32 microprocessor, a 5G communication module, an unwinding device, a rewinding device, a camera, a temperature sensor, a speed sensor, a tension sensor, a position sensor and a signal conditioning circuit corresponding to the sensor; the software of the winding control device mainly realizes the collection, control and communication of the parameters of the winding control device. The software is designed in C language with high compatibility, which greatly improves the work efficiency of software design and development and enhances the reliability, readability and portability of the program code. See the structure of the winding control device. Figure 4 .

[0072] 2. Gateway node design

[0073] The gateway node includes a 5G communication module, an NB-IoT module, an STM32 microprocessor and a USB interface. The gateway node realizes wireless network communication with the winding control device through the 5G communication module. The NB-IoT module realizes two-way data interaction between the gateway node and the cloud platform. The USB interface connects to the on-site monitoring terminal to realize information interaction between the gateway node and the on-site monitoring terminal. The structure and function of the gateway node are shown in Figure 5 .

[0074] 3. Software design for on-site monitoring

[0075] The field monitoring terminal is an industrial control computer. It mainly realizes the intelligent collection and control of the operating parameters of the winding control device and realizes the information interaction with the winding control device and the gateway node. The main functions of the field monitoring terminal are communication parameter setting, data analysis and data management, and the decoupling subsystem of winding mechanism detection and intelligent control. The software functions of the field monitoring terminal are shown in Figure 6 .

[0076] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent detection and control method for a winding mechanism, characterized in that: The following steps are involved: Step 1: Build parameter detection module The parameter detection module includes TDL beat delay device, AANN autoassociative neural network model-metabolism grey GM (1,1) model, PSO BiLSTM neural network model and PSO DRNN neural network model; Step 2: Decoupling subsystem of winding mechanism detection and intelligent control The system includes a TDL beat delay device, a TS neural network tension and speed decoupling setting model, a PSO NARX neural network tension dynamic setting model, a PI controller-PSO BiLSTM neural network tension controller, a PSO NARX neural network speed dynamic setting model, a PI controller-PSO BiLSTM neural network speed controller, a PSO DRNN neural network unwinding controller, a PSO NARX neural network speed dynamic tracking model, a PI controller-PSO BiLSTM neural network speed tracking controller and a parameter detection module; The tension given value is used as the input of TDL beat delay device A and the corresponding input of PSO's NARX neural network tension dynamic setting model, the speed given value is used as the input of TDL beat delay device B and the corresponding input of PSO's NARX neural network speed dynamic setting model, the output of TDL beat delay device A and TDL beat delay device B are used as the corresponding input of TS neural network tension and speed decoupling setting model, the decoupling setting value of tension and speed output by TS neural network tension and speed decoupling setting model are used as the corresponding input of PSO's NARX neural network tension dynamic setting model and PSO's NARX neural network speed dynamic setting model, The error and error change rate of the output of the NARX neural network tension dynamic setting model of PSO and the output of the parameter detection module 4 are respectively used as the corresponding inputs of the NARX neural network tension dynamic setting model of PSO and the PI controller-PSO's BiLSTM neural network tension controller, the outputs of the tension sensor and temperature sensor measuring the winding object are used as the corresponding inputs of the parameter detection module 4, the outputs of the speed sensor 3 and the temperature sensor measuring the winding object are used as the corresponding inputs of the parameter detection module 3, and the output of the parameter detection module 3 is used as the corresponding input of the TS neural network tension and speed decoupling setting model.

2. According to claim 1, the intelligent detection and control method of winding mechanism is characterized in that: In the step 1, the parameter sensor output is used as the input of TDL beat delay device A, the outputs of the temperature sensor and TDL beat delay device A are used as the corresponding inputs of the AANN self-associative neural network model-metabolism gray GM (1,1) model respectively, the time series difference between the parameter sensor output and the output of the AANN self-associative neural network model-metabolism gray GM (1,1) model is used as the input of TDL beat delay device B and the corresponding input of the DRNN neural network model of PSO respectively, the output of the AANN self-associative neural network model-metabolism gray GM (1,1) model is used as the input of TDL beat delay device C, the output of TDL beat delay device C and the output of TDL beat delay device B are used as the corresponding inputs of the BiLSTM neural network model of PSO respectively, the output of the BiLSTM neural network model of PSO is used as the corresponding input of the DRNN neural network model of PSO, and the DRNN neural network model of PSO outputs the parameter sensor output prediction value.

3. The intelligent detection and control method for winding mechanism according to claim 1, characterized in that: In the step 2, the error and error change rate of the output of the NARX neural network speed dynamic setting model of the PSO and the output of the parameter detection module 1 are respectively used as the corresponding inputs of the NARX neural network speed dynamic setting model of the PSO and the PI controller-PSO BiLSTM neural network speed controller, the cumulative sum of the output of the PI controller-PSO BiLSTM neural network speed controller and the output of the PI controller-PSO BiLSTM neural network tension controller is used as the corresponding input of the DRNN neural network unwinding controller of the PSO, the output of the DRNN neural network unwinding controller of the PSO is respectively used as the input of the TDL beat delay device C and the unwinding device, the output of the TDL beat delay device C is used as the corresponding input of the DRNN neural network unwinding controller of the PSO, the output of the speed sensor 1 and the temperature sensor of the unwinding device is used as the corresponding input of the parameter detection module 1, and the output of the parameter detection module 1 is respectively used as the corresponding input of the TS neural network tension and speed decoupling setting model and the corresponding input of the NARX neural network speed dynamic tracking model of the PSO. The output of the NARX neural network speed dynamic tracking model of PSO is used as the input of the TDL beat delay device D, and the output of the TDL beat delay device D is used as the corresponding input of the NARX neural network speed dynamic tracking model of PSO. The error and error change rate of the output of the NARX neural network speed dynamic tracking model of PSO and the output of the parameter detection module 2 are used as the corresponding input of the NARX neural network speed dynamic tracking model of PSO and the input of the PI controller-PSO BiLSTM neural network speed tracking controller respectively. The output of the PI controller-PSO BiLSTM neural network speed tracking controller is used as the input of the winding device, and the output of the speed sensor 2 and the temperature sensor measuring the winding device are used as the corresponding input of the parameter detection module 2. The output of the parameter detection module 2 is used as the corresponding input of the TS neural network tension and speed decoupling setting model.

4. Intelligent detection and control device for winding mechanism, characterized in that: The detection and control device includes a winding control device, a gateway node, a field monitoring terminal, a cloud platform, a remote monitoring terminal and a mobile phone App. The winding control device sends the detected operating parameters to the field monitoring terminal and the cloud platform through the gateway node. The winding control device, the gateway node, the cloud platform, the field monitoring terminal, the remote monitoring terminal and the mobile phone App realize two-way transmission of the operating parameters and control information of the winding control device. The field monitoring terminal, the remote monitoring terminal and the mobile phone App realize intelligent monitoring of the operating parameters of the winding control device. The device is loaded with computer program steps that realize the intelligent detection and control method of the winding mechanism as described in any one of claims 1-3.

Images