Intelligent forecasting method, device and equipment for working condition indexes of high-pressure roller mill and storage medium

Abstract

The invention provides an intelligent forecasting method, device and equipment for working condition indexes of a high-pressure roller mill and a storage medium, and the intelligent forecasting method for the working condition indexes of the high-pressure roller mill comprises the steps: building an operation index dynamic model according to the dynamic characteristics of the working condition of the high-pressure roller mill and the change of the operation indexes, the operation index dynamic model comprises a mechanism model and deep learning; and combining identification errors of parameters of the mechanism model in the operation index dynamic model with non-mechanism model variables in the deep learning model to construct a nonlinear deep learning dynamic system and the like. In order to solve the problem that the working condition indexes of the high-pressure roller mill are difficult to forecast, the intelligent forecasting method for the working condition indexes of the high-pressure roller mill is provided by combining the mechanism model with a deep learning method based on big data, and the forecasting problem of the working condition indexes of the high-pressure roller mill is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent detection of mining machinery and equipment, and relates to an intelligent prediction method, device, equipment and storage medium for the working condition indexes of a high-pressure roller mill. Background Art

[0002] High-pressure roller mills play an important role in the ore dressing production process. In the ore dressing process, the large scale and complexity of mechanical equipment and the harsh production environment lead to strong nonlinearity and dynamics of the indexes characterizing the working condition changes. Therefore, it is crucial to accurately predict the indexes characterizing the working condition changes for realizing the stable operation of the ore dressing production process. Since the indexes characterizing the working condition changes show strong nonlinearity and strong coupling with the production process parameters, and the operation process of the equipment is affected by the change of raw materials, the establishment of the mechanism model cannot fully characterize the working condition changes of the equipment operation. At the same time, the mechanical equipment in the ore dressing process is in a dynamic change process during production, and the production process data generated changes in real time, resulting in the uncertainty of the prediction model with a complete information space at present and being unable to accurately predict the production indexes characterizing the working condition changes. Summary of the Invention

[0003] The purpose of the present invention is to provide an intelligent prediction method, device, equipment and storage medium for the working condition indexes of a high-pressure roller mill aiming at the problems existing in the prior art, and solve the problem that in the existing ore dressing process, when the prediction model is used for the mechanical equipment to characterize the working condition changes, the production indexes characterizing the working condition changes of the mechanical equipment cannot be accurately predicted.

[0004] To this end, the present invention adopts the following technical solutions:

[0005] An intelligent prediction method for the working condition indexes of a high-pressure roller mill includes the following steps:

[0006] Establish an intelligent prediction model for the working condition indexes by using the operation characteristics of the high-pressure roller mill equipment. The intelligent prediction model for the working condition indexes is a dynamic model between the mechanism model and the input and output data of the control system of the high-pressure roller mill. The dynamic model includes two parts: a mechanism model and a deep learning model;

[0007] Combine the identification error of the mechanism model parameters in the mechanism model of the high-pressure roller mill with the deep learning model in the dynamic model of the operation indexes into a key index prediction dynamic system;

[0008] Establish an online intelligent prediction model for the working condition indexes of the key index identification dynamic system by using adaptive deep learning;

[0009] The predicted value of the operating condition index is obtained from the output of the mechanism model in the intelligent prediction model of the operating condition index and the output of the online intelligent prediction model of the operating condition index of the non-linear deep learning dynamic system;

[0010] Preferably, the online intelligent prediction model of the operating condition index includes an online deep learning prediction model, a mechanism model and an adaptive mechanism;

[0011] The online deep learning prediction model is established using the Transformer architecture;

[0012] When the error between the output of the online deep learning prediction model and the target data is greater than the set threshold, the adaptive mechanism is used to adaptively adjust the weights and biases of the online deep learning prediction model;

[0013] Preferably, the training data used by the deep learning model is more than the actual prediction data used by the online deep learning prediction model; the operating condition index is the operating condition change index in the production process of the high-pressure roller mill, that is, the current of the fixed and moving rollers of the high-pressure roller mill.

[0014] Preferably, both the online deep learning prediction model and the deep learning adaptive mechanism include an input layer, a hidden layer, a fully connected layer and an output layer, where the number of hidden layers is N, N is a positive integer, and N>1;

[0015] Fix the weights and biases of the hidden layer in the online deep learning prediction model, and adaptively adjust the weights and biases of the fully connected layer online;

[0016] Online train the weights and biases in the adaptive mechanism model;

[0017] When the error between the output of the online deep learning prediction model and the target data is greater than the set threshold, the adaptive mechanism is used to replace the weights and biases of the hidden layer and the fully connected layer of the online deep learning prediction model with the weights and biases in the adaptive mechanism model.

[0018] An intelligent prediction device for the operating condition index of a high-pressure roller mill, comprising:

[0019] An operating condition index intelligent prediction model modeling module, used to establish an operating condition index intelligent prediction model using the characteristics of the production process control system of the high-pressure roller mill. The operating condition index intelligent prediction model is a dynamic model between the operating condition index and the input and output data of the mechanical equipment control system. The operating condition index intelligent prediction model includes two parts: a mechanism model and deep learning;

[0020] A dynamic system acquisition module for combining the identification error of the mechanism model parameters in the intelligent prediction model of the working condition indicators with the deep learning in the intelligent prediction model of the working condition indicators into a key index prediction dynamic system;

[0021] An online deep learning working condition index intelligent prediction model modeling module for establishing an online working condition index intelligent prediction model of the key index prediction dynamic system by using adaptive deep learning;

[0022] A prediction module for obtaining the predicted value of the working condition indicators from the output of the mechanism model in the intelligent prediction model of the working condition indicators and the output of the online working condition index intelligent prediction model of the deep learning dynamic system;

[0023] The online working condition index intelligent prediction model modeling module includes a prediction model modeling module, a mechanism model modeling module, and an adaptive module;

[0024] The prediction model modeling module uses the Transformer architecture to establish an online deep learning prediction model;

[0025] The mechanism model modeling module is constructed by using the mechanical and physical performance principles during the operation of the high-pressure roller mill equipment;

[0026] The adaptive module is used to, when the error between the output of the online deep learning prediction model and the target data is greater than a set threshold, adopt an adaptive mechanism to replace the weights and biases of the hidden layer and the fully connected layer of the online deep learning prediction model with the weights and biases in the adaptive mechanism model;

[0027] Preferably, both the online deep learning prediction model and the deep learning adaptive mechanism include an input layer, a hidden layer, a fully connected layer, and an output layer, where the number of hidden layers is N, N is a positive integer, and N>1;

[0028] The online working condition index intelligent prediction model modeling module fixes the weights and biases of the hidden layer in the online deep learning prediction model and adaptively adjusts the weights and biases of the fully connected layer online;

[0029] The calibration model modeling module trains the weights and biases in the adaptive mechanism model online;

[0030] Preferably, the training data used by the deep learning model is more than the actual prediction data used by the online deep learning prediction model; the high-pressure roller mill is a high-pressure roller mill, and the working condition indicators are the working condition change indicators during the production process of the high-pressure roller mill, namely the fixed and moving roller currents of the high-pressure roller mill.

[0031] An equipment for intelligent prediction of working condition indexes of a high-pressure roller mill, characterized in that the equipment comprises: an end-side sub-equipment, an edge-side sub-equipment and a cloud-side sub-equipment;

[0032] The end-side sub-equipment is used for collecting input data, output data in the production process of the high-pressure roller mill and equipment mechanical data required for equipment mechanism modeling;

[0033] The edge-side sub-equipment uses the online deep learning prediction model to perform online prediction of the working condition indexes;

[0034] The cloud-side sub-equipment is used for training the adaptive mechanism model and realizing the adaptability of the online prediction.

[0035] An upper computer-readable storage medium stores an upper computer program, characterized in that when the program is executed by a processor, it realizes a method for intelligent prediction of working condition indexes of a high-pressure roller mill.

[0036] The beneficial effects of the present invention are as follows:

[0037] Aiming at the problem that it is difficult to predict the working condition indexes of high-pressure roller mill equipment, the present invention combines the mechanism model with the deep learning intelligent prediction method based on big data, and uses the non-linear dynamic data representing the change of working condition indexes in the production operation process of high-pressure roller mill equipment to propose an intelligent prediction method for key indexes of the operating condition of high-pressure roller mill equipment combining the mechanism model with deep learning, realizing the prediction of key indexes representing the change of working condition in the production process of high-pressure roller mill. Description of the Drawings

[0038] Figure 1 It is a flowchart for realizing the intelligent prediction method of working condition indexes of high-pressure roller mill equipment according to an embodiment of the present invention;

[0039] Figure 2 It is a flowchart for realizing the intelligent prediction method of dynamic and fixed roller currents during the operation of a high-pressure roller mill according to an embodiment of the present invention;

[0040] Figure 3 It is a structural diagram of the Transformer network of the deep learning intelligent prediction algorithm according to an embodiment of the present invention;

[0041] Figure 4 It is a prediction result diagram of the dynamic roller current of a high-pressure roller mill according to an embodiment of the present invention;

[0042] Figure 5 It is a prediction result diagram of the fixed roller current of a high-pressure roller mill according to an embodiment of the present invention;

[0043] Figure 6 It is a schematic diagram of the prediction result error of the dynamic and fixed roller currents of a high-pressure roller mill according to an embodiment of the present invention;

[0044] Figure 7 Structural schematic diagram of the intelligent prediction device for the operating condition indexes of the high-pressure roller mill according to an embodiment of the present invention;

[0045] Figure 8 Structural schematic diagram of the intelligent prediction equipment for the operating condition indexes of the high-pressure roller mill according to an embodiment of the present invention. Specific embodiments

[0046] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0047] As Figure 1 Flowchart for realizing the intelligent prediction method for the operating condition indexes of the high-pressure roller mill equipment according to an embodiment of the present invention. The method includes the following steps:

[0048] Step 1: Establish an intelligent prediction model for the operating condition indexes by using the operating characteristics of the high-pressure roller mill equipment. The intelligent prediction model for the operating condition indexes includes a mechanism model and a deep learning model.

[0049] The establishment of the model in the specific embodiment is as Figure 2 shown. At this time, step 1 includes step A and step B.

[0050] Step A is to establish a mechanism model of the high-pressure roller mill. Specifically, by using the physical dynamic characteristics during the operation of the high-pressure roller mill equipment, a mechanism model for the operation of the high-pressure roller mill equipment is constructed according to the mechanism relationship of the equipment operation, and the parameters of the mechanism model are identified to obtain the identification error.

[0051] Specifically, according to the basic equation of grinding kinetics:

[0052]

[0053] In the formula, is the reduction rate of the coarse fraction particle size per unit time, i.e., the grinding speed; R is the percentage of the content of the ground and medium-coarse fractions after the grinding time t; k is a proportionality coefficient related to the particle size of the grinding product. Integrating the above formula gives:

[0054]

[0055] It is obtained that lnR=-kt + C, where C is a constant in the grinding process; when t = 0, at this time R = R0, substituting it into the above formula gives:

[0056] lnR=-kt + lnR0 (3)

[0057] R = R0e -kt (4)

[0058] Therefore, it can be seen from Equation (4) that the change in grinding particle size is directly proportional to the grinding speed.

[0059] From the mathematical relationship between current and speed, it can be obtained that:

[0060]

[0061] where U is the voltage, I t is the current value at time t, and cosα is the power factor; it can be seen from Equation (5) that at time t, the grinding current directly affects the particle size of the final product.

[0062] Step B: Establish a dynamic model between the operating condition indicators and the input and output data of the control system during the production process of the high-pressure roller mill; specifically, adopt the dynamic model of the production control system of the high-pressure roller mill, and utilize the characteristics of the production process control system of the equipment to control the control system parameters and variables of the equipment production process within a certain range, and construct a prediction algorithm for the operating condition indicators of the high-pressure roller mill equipment based on deep learning.

[0063] Specifically, the production process data of the high-pressure roller mill in the concentrator not only includes mechanism data such as pressure and roll gap, but also includes more than thirty-dimensional process data such as valve opening, frequency, bearing temperature, buffer bin level, moving roll current, and fixed roll current. The above process data is used as the input and output data of the established high-pressure roller mill operating condition identification and prediction model.

[0064] By establishing a corresponding quantitative relationship between the input and output data, the original process data can be represented by the matrix H m×n as follows:

[0065]

[0066] where m > n, and m represents the number of sampling points of the input data, and n represents the number of variables of the input data; performing singular value decomposition on the above m×n order Hankel matrix, we can obtain:

[0067] H m×n = UΣV T (7)

[0068] where U is an m×m order orthogonal matrix that satisfies U×U T = E; V is an n×n order orthogonal matrix that satisfies V×V T = E; and Σ is an m×n order matrix:

[0069]

[0070] Λ = diag(λ1, λ2, …, λ l ) λ1 ≥ λ2 ≥ … ≥ λ l , and l is the rank of the Hankel matrix.

[0071] The process of performing singular value decomposition on the above-mentioned m×n order Hankel matrix is as follows:

[0072] First, reserve the first r (r < l) effective singular values of the diagonal matrix, and then set the remaining singular values to 0. By using the inverse process of SVD, a new matrix is obtained, which solves the problem that the current singular value selection method is prone to losing the eigenvectors of mutations, and at the same time, it is impossible to fix and effectively select the number of singular values to reconstruct the original signal. Therefore, only the first i singular values are retained, and the remaining singular values are all set to 0, thus obtaining equations (9) and (10).

[0073]

[0074] H′ i = UΛ′ i V T (10)

[0075] The new Hankel matrix H′ is calculated through formula (10) i , and the original signal is reconstructed into signal X through equation (10) i (i = 1, 2, …, k), and k < n.

[0076] Step 2: Combine the identification error of the mechanism model parameters with the deep learning model as a key index to predict the dynamic system.

[0077] Specifically, the error of identifying parameters using the mechanism model is combined with the deep learning model of the parameters of the high-pressure roller mill production process control system as a key index intelligent prediction system.

[0078] When the error between the output of the online deep learning prediction model and the target data is greater than the set threshold, an adaptive mechanism is adopted to adaptively adjust the weights and biases of the online deep learning intelligent working condition index prediction model.

[0079] As Figure 3 shown, the above-mentioned reconstructed signal is input into the Encoder model of the Transformer architecture, where the Encoder is composed of N identical units. Each unit is composed of two sub-units, namely the multi-head attention mechanism and the fully connected layer; among them, each sub-unit is added with a residual connection and normalization, so the output of the sub-unit can be expressed as:

[0080] sub_layer_output = LayerNorm(x + (SubLayer(x))) (11)

[0081] According to Google's definition, Attention is defined as follows:

[0082]

[0083] Among them, From the perspective of the dimension information of the matrix, it can be considered that Attention encodes a sequence Q of m×d k into a new sequence of n×d v . And K and V are in one-to-one correspondence. Looking at a single vector in sequence Q, it can be expressed as:

[0084]

[0085] where t ∈ (0, 1, …, m), and Z is the normalization factor of the softmax function; it can be seen from the above formula that each q t is encoded into a weighted sum of v1, v2, …, v n , and the weight of v s depends on the inner product (dot product) of q t and k s . The scaling factor plays a certain regulatory role to avoid the gradient of the softmax function being very small when the inner product is large. On this basis, Multi-head Attention is defined as follows:

[0086] MultiHead(Q, K, V) = Concat(head1, head2, … head h ) (14)

[0087] head i = Attention(Q i , K i , V i ) (15)

[0088]

[0089] Among them, and

[0090] are calculated h times repeatedly, and the results of the h Attention are concatenated, and finally a sequence is output.

[0091] The feature vector v i (i = 1, …, q) of the process data is obtained through the spatial attention weight vector to obtain the feature matrix v = [h1, …, h p, in order to obtain the dynamic and timely characteristics of the conversion process data, a time sample attention mechanism is introduced in the Decoder to adaptively determine the relevant hidden states generated by the Decoder at all time instants. The measurement can refer to the previous Decoder hidden states. Each Decoder hidden state is assigned a time attention value. Then, an adaptively weighted content vector is obtained as the input of the Decoder. In this way, this attention mechanism breaks the limitation of the traditional Encoder-Decoder structure that depends on fixed-length vectors during encoding and decoding. In the time dimension, there is a correlation between the working conditions in different time periods, and the correlation is also different in different situations. The attention mechanism is used to adaptively assign different weights to the data. Therefore, the time attention value of the hidden state at time t can be calculated as follows:

[0092]

[0093] where obtained from equations (20) and (21):

[0094]

[0095] In the formula, is the output result of the (i - 1)-th Multi-head Attention unit, and u is the set Decoder feature scale. Thus, the time feature matrix of the corresponding process variable is obtained. Finally, the obtained time feature matrix is decoded by the Decoder method, and the prediction result is finally obtained.

[0096] Step 4: Establish an online intelligent prediction model for the working condition index of the adaptive high-pressure roller mill by using adaptive deep learning.

[0097] Specifically, both the online deep learning prediction model and the deep learning adaptive mechanism include an input layer, a hidden layer, a fully connected layer, and an output layer. Among them, the number of layers of the hidden layer is N, N is a positive integer, and N > 1;

[0098] Fix the weights and biases of the hidden layer in the online deep learning prediction model, and adaptively adjust the weights and biases of the fully connected layer online; train the weights and biases in the adaptive mechanism model online; when the error between the output of the online deep learning prediction model and the target data is greater than the set threshold, use the adaptive mechanism to replace the weights and biases of the hidden layer and the fully connected layer of the online deep learning prediction model with the weights and biases in the adaptive mechanism model.

[0099] Step Five: Obtain the predicted value of the operating index from the output of the identifiable model in the operating index dynamic model and the output of the online intelligent prediction model of the unknown nonlinear dynamic system.

[0100] In one embodiment, the intelligent prediction method for the operating condition indexes of a high-pressure roller mill can be used for the intelligent prediction of the dynamic and static roller currents, which are the operating condition indexes of the high-pressure roller mill in a beneficiation plant.

[0101] As Figures 4 to 6 shown, by using the above-mentioned prediction method for the dynamic and static roller currents, which are the characterization indexes of the operating conditions of the high-pressure roller mill, the prediction accuracy of the dynamic roller current is 91.67%, and the prediction accuracy of the static roller current is 92.83%. Moreover, within the error range of ±0.5%, the prediction accuracy rate of the dynamic and static roller currents reaches 99.9%, meeting the prediction accuracy requirements for the operating condition indexes of the high-pressure roller mill equipment during operation.

[0102] In one embodiment, as Figure 7 shown, an intelligent prediction device for the operating condition indexes of a high-pressure roller mill equipment is provided, including: a mechanism model parameter identification module, an adaptive module, a deep learning prediction algorithm module, an online intelligent prediction system modeling module, and an operating condition index prediction module, where:

[0103] The operating condition index intelligent prediction model modeling module is used to establish an operating condition index intelligent prediction model by using the characteristics of the production process control system of the high-pressure roller mill. The operating condition index intelligent prediction model is a dynamic model between the operating condition indexes and the input and output data of the mechanical equipment control system. The operating condition index intelligent prediction model includes two parts: a mechanism model and deep learning.

[0104] The dynamic system acquisition module is used to combine the identification error of the mechanism model parameters in the operating condition index intelligent prediction model with the deep learning in the operating condition index intelligent prediction model into a key index prediction dynamic system.

[0105] The online deep learning operating condition index intelligent prediction model modeling module is used to establish an online operating condition index intelligent prediction model of the key index prediction dynamic system by using adaptive deep learning.

[0106] The prediction module is used to obtain the predicted value of the operating condition index from the output of the mechanism model in the operating condition index intelligent prediction model and the output of the online operating condition index intelligent prediction model of the deep learning dynamic system.

[0107] In one of the embodiments, the intelligent prediction device for the operating condition indexes of the high-pressure roller mill equipment is used for the intelligent prediction of the operating condition indexes of the high-pressure roller mill.

[0108] For the specific limitations of the intelligent prediction device for industrial process operation indicators, reference can be made to the limitations of the intelligent prediction method for industrial process operation indicators in the above text, which will not be elaborated here. Each module in the above intelligent prediction device for industrial process operation indicators can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of the host device in hardware form or be independent of it, or can be stored in the memory of the host device in software form, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0109] In one embodiment, as Figure 8 shown, there is provided a high-pressure roller mill condition indicator intelligent prediction device for implementing the high-pressure roller mill condition indicator intelligent prediction method in each of the above embodiments, including:

[0110] An edge-side sub-device, a cloud-side sub-device; the edge-side sub-device is used to collect the mechanical signals of the high-pressure roller mill device and the input and output data during the production process;

[0111] The edge-side sub-device uses the online deep learning prediction model to perform online prediction of the condition indicators;

[0112] The cloud-side sub-device is used to train the deep learning adaptive model and implement the adaptive mechanism.

[0113] In one embodiment, there is provided a host-readable storage medium storing a host program, and when the program is executed by a processor, it implements the high-pressure roller mill device condition indicator intelligent prediction method in each of the above embodiments.

[0114] In summary, the high-pressure roller mill device condition indicator intelligent prediction method, device, and equipment proposed in the embodiments of the present invention are directed at the existing model-based prediction method and deep learning method, which cannot accurately predict the production indicators representing the working conditions due to the real-time changes in the production process data generated by the working conditions, resulting in the uncertainty of the existing prediction model with a complete information space.

Claims

1. An intelligent prediction method for operating condition indicators of a high pressure roller grinding mill, characterized in that: The steps include: Establish an intelligent prediction model for operating condition indicators, which includes adopting a dynamic model of the production control system of the high-pressure roller mill equipment, taking advantage of the characteristics of the equipment production process control system that the control system parameters and variables are controlled within a certain range, and constructing a prediction algorithm for the operating condition indicators of the high-pressure roller mill equipment based on deep learning; The intelligent prediction model of operating condition indicators is a dynamic model between the mechanism model and the input and output data of the high pressure roller grinding mill control system, which includes the mechanism model and the deep learning model; Using the mechanism characteristics of the high pressure roller grinding machine, a mechanism model of the equipment operation process is established; Identify the parameters of the mechanism model and obtain the identification error; Combine the mechanism model identification error and deep learning model into a dynamic system for key indicator prediction; Adaptive deep learning is used to establish an online operating condition indicator intelligent prediction model for the key indicator prediction dynamic system, which obtains the predicted value of the operating condition indicator from the output of the mechanism model and the output of the online operating condition indicator intelligent prediction model; The online operating condition index intelligent prediction model includes online deep learning prediction model, mechanism model and adaptive mechanism model; An online deep learning prediction model is established using the Transformer architecture of the deep learning algorithm. When the error between the output of the online deep learning prediction model and the target data is greater than the set threshold, an adaptive mechanism model is used to adjust the weight and bias of the online deep learning prediction model. Among them, the training data used by the deep learning model is more than the actual forecast data used by the online deep learning forecast model.

2. The intelligent prediction method for operating condition indicators of a high pressure roller grinding mill according to claim 1 is characterized in that: The online deep learning prediction model and the adaptive mechanism model both include an input layer, a hidden layer, a fully connected layer and an output layer, wherein the number of hidden layers is N, N is a positive integer, and N>1; Fix the weights and biases of the hidden layers in the online deep learning prediction model, and adapt the weights and biases of the fully connected layers online; train the weights and biases in the adaptive mechanism model online; When the error between the output of the online deep learning prediction model and the target data is greater than a set threshold, an adaptive mechanism model is used to replace the weights and biases of the hidden layer and the fully connected layer in the online deep learning prediction model with the weights and biases in the adaptive mechanism model.

3. An intelligent forecasting device for operating condition indicators of a high pressure roller mill, characterized in that: The device comprises: The modeling module of the intelligent prediction model of the working condition index is used to establish the intelligent prediction model of the working condition index by utilizing the characteristics of the production process control system of the high-pressure roller mill. The intelligent prediction model of the working condition index is a dynamic model between the working condition index and the input and output data of the mechanical equipment control system. The intelligent prediction model of the working condition index includes a mechanism model and a deep learning model; Dynamic system acquisition module, used to combine the identification error of mechanism model parameters with the deep learning model into a key indicator prediction dynamic system; The online deep learning working condition indicator intelligent prediction model modeling module is used to establish an online working condition indicator intelligent prediction model of the key indicator prediction dynamic system using adaptive deep learning; A prediction module, used to obtain the predicted value of the operating condition index from the output of the mechanism model and the output of the online operating condition index intelligent prediction model; Online working condition index intelligent forecasting model modeling module, including forecasting module modeling module, mechanism model modeling module and adaptive module; The forecast model building module uses the Transformer architecture to build an online deep learning forecast model; The mechanism model modeling module is constructed using the mechanical and physical performance principles of the high-pressure roller mill equipment operation process; The adaptive module is used to replace the weights and biases of the hidden layer and the fully connected layer in the online deep learning prediction model with the weights and biases in the adaptive mechanism model when the error between the output of the online deep learning prediction model and the target data is greater than the set threshold.

4. The intelligent forecasting device for operating condition indicators of a high pressure roller grinding mill according to claim 3 is characterized in that: The online deep learning prediction model and the adaptive mechanism model both include an input layer, a hidden layer, a fully connected layer and an output layer, wherein the number of hidden layers is N, N is a positive integer, and N>1; The forecast model building module fixes the weights and biases of the hidden layers in the online deep learning forecast model, and adapts the weights and biases of the fully connected layers online; A correction model modeling module is also included, which is used to online train the weights and biases in the adaptive mechanism model.

5. An intelligent prediction device for high pressure roller mill operating condition indicators for implementing the method described in claim 1 or 2, characterized in that: The device includes: a terminal side sub-device, an edge side sub-device and a cloud side sub-device; The terminal side sub-device is used to collect input data and output data in the production process of the high pressure roller grinding mill and equipment mechanical data required for equipment mechanism modeling; The edge-side sub-device uses an online deep learning prediction model to perform online prediction of the operating condition index; The cloud-side sub-device is used to train the adaptive mechanism model and realize the adaptation of the online forecast.

6. A host computer readable storage medium storing a host computer program, characterized in that: When the program is executed by a processor, the method according to claim 1 or 2 is implemented.

Images