A control system and control method for a wood defibration plant

Abstract

The application relates to a control system of wood defibration equipment, comprising a detection module, a machine learning module and a control module. The detection module comprises a pressure detection unit, a thickness detection unit, a width detection unit, a surface feature extraction unit and a moisture content detection unit. The machine learning module is used to establish a mapping relationship between input data and output data. The input data comprises data detected by the signal detection module. The output data comprises the pressure between the defibration roller and the veneer, the height of the defibration roller, the rotating speed of the defibration roller and the press roller, and the nip between the defibration roller and the press roller. The control module is used to control the parameters of the defibration roller and the press roller according to the output data, so that the defibration unit is within the range of the defibration degree required by production. The application can quantize the key technical indexes in the veneer defibration process, help improve the production efficiency, improve the defibration processing precision, and help improve the defibration uniformity in the veneer defibration production.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wood defibration technology, and in particular to a control system and a control method of a wood defibration equipment. BACKGROUND

[0002] Wood reconstituted material is a new type of biomass composite material which is made of biomass materials such as fast-growing forest wood and desert shrubs through directional reconstitution and compounding. After nearly 20 years of continuous research and development, the key technologies and industrialization of reconstituted wood manufacturing have made significant breakthroughs, and large-scale industrial production has been realized in China. Defibration equipment is a key equipment for preparing wood bundles in the production of reconstituted wood, which is used for rolling, cutting and directional fiber separation of single boards. The defibration equipment mainly adjusts the gap between the defibration roll teeth to produce a series of point or line segment cracks on the surface of the single board, so as to prepare a fiberized wood bundle single board with unbroken longitudinal fibers and uniform thickness.

[0003] The current defibration equipment is in the stage of mechanization and semi-automation. In the prior art, the mechanical screw is generally adjusted by experience, and the defibration pressure and effect of the single board are adjusted by changing the roll gap of the upper and lower rolls. However, this method lacks quantitative indexes of key technical parameters, which causes low production efficiency, defibration processing precision and uniformity. SUMMARY

[0004] In order to overcome the problem that the key technical parameters of the existing wood defibration equipment lack quantitative indexes, resulting in low production efficiency, defibration processing precision and uniformity, on the one hand, the present application provides a control system of a wood defibration equipment.

[0005] A control system of a wood defibration equipment, comprising a detection module, a machine learning module and a control module; the detection module comprises a pressure detection unit, a thickness detection unit, a width detection unit, a surface feature extraction unit and a moisture content detection unit; the pressure detection unit is used for real-time detection of the pressure between the defibration roller and the veneer; the thickness detection unit is used for detection of the thickness of the veneer and the defibration unit; the width detection unit is used for detection of the width of the veneer and the defibration unit; the surface feature extraction unit is used for extraction of the surface features of the defibration unit and evaluation of the defibration quality of the defibration unit; the moisture content detection unit is used for real-time detection of the moisture content of the veneer; the machine learning module is in communication connection with the pressure detection unit, the thickness detection unit, the width detection unit and the moisture content detection unit; the machine learning module is used for establishment of a mapping relationship between input data and output data; wherein the input data comprises the data detected by the signal detection module; the output data comprises the pressure between the defibration roller and the veneer, the height of the defibration roller, the rotating speed of the defibration roller and the pressure roller, and the nip between the defibration roller and the pressure roller; the machine learning module is in communication connection with the control module; the control module is used for control of the lifting of the defibration roller, adjustment of the pressure between the defibration roller and the veneer, and adjustment of the rotating speed of the defibration roller and the pressure roller according to the output data, so that the defibration unit is within the range of the required defibration degree.

[0006] Optionally, the pressure detection unit comprises a pressure sensor; the width detection unit and the surface feature extraction unit each comprise an industrial camera; the thickness detection unit comprises a photoelectric sensor; and the moisture content detection unit is a contact moisture analyzer or a non-contact moisture analyzer.

[0007] Optionally, the surface feature extraction unit has a pre-processing filter subunit; the pre-processing filter subunit is used for reduction of edge unevenness in feature extraction.

[0008] Optionally, the moisture content detection unit is a contact moisture analyzer; the contact moisture analyzer is driven by a mechanical arm, used for moisture content detection of multiple sites on the veneer and averaging to obtain the moisture content of the veneer.

[0009] Optionally, an optimization module is further included; the optimization module is in communication connection with the detection module and the machine learning module, used for real-time tracking and recording of the data detected by the detection module, and optimization of the mapping relationship in the machine learning module.

[0010] Optionally, a man-machine interaction module is further included; the man-machine interaction module is in communication connection with the detection module and the control module, and has a display unit and an operation unit; the display unit is used for displaying the data detected by the detection module; and the operation unit is used for inputting instructions by a user and transmitting the instructions to the control unit, so as to control the wood defibration equipment by the user.

[0011] Optionally, the mapping relationship includes:

[0012]

[0013] F = γ·g + δ(g min ≤ g ≤ g max );

[0014] In the formula, F is the pressure between the defibration roller and the veneer; d s is the lifting distance of the lead screw lifter for driving the defibration roller to lift; g is the gap size between the defibration roller and the press roller; and α, β and γ are all coefficients, which are all fixed values measured by experiments.

[0015] The mapping relationship further includes:

[0016]

[0017] In the formula, R scrimber is the defibration degree of the defibration unit. H A is the thickness of the defibration unit. H P is the thickness of the veneer. W P is the width of the defibration unit. W A is the width of the veneer. R compression is the compression rate of the defibration unit. θ, η and ω are all coefficients, which need to be adaptively adjusted according to the modeling data and test results of the machine learning regression model.

[0018] wherein,

[0019]

[0020] In the formula, R compression is the compression rate of the defibration unit. H P is the thickness of the veneer. g is the gap size between the defibration roller and the press roller.

[0021] On the other hand, based on the above idea, the application further provides a control method of the wood defibration equipment, which comprises:

[0022] S1, acquiring training data; the training data includes input data and output data;

[0023] The input data includes the width of the single board, the width of the defibration unit, the thickness of the single board, the surface characteristics of the single board, the moisture content of the single board, and environmental data; the environmental data includes temperature, humidity;

[0024] The output data includes the pressure between the defibration roller and the single board, the height of the defibration roller, the rotation speed of the defibration roller and the compression roller, and the nip between the defibration roller and the compression roller;

[0025] S2, training a machine learning regression model; inputting the training data into the machine learning regression model to establish a mapping relationship between the input data and the data;

[0026] S3, before defibrating the single board, obtaining the width, thickness, moisture content of the single board to be defibrated and the environmental data, and inputting them into the machine learning regression model;

[0027] S4, based on the mapping relationship, obtaining the pressure between the defibration roller and the single board, the height of the defibration roller, the rotation speed of the defibration roller and the compression roller, and the nip between the defibration roller and the compression roller required for defibrating the single board to be defibrated, and controlling the wood defibration equipment to make corresponding adjustments.

[0028] Optionally, in step S1, the training data is divided into a training set and a test set according to a ratio of 8:2; the training set is used to train the machine learning regression model; and the test set is used to evaluate the performance of the machine learning regression model

[0029] Optionally, in step S2, during the training of the machine learning regression model, the defibration quality of the defibration unit is evaluated, specifically including:

[0030] S201, obtaining a sample of a defibration unit surface image; the sample of the defibration unit surface image is obtained based on multiple repeated experiments on different pressures between the defibration roller and the defibration unit and different types of defibration units;

[0031] S202, preprocessing the sample of the defibration unit surface image;

[0032] S203, processing the sample of the defibration unit surface image by horizontal mirror flipping, translation, scaling, cropping, and rotation to increase the number of samples;

[0033] S204, data cleaning is performed on the samples obtained in step S203 to obtain an image data set; the surface characteristics in the defibration unit surface image in the image data set are labeled by a labeling tool;

[0034] S205, construct a deep learning network, and train through the image data set; the deep learning network is used to extract the surface features of the defibration unit based on the surface image of the defibration unit, and then evaluate the defibration quality of the defibration unit.

[0035] As described above, the control system of the wood defibration equipment has at least the following beneficial effects:

[0036] The control system of the wood defibration equipment of the present application first detects the parameters before and after the veneer defibration through the detection module, then inputs the detected parameters as input data into the machine learning module for training, and establishes the mapping relationship between the input data and the output. Finally, when the wood defibration equipment is running, the parameters of the veneer to be defibrated are input into the machine learning module as input data, and the pressure between the press roll and the veneer, the height of the press roll, the speed of the press roll and the nip between the two press rolls suitable for the defibration of the veneer to be defibrated are obtained. The defibration unit is in the defibration degree range required by production. The system makes the key technical indicators in the veneer defibration process quantified, thereby helping to improve the production efficiency, improve the precision of defibration processing, and help to improve the defibration uniformity in the veneer defibration production. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is the architecture diagram for embodying the control system architecture of the wood defibration equipment according to the embodiment of the present application.

[0038] Figure 2 is the flowchart for embodying the control method of the wood defibration equipment according to the embodiment of the present application. DETAILED DESCRIPTION

[0039] The embodiments of the present application will be described below through specific concrete examples, and those skilled in the art can easily understand other advantages and effects of the present application from the disclosure of the present specification. The present application can also be implemented or applied by means of other different specific embodiments, and the details in the present specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. It should also be understood that the terms used in the embodiments of the present application are for describing specific specific embodiments, and are not intended to limit the protection scope of the present application. The test methods in the following embodiments are not specified, and are usually performed according to the conventional conditions or the conditions recommended by the manufacturers.

[0040] It is to be understood that the structures, proportions, sizes, etc. shown in the drawings accompanying the present specification are merely intended to facilitate the understanding of the content disclosed in the present specification for those skilled in the art to understand and read, and are not intended to limit the defined conditions under which the present application can be implemented, and therefore do not have technical significance. Any modification of the structure, change of the proportional relationship or adjustment of the size, without affecting the effects that can be produced by the present application and the purposes that can be achieved, should still fall within the scope of the technical content disclosed by the present application. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" used in the present specification are only for the purpose of clear understanding of the description, and are not intended to limit the scope of the present application, and the change or adjustment of the relative relationship, without substantially changing the technical content, is also considered as the scope of the present application.

[0041] It is also necessary to explain in advance that the defibration unit is a single board after defibration. The wood defibration equipment defibrates the wood through defibration rollers and pressure rollers. The defibration roller is located directly above the pressure roller and can be adjusted in height by a screw rod lifter. The defibration roller is provided with equally spaced defibration teeth, the tooth depth can be designed according to the needs, and the defibration teeth have cutting edges, and each circle of defibration blades is ladder-shaped along the circumferential direction. The outer surface of the pressure roller is provided with a plurality of spherical protruding units arranged around the annular surface, or a protruding unit densely covered with a mesh pattern. The single board is separated along the longitudinal weakest link of the fiber structure under the action of the defibration roller pressure and the defibration tooth cutting force.

[0042] Please refer to Figure 1 The present application discloses a control system of a wood defibration equipment, which comprises a detection module, a machine learning module and a control module. The detection module comprises a pressure detection unit, a thickness detection unit, a width detection unit, a surface feature extraction unit and a moisture content detection unit.

[0043] The pressure detection unit is used for real-time detection of the pressure between the defibration roller and the single board. The pressure detection unit can use a pressure sensor. The thickness detection unit is used for detecting the thickness of the single board. The thickness detection unit can use a photoelectric sensor.

[0044] The width detection unit is used for detecting the thickness of the single board and the defibration unit. The surface feature extraction unit is used for extracting the surface features of the defibration unit and evaluating the defibration quality of the single board. The width detection unit and the surface feature extraction unit can both be industrial cameras. The moisture content detection unit is used for real-time detection of the moisture content of the single board, and can use a contact moisture analyzer or a non-contact moisture analyzer.

[0045] The machine learning module is in communication connection with the pressure detection unit, the thickness detection sensor, the width detection sensor and the moisture content detection sensor. The machine learning module is used for establishing a mapping relationship between input data and output data.

[0046] The input data includes data detected by the signal detection module. The output data includes the pressure between the defibrator roller and the veneer, the height of the defibrator roller, the rotation speed of the defibrator roller and the press roller, and the nip between the defibrator roller and the press roller. The machine learning module is in communication connection with the control module.

[0047] The control module is configured to control the press roller lifting, adjust the pressure between the defibrator roller and the veneer, and adjust the rotation speed of the defibrator roller and the press roller, so that the defibrator unit is within the range of the required defibration degree.

[0048] The implementation principle of the control system of the wood defibration equipment is that the detection module detects various parameters of the veneer and the defibrator unit first, and then inputs the parameters as input data into the machine learning module for training, so that the machine learning module establishes a mapping relationship between the input data and the output data. Before defibrating the veneer, the trained machine learning module receives the input data corresponding to the veneer to be defibrated, and obtains the corresponding output data through the mapping relationship. According to the output data, the control module controls the defibrator roller lifting, adjusts the pressure between the defibrator roller and the veneer, the rotation speed of the defibrator roller and the press roller, and the nip between the defibrator roller and the press roller, so that the defibration degree of the defibrator unit is within the range of the required defibration degree.

[0049] It should be noted that the defibration degree is related to the compression rate and water absorption rate of wood, and can evaluate the defibration degree of the defibrator unit. The defibration degree can be used to represent the defibration effect of the veneer, such as the degree of fiber breakage, the softness and hardness of the veneer after defibration, etc.

[0050] The parts of the control system of the wood defibration equipment will be described in detail as follows:

[0051] When the industrial camera is used as the width detection unit to measure the width of the defibrator unit, the edge of the defibrator unit may be uneven, so the image collected by the industrial camera needs to be preprocessed by a filter to reduce the unevenness of the edge in the image.

[0052] Specifically, the surface feature extraction unit has a pre-processing filter subunit. The pre-processing filter subunit is used to reduce the unevenness of the edge in the feature extraction. The filter parameters of the pre-processing filter subunit can be determined by "medianization", "averaging", "medianization + averaging", or "averaging + medianization", or other suitable filtering techniques, repeated measurement of the defibrator unit in a stationary state, and comparison of the average value and standard deviation of the error index obtained under various filtering techniques, so as to determine the best filter parameters required for measuring the width of the defibrator unit.

[0053] The surface feature extraction unit can classify the defibration type of the defibration unit by the Attention-UNet method. The defibration type of the defibration unit is mainly divided into three types: under-defibration, optimal defibration and over-defibration. Under-defibration represents that the defibration unit needs to be defibrated again. Optimal defibration represents that the defibration unit is properly defibrated and can enter the next process. Over-defibration represents that the defibration unit is defibrated too much, and the pressure between the defibration roller and the single board can be appropriately reduced.

[0054] The Attention-UNet method is used to classify the defibration type of the defibration unit, which includes the following steps:

[0055] Step a: Through the industrial camera, the surface image samples of the defibration unit are repeatedly collected under different pressures between the defibration roller and the single board and different types of single boards.

[0056] Step b: The surface image samples of the defibration unit are preprocessed by grayscale, HE transformation, CLAHE transformation method, or other appropriate methods.

[0057] Step c: The number of surface image samples of the defibration unit is increased by horizontal mirror flipping, translation, scaling, cropping, rotation and other transformation image operations.

[0058] Step d: The surface image samples of the defibration unit after expansion in step c are data cleaned to obtain an image dataset. The images in the image dataset are labeled using a labeling tool for subsequent training of the neural network.

[0059] Step e: The surface image samples of the defibration unit in the image dataset are divided into training set, validation set and test set according to the ratio of 7:2:1. The training set is used to train the neural network. The validation set is used to verify the training result of the neural network. The test set is used to test the neural network before formal production.

[0060] Step f: An Attention-Unet neural network is constructed, which adopts an Attention Gate structure to realize the Attention mechanism by supervising the features of the previous level by the features of the next level. The activated part is limited to the segmented area to reduce the activation value of the background to optimize the segmentation and realize end-to-end segmentation. After training, point-like or line segment-like fiber crack pictures are obtained.

[0061] Step g: Set the batch size in the Attention-Unet neural network to 2-4 and deploy the Adam optimizer. Perform forward inference on the surface image samples of the loosening unit to obtain the output results. Calculate the cross-entropy loss between the output results and the true labels, and calculate the gradient and backpropagation. Every certain period, print the current training loss and save the model weights. In addition, in each batch, the input, true label and output result are merged into an image, which is saved to the specified path for observing the training effect.

[0062] The main function of the Adam optimizer is to update the neural network parameters according to the gradient information, so as to minimize the loss function. The true label refers to the label marked on the surface image sample of the loosening unit in step d by using the marking tool to mark the images in the image dataset. By comparing the output results of the Attention-Unet neural network with the true labels, the loss function is calculated, which can evaluate the training effect of the Attention-Unet neural network.

[0063] Step h: Stop the model training after the loss function converges to obtain the trained model. Input the surface image of the loosening unit into the trained model, and output the surface feature evaluation indexes such as the number of cracks, the length of cracks, and the average width of cracks, and then obtain the loosening type of the loosening unit and evaluate the loosening quality of the loosening unit.

[0064] After the Attention-Unet neural network is trained, it can be used for online detection of the loosening quality of the loosening unit. During online detection, the loosening unit is conveyed by the conveying belt and the roller. In the process of conveying, the industrial camera of the feature extraction unit captures the surface image of the static or dynamic loosening unit and inputs it into the trained Attention-Unet neural network to extract the surface features of the loosening unit, and then evaluates the loosening quality of the loosening unit.

[0065] In this embodiment, the moisture content detection unit uses a contact moisture analyzer. The moisture content of the single board is detected by the following method: first, select 4-6 sites on the single board as detection sites, then use the mechanical arm to grab the contact moisture analyzer to detect the detection sites in turn, and finally calculate the average value of the moisture content at each detection site as the moisture content of the single board.

[0066] Taking the average value of the moisture content of multiple detection sites as the moisture content of the single board can reduce the error caused by detection calibration.

[0067] Please refer to Figure 1The control system of the wood defibration equipment also includes an optimization module. The optimization module is in communication connection with the detection module and the machine learning module, and is used for tracking and recording the data detected by the detection module in real time, and optimizing the mapping relationship in the machine learning module. Through the optimization module, the performance of the machine learning module can be ensured to be stable, and the change of data can be adapted. The machine learning module can automatically adjust the mapping relationship according to the actual running data and feedback information, thereby improving the stability, response speed and energy efficiency of the control system of the wood defibration equipment.

[0068] Please continue to refer to Figure 1 The control system of the wood defibration equipment also includes a man-machine interaction module. The man-machine interaction module is in communication connection with the detection module and the control module, and has a display unit and an operation unit. The display unit is used for displaying the data detected by the detection module. The operation unit is used for inputting instructions by the user and transmitting the instructions to the control unit, so as to control the wood defibration equipment by the user.

[0069] The man-machine interaction module has the function of the host computer, which is convenient for the user to operate the equipment, is used for receiving the input of the user and transmitting it to the detection module and the control module. The man-machine interaction module can timely and clearly provide error information to the user, and give corresponding solutions or suggestions, and can display the running state of the current equipment to monitor the production process. The man-machine interaction module can improve the real-time performance of the system, so that the user can timely acquire and process information.

[0070] The control system of the wood defibration equipment also can deploy a cloud module. The cloud module is in communication connection with the detection module and the control module. The function of the cloud module is the same as that of the man-machine interaction module, but the cloud module can be deployed in a location far away from the production site through network connection, and can remotely monitor and control the production process.

[0071] The control system of the wood defibration equipment first detects the parameters of the veneer and the defibration unit corresponding to the veneer through the detection module, then inputs the detected parameters as input data into the machine learning module for training, and establishes the mapping relationship between the input data and the output. Finally, when the wood defibration equipment is running, the parameters of the veneer to be defibrated are input as input data into the machine learning module, and the pressure between the defibration roller and the veneer, the height of the defibration roller, the rotation speed of the defibration roller and the pressure roller, and the nip between the defibration roller and the pressure roller suitable for the defibration of the veneer to be defibrated are obtained. The defibration unit is in the defibration degree range required by production. The system quantifies the key technical indicators in the veneer defibration process, thereby helping to improve the production efficiency, improve the precision of defibration processing, and help to improve the defibration uniformity in the veneer defibration production.

[0072] Please refer to Figure 2The application also discloses a control method of the wood defibration equipment, and the method comprises the following steps:

[0073] S1, obtaining training data; the training data comprises input data and output data;

[0074] The input data comprises the width of the veneer, the width of the defibration unit, the thickness of the veneer, the surface feature of the veneer, the moisture content of the veneer, and environmental data; the environmental data comprises temperature and humidity.

[0075] The output data comprises the pressure between the defibration roller and the veneer, the height of the defibration roller, the rotating speed of the defibration roller and the compression roller, and the nip between the defibration roller and the compression roller.

[0076] S2, training a machine learning regression model; inputting the training data into the machine learning regression model to establish a mapping relationship between the input data and the data.

[0077] S3, before defibrating the veneer, obtaining the width, thickness, moisture content of the veneer to be defibrated and environmental data, and inputting the machine learning regression model.

[0078] S4, based on the mapping relationship, obtaining the pressure between the defibration roller and the veneer, the height of the defibration roller, the rotating speed of the defibration roller and the compression roller, and the nip between the defibration roller and the compression roller required for defibrating the veneer to be defibrated, and controlling the wood defibration equipment to make corresponding adjustment.

[0079] The implementation principle of the control method of the wood defibration equipment is described in detail by taking the control method of the wood defibration equipment realized by using the XGBoost algorithm as an example.

[0080] In step S1, the compression rate of the defibration unit can be calculated by formula 1, which is used to represent the size of the compression amount, and provides a reference for analyzing the defibration degree of the defibration unit.

[0081]

[0082] In the formula, R compression is the compression rate of the defibration unit. H P is the thickness of the veneer. g is the gap size between the defibration roller and the compression roller.

[0083] The defibration degree of the defibration unit is related to the compression rate of the veneer, which is calculated by formula 2.

[0084]

[0085] In the formula, R scrimber is the defibration degree of the defibration unit; H A is the thickness of the defibration unit; H P is the thickness of the veneer; W PW is the width of the single board; R is the width of the defibration unit; θ, η, ω are coefficients, which need to be adjusted adaptively according to the modeling data and test results of the machine learning regression model. A W is the width of the single board; R is the width of the defibration unit; θ, η, ω are coefficients, which need to be adjusted adaptively according to the modeling data and test results of the machine learning regression model. compression W is the width of the single board; R is the width of the defibration unit; θ, η, ω are coefficients, which need to be adjusted adaptively according to the modeling data and test results of the machine learning regression model.

[0086] The pressure between the defibration roller and the single board, the lifting distance of the screw lifting machine driving the defibration roller to lift, and the gap size between the defibration roller and the compression roller are calculated by formula 3 and formula 4.

[0087]

[0088] F = γ·g + δ(g min ≤ g ≤ g max ) (Formula 4).

[0089] In the formula, F is the pressure between the defibration roller and the single board; d s is the lifting distance of the screw lifting machine driving the defibration roller to lift; g is the gap size between the defibration roller and the compression roller; α, β, γ are all coefficients. The rotation speed of the compression roller is generally set to be large enough to enable the single board to be smoothly rolled into the gap between the defibration roller and the compression roller, so α, β, γ are all fixed values that can be measured by experiments.

[0090] More specifically, in step S1, the input data includes the width of the single board, the width of the defibration unit, the thickness of the single board, the surface characteristics of the defibration unit, the moisture content of the single board, and environmental data; the environmental data includes temperature and humidity. The output data includes the pressure between the defibration roller and the single board, the height of the defibration roller, the rotation speed of the defibration roller and the compression roller, and the nip between the defibration roller and the compression roller.

[0091] Because the defibration effect of single boards made of different tree species will also differ under the same conditions, in other embodiments of the present application, the tree species can also be used as one of the input data.

[0092] After obtaining the input data and the output data, data cleaning and preprocessing can be performed thereon, such as processing missing values and abnormal values, and standardizing or normalizing the data as much as possible.

[0093] In step S1, the training data can also be divided into a training set and a test set in a ratio of 8:2. The training set is used to train the machine learning regression model, and the test set is used to evaluate the performance of the machine learning regression model.

[0094] Before step S2, the adjustment level of the screw lifting machine can also be feature-encoded according to the standard thickness of the single board and the precision of the screw lifting machine controlling the compression roller, so as to facilitate subsequent quantitative control of the screw lifting machine.

[0095] S2, training a machine learning regression model; inputting training data into the machine learning regression model to establish a mapping relationship between the input data and the data.

[0096] In step S2, the machine learning regression model is modeled using the XGBoost distributed gradient boosting library method. During the training of the machine learning regression model, the defibration quality of the defibration unit is evaluated, specifically including:

[0097] S201, obtaining samples of defibration unit surface images; the samples of defibration unit surface images are obtained based on repeated experiments on different pressures between the defibration roller and the defibration unit and different types of defibration units;

[0098] S202, preprocessing the samples of defibration unit surface images;

[0099] S203, processing the samples of defibration unit surface images by horizontal mirror flipping, translation, scaling, cropping, and rotation to increase the number of samples;

[0100] S204, data cleaning of the samples obtained in step S203 to obtain an image data set; labeling the surface features in the defibration unit surface images in the image data set using a labeling tool;

[0101] S205, constructing a deep learning network and training it using the image data set; the deep learning network is used to extract surface features of the defibration unit based on the defibration unit surface images, and then evaluate the defibration quality of the defibration unit.

[0102] In addition, in step S2, the learning rate can be set to between 0.04 and 0.06, such as 0.04, 0.05, and 0.06, to control the learning speed of the machine learning regression model and prevent overfitting, as a smaller learning rate means that the machine learning regression model needs more iterations to converge.

[0103] The subsample ratio can also be set to between 0.5 and 1, for example, 0.5 and 1. The subsample ratio and the column subsample ratio both represent a certain proportion of samples randomly selected for training trees at each iteration, which is used to control the sample proportion of each tree used to train the machine learning regression model, and helps to prevent overfitting.

[0104] The initial value of the regularization parameter can also be set to 1 or a smaller positive number to control the strength of L1 and L2 regularization and prevent the complexity of the machine learning regression model from being too high. The initial value of the regularization parameter can be adjusted by cross-validation.

[0105] S3, in the process of the single board to be defibrated, the width of the single board to be defibrated, the thickness of the single board to be defibrated, the moisture content of the single board to be defibrated and the environmental data are obtained and input into the machine learning regression model.

[0106] S4, based on the mapping relationship, the pressure between the defibration roller and the single board, the height of the defibration roller, the rotating speed of the defibration roller and the press roller, the nip between the defibration roller and the press roller required for the single board to be defibrated are obtained, and the wood defibration equipment is controlled to make corresponding adjustment.

[0107] In another embodiment of the present application, the control method of the wood defibration equipment further comprises:

[0108] S5, in the process of the machine learning regression model working, the parameters of the single board, the parameters of the defibration unit and the parameters of the wood defibration equipment are fed back to the machine learning regression model, and the machine learning regression model is continuously optimized, so that the machine learning regression model can automatically adjust the control parameters and strategies (mapping relationship), thereby improving the stability and response speed of the machine learning regression model, and reducing the energy consumption of the machine learning regression model.

[0109] The control method of the wood defibration equipment of the present application, by establishing the machine learning regression model, the trained machine learning regression model is deployed to the production environment for the prediction and decision of the control system of the wood defibration equipment. The machine learning regression model analyzes and processes real-time data, automatically decides to control the wood defibration equipment according to the change of the environment and the running state of the wood defibration equipment, so as to realize the optimal control effect. Thus, the key technical indexes in the process of single board defibration are quantified, which helps to improve the production efficiency, improve the precision of defibration processing, and help to improve the defibration uniformity in the process of single board defibration.

[0110] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not used to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those skilled in the art without departing from the spirit and technical thought disclosed by the present application should be covered by the claims of the present application.

Claims

1. A control system of a wood defibrating apparatus, characterized in that, The detection module, the machine learning module and the control module are included. The detection module includes a pressure detection unit, a thickness detection unit, a width detection unit, a surface feature extraction unit and a moisture content detection unit. The pressure detection unit is used to detect the pressure between the defibration roller and the single board in real time. The thickness detection unit is used to detect the thickness of the single board and the defibration unit. The width detection unit is used to detect the width of the single board and the defibration unit. The surface feature extraction unit is used to extract the surface features of the defibration unit and evaluate the defibration quality of the defibration unit. The moisture content detection unit is used to detect the moisture content of the single board in real time. The machine learning module is in communication connection with the pressure detection unit, the thickness detection unit, the width detection unit and the moisture content detection unit. The machine learning module is used to establish a mapping relationship between input data and output data; wherein, The input data includes the data detected by the signal detection module; the output data includes the pressure between the defibration roller and the single board, the height of the defibration roller, the rotating speed of the defibration roller and the pressure roller, and the nip between the defibration roller and the pressure roller. The machine learning module is in communication connection with the control module. The control module is used to control the lifting of the defibration roller, adjust the pressure between the defibration roller and the single board, and adjust the rotating speed of the defibration roller and the pressure roller according to the output data, so that the defibration unit is within the range of the required defibration degree. The mapping relationship includes: F = a - d s + β, where d s ranges from F = γ · g + δ, where g is in the range g min ≤ g ≤ g max ; In the formula, F is the pressure between the defibrator roller and the veneer; d s is the lifting distance of the screw lifting machine for driving the defibrator roller to lift; g is the gap size between the defibrator roller and the press roller; α, β, and γ are all coefficients, which are all constant values measured by experiments; The mapping relationship also includes: In the formula, R scrimber is the degree of defibration of the defibration unit; H A is the thickness of the defibration unit; H P is the thickness of the single board; W P is the width of the defibration unit; W A is the width of the single board; R compression is the compression rate of the defibration unit; θ, η, ω are all coefficients, which need to be adaptively adjusted according to the modeling data and test results of the machine learning regression model. Wherein, wherein R compression is the compression ratio of the defibrator; H P is the thickness of the veneer; and g is the gap size between the defibrator roll and the press roll.

2. A control system of wood fiber liberation equipment according to claim 1, characterized in that, The pressure detection unit includes a pressure sensor. The width detection unit and the surface feature extraction unit both include an industrial camera. The thickness detection unit includes a photoelectric sensor. The moisture content detection unit is a contact moisture analyzer or a non-contact moisture analyzer.

3. A control system of wood fiber liberation equipment according to claim 2, characterized in that, The surface feature extraction unit has a pre-processing filter subunit. The pre-processing filter subunit is used to reduce the edge unevenness in feature extraction.

4. The control system of wood fiber liberation equipment according to claim 2, characterized in that, The moisture content detection unit is a contact moisture analyzer. The contact moisture analyzer is driven by a mechanical arm and is used to detect the moisture content of multiple sites on the single board and then take the average value to obtain the moisture content of the single board.

5. The control system of wood fiber liberation equipment according to claim 1, characterized in that, An optimization module is also included. The optimization module is in communication connection with the detection module and the machine learning module, and is used to track and record the data detected by the detection module in real time, and optimize the mapping relationship in the machine learning module.

6. The control system of wood fiber liberation equipment according to claim 1, characterized in that, A human-computer interaction module is also included. The human-computer interaction module is in communication connection with the detection module and the control module, and has a display unit and an operation unit. The display unit is used to display the data detected by the detection module. The operation unit is used for user to input instructions and transmit the instructions to the control unit for user to control the wood defibration equipment.

7. A method of controlling a wood defibrating apparatus, characterized by The method includes: S1, obtaining training data; the training data includes input data and output data; The input data includes the width of the single board, the width of the defibration unit, the thickness of the single board, the surface features of the single board, the moisture content of the single board, and environmental data; the environmental data includes temperature and humidity. The output data includes pressure between the defibrator roller and the veneer, height of the defibrator roller, rotation speed of the defibrator roller and the press roller, and roll gap between the defibrator roller and the press roller; S2, training a machine learning regression model; inputting the training data into the machine learning regression model to establish a mapping relationship between the input data and the output data; S3, before defibrating the veneer, obtaining the width, thickness, moisture content of the veneer to be defibrated and the environmental data, and inputting them into the machine learning regression model; S4, based on the mapping relationship, obtaining the pressure between the defibrator roller and the veneer, the height of the defibrator roller, the rotation speed of the defibrator roller and the press roller, and the roll gap between the defibrator roller and the press roller required for defibrating the veneer to be defibrated, and controlling the wood defibration equipment to make corresponding adjustments; The mapping relationship includes: F = a - d s + β, where d s ranging from F = y-g + d, where g ranges from g min ≤ g ≤ g max ; In the formula, F is the pressure between the defibrator roller and the veneer; d s is the lifting distance of the screw lifting machine for driving the defibrator roller to lift; g is the gap size between the defibrator roller and the press roller; α, β, and γ are all coefficients, which are all constant values measured by experiments; The mapping relationship also includes: In the formula, R scrimber is the degree of defibration of the defibration unit; H A is the thickness of the defibration unit; H P is the thickness of the single board; W P is the width of the defibration unit; W A is the width of the single board; R compression is the compression rate of the defibration unit; θ, η, ω are all coefficients, which need to be adaptively adjusted according to the modeling data and test results of the machine learning regression model; Wherein, wherein R compression is the compression ratio of the defibrator; H P is the thickness of the veneer; g is the size of the gap between the defibrator roll and the press roll.

8. The method of claim 7, wherein, In the step S1, The training data is divided into training set and test set according to the ratio of 8:2; The training set is used to train the machine learning regression model; The test set is used to evaluate the performance of the machine learning regression model.

9. The method of claim 7, wherein: In the step S2, during the training of the machine learning regression model, the defibration quality of the defibration unit is evaluated, which specifically includes: S201, obtaining samples of defibration unit surface images; the samples of defibration unit surface images are obtained based on repeated experiments on different pressures between the defibrator roller and the defibration unit and different types of defibration units; S202, preprocessing the samples of defibration unit surface images; S203, processing the samples of defibration unit surface images by horizontal mirror image flipping, translation, scaling, cropping and rotation to increase the number of samples; S204, data cleaning is performed on the samples obtained in the step S203 to obtain an image data set; the surface features in the defibration unit surface images in the image data set are labeled by a labeling tool; S205, constructing a deep learning network and training it by using the image data set; the deep learning network is used to extract surface features of the defibration unit based on the defibration unit surface images, and then evaluate the defibration quality of the defibration unit.

Images