Raw material moisture content prediction and adjustment method for shaving board production

Abstract

The invention relates to the technical field of intelligent regulation and control, in particular to a method for predicting and adjusting the moisture content of raw materials for shaving board production. According to the method, the raw material drying section disturbance parameter set and the moisture content detection sequence record in the shaving board production process are obtained, symbol matching, amplitude sorting and disturbance direction and influence intensity marking are carried out in combination with the prediction sequence and the disturbance characteristic sequence, so that the raw material moisture content prediction has the dynamic trend judgment capability; tracing correlation can be carried out on a disturbance source, prediction sequence amplitude adjustment is carried out through a disturbance direction and an amplitude mark, prediction correction of disturbance response and trend inversion abnormal point detection are achieved, and meanwhile local residual analysis is introduced to construct a trend residual sorting table; the abnormal inflection point is taken as the core to carry out weight sorting of the correction result and guide scaling and direction adjustment of the moisture content correction trend, and finally, adjustment instructions of the hot air temperature, the heating source power and the drying channel humidity are generated according to the deviation degree of the prediction result and the target set value.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent regulation and control, and particularly to a method for predicting and adjusting the moisture content of raw materials used in particleboard production. Background Art

[0002] The technical field of intelligent regulation and control involves precise parameter control and real-time adjustment of various industrial processes to improve the automation level and production efficiency of system operation, including real-time monitoring of process parameters, prediction and control of key factors, dynamic adjustment of the production process, and intelligent analysis and optimization of the operating state of equipment.

[0003] Among them, the traditional method for predicting and adjusting the moisture content of raw materials used in particleboard production refers to the method of measuring the moisture content of raw materials by drying experiments in the particleboard production process. By physically sampling and combining the weighing analysis after heating by a drying device, the change data of the moisture content of the raw materials is obtained, and then the moisture state of the raw materials is adjusted based on manual experience or a preset moisture content standard.

[0004] The existing technology relies on the weighing analysis method combined with physical sampling after heating by a drying device to obtain the moisture content data of raw materials during the particleboard production process. This method can only make manual inferences based on the static detection results at a single time point, and cannot capture the influence direction and amplitude of disturbance factors on the moisture content of raw materials in real time, which easily leads to lagging prediction results and untimely response to disturbance changes. Moreover, in actual operation, only relying on experience or preset standards for manual adjustment, it is easy to have the situation of lagging or excessive moisture content adjustment. For example, the non-linear coupling effect of hot air temperature, heating source power or drying channel humidity changes on the moisture content of raw materials during production cannot be timely feedback and corrected, resulting in large fluctuations in the moisture content of raw materials and poor drying uniformity, which in turn leads to the risk of fluctuations in the quality of the boards in subsequent pressing and forming processes. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art, and propose a method for predicting and adjusting the moisture content of raw materials used in particleboard production.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: A method for predicting and adjusting the moisture content of raw materials used in particleboard production, comprising the following steps: S1: Obtain the disturbance parameter set detected in real time in the drying section of the particleboard raw materials and the moisture content detection sequence record of a specified period, and construct a moisture content prediction model through the XGBoost algorithm. The model predicts based on the disturbance parameter set and the moisture content detection sequence record, and outputs an initial moisture content prediction sequence; S2: Perform symbol matching and amplitude sorting on the initial moisture content prediction sequence and the disturbance parameter set, and output a set of disturbance direction and influence strength flags; S3: Adjust the amplitude of the initial moisture content prediction sequence through the disturbance direction and the influence intensity flag set, and correct the moisture content prediction through the LSTM algorithm, and output a disturbance-corrected prediction sequence; S4: Detect the reversal of the sign of the increase or decrease of the moisture content prediction value in the disturbance-corrected prediction sequence, extract the abnormal response inflection point and calculate the local trend residual, and construct an abnormal trend residual ranking table; S5: Scale the amplitude and adjust the direction of the disturbance-corrected prediction sequence according to the abnormal trend residual ranking table, and generate a drying instruction for the particleboard raw material according to the adjustment result.

[0007] As a further solution of the present invention, the initial moisture content prediction sequence includes a time-step prediction value sequence, a prediction change trend, and a prediction time axis. The disturbance direction and influence intensity flag set includes a disturbance direction flag, an influence intensity flag, and a disturbance parameter weight. The disturbance-corrected prediction sequence is specifically a predicted moisture content value sequence, a disturbance correction amplitude, and a prediction correction time node. The abnormal trend residual ranking table specifically refers to the abnormal response inflection point position, the local trend residual, and the residual ranking weight. The adjustment of the particleboard raw material drying process instruction includes the hot air temperature adjustment amplitude, the heating source power adjustment amplitude, and the drying channel humidity adjustment setting.

[0008] As a further solution of the present invention, the acquisition steps of the initial moisture content prediction sequence are specifically as follows: S111: Obtain the disturbance parameter set detected in real time in the drying section of the particleboard raw material, and the moisture content detection sequence is recorded as the moisture content detection sequence of the particleboard raw material at the same time. Perform maximum-minimum normalization processing on the disturbance parameter set to generate a normalized disturbance parameter sequence. Among them, the disturbance parameter set includes the hot air temperature change rate, the heating source power fluctuation amplitude, and the instantaneous offset of the drying channel humidity gradient; S112: Perform time-step splicing on the normalized disturbance parameter sequence and the moisture content detection sequence and use them as the input features of XGBoost. Use the moisture content detection value in the next time step as the supervision label, calculate the feature gain through error backpropagation and complete the training iteration to generate a disturbance and moisture content feature gain sequence; S113: Perform weighted calculation on the normalized disturbance parameter sequence and the moisture content detection sequence record according to the disturbance and moisture content feature gain sequence to generate an initial moisture content prediction sequence.

[0009] As a further solution of the present invention, the acquisition steps of the disturbance direction and influence intensity flag set are specifically as follows: S211: Obtain the increase or decrease difference between the predicted value of each time step and the predicted value of the previous time step from the initial moisture content prediction sequence, and calculate the sign mark of each difference to generate a moisture content prediction change sign sequence; S212: Based on the predicted moisture content change sign sequence and the normalized perturbation parameter sequence, perform element-by-element sign comparison, determine that the same sign is positive and different signs are negative, and mark a direction flag for each perturbation parameter to obtain a perturbation direction flag sequence; S213: According to the perturbation direction flag sequence, sort the absolute values of the normalized perturbation parameter sequence by amplitude size, and combine the direction flag to output a set of perturbation directions and influence intensity flags.

[0010] As a further solution of the present invention, the steps for obtaining the perturbation correction prediction sequence are specifically as follows: S311: Based on the set of perturbation directions and influence intensity flags, obtain the perturbation direction flag bit and amplitude sorting weight for each time step, and weight the direction flag bit and sorting weight with the predicted value of the time step of the initial moisture content prediction sequence to generate a perturbation weighted prediction sequence; S312: Continuously splice the predicted values of each time step in the perturbation amplitude adjustment sequence and match them with the corresponding time step numbers to form a perturbation correction input sequence; S313: Input the perturbation correction input sequence into the LSTM model, calculate the corrected predicted value of the moisture content at the current time step with the predicted value of the time step at a specified period and the current input vector, and output a perturbation correction prediction sequence.

[0011] As a further solution of the present invention, the steps for obtaining the abnormal trend residual sorting table are specifically as follows: S411: Based on the continuously increasing and decreasing change trend of the corrected predicted value of the moisture content at each time step in the perturbation correction prediction sequence, calculate the positive and negative signs of the difference between the predicted values of adjacent time steps to generate a moisture content increase and decrease change sign sequence; S412: Based on the moisture content increase and decrease change sign sequence, compare the difference between the sign at the current time step and the sign at the previous time step, determine the position where the sign changes from positive to negative or from negative to positive as the sign reversal point, and screen the position of the sign reversal time step to generate an abnormal response inflection point position sequence; S413: Based on the abnormal response inflection point position sequence, select the perturbation correction prediction sequence within a fixed time window before and after the inflection point, perform linear fitting and calculate the fitting residual value at each time step, and obtain an abnormal trend residual sorting table according to the residual value sorting.

[0012] As a further solution of the present invention, the steps for obtaining the drying instruction for the particleboard raw material are specifically as follows: S511: Based on the time step position of the specified ranking in the abnormal trend residual sorting table, select the corrected predicted value of the moisture content at the corresponding time step, and use the sorting position as a weight factor to adjust the scaling coefficient and direction sign of the predicted value to obtain a scaled and direction-adjusted prediction sequence; S512: Based on the adjusted prediction sequence in the scaling direction, calculate the deviation between the corrected predicted moisture content value and the target moisture content set value at each time step, and perform amplitude grading according to the deviation amount to generate a moisture content prediction deviation grading result; S513: Based on the moisture content prediction deviation grading result, calculate the deviation amplitude to perform corresponding adjustment amplitudes of hot air temperature, increase or decrease amplitudes of heating source power, and set parameters for drying channel humidity control, and generate a drying adjustment instruction for particleboard raw materials.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by obtaining the disturbance parameter set and moisture content detection sequence record in the raw material drying section during the production process of particleboard, combining the prediction sequence and the disturbance feature sequence for symbol matching, amplitude sorting, and marking the disturbance direction and influence intensity, the prediction of the raw material moisture content not only has the ability to judge the dynamic trend but also can trace the source of the disturbance. By adjusting the amplitude of the prediction sequence through the disturbance direction and amplitude mark, the prediction correction of the disturbance response and the detection of abnormal points of trend reversal are realized. At the same time, local residual analysis is introduced to construct a trend residual sorting table, and the weight sorting of the correction results is carried out with the abnormal inflection point as the core to guide the scaling and direction adjustment of the moisture content correction trend. Finally, based on the deviation degree between the prediction result and the target set value, adjustment instructions for hot air temperature, heating source power, and drying channel humidity are generated, enabling the moisture content prediction adjustment to adaptively respond to multi-source disturbance changes and having the capabilities of trend correction, abnormal inflection point identification, disturbance tracing, error feedback, and linkage adjustment of drying process parameters, realizing the dynamic linkage, intelligent correction, and precise feedback of the raw material moisture content prediction and adjustment process, and significantly improving the accuracy of prediction correction and the adaptability of the drying process. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the main steps of the present invention; Figure 2 It is a flowchart of step S1 of the present invention; Figure 3 It is a flowchart of step S2 of the present invention; Figure 4 It is a flowchart of step S3 of the present invention; Figure 5 It is a flowchart of step S4 of the present invention; Figure 6 It is a flowchart of step S5 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0015] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0016] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality" is two or more, unless otherwise specifically defined.

[0017] Please refer to Figure 1 , the present invention provides a technical solution: a method for predicting and adjusting the moisture content of raw materials for particleboard production, including the following steps: S1: Obtain the disturbance parameter set detected in real time in the drying section of the particleboard raw material and the moisture content detection sequence record of a specified period. Construct a moisture content prediction model through the XGBoost algorithm. The model predicts based on the disturbance parameter set and the moisture content detection sequence record, and outputs the initial moisture content prediction sequence; S2: Perform symbol matching and amplitude sorting on the initial moisture content prediction sequence and the disturbance parameter set, and output the disturbance direction and influence strength flag set; S3: Adjust the amplitude of the initial moisture content prediction sequence through the disturbance direction and influence strength flag set, and perform moisture content prediction correction through the LSTM algorithm, and output the disturbance correction prediction sequence; S4: Perform reverse detection on the increase and decrease change signs of the moisture content prediction values in the disturbance correction prediction sequence, extract the abnormal response inflection points and calculate the local trend residuals, and construct an abnormal trend residual sorting table; S5: Perform amplitude scaling and direction adjustment on the disturbance correction prediction sequence according to the abnormal trend residual sorting table, and generate a particleboard raw material drying instruction according to the adjustment result; The initial moisture content prediction sequence includes a time step prediction value sequence, a prediction change trend, and a prediction time axis. The disturbance direction and influence strength flag set includes a disturbance direction flag, an influence strength flag, and a disturbance parameter weight. The disturbance correction prediction sequence is specifically a predicted moisture content value sequence, a disturbance correction amplitude, and a prediction correction time node. The abnormal trend residual sorting table specifically refers to the abnormal response inflection point position, the local trend residual, and the residual sorting weight. The adjustment of the particleboard raw material drying process instruction includes the adjustment amplitude of the hot air temperature, the adjustment amplitude of the heating source power, and the setting of the drying channel humidity adjustment.

[0018] Please refer to Figure 2 , the specific steps for obtaining the initial moisture content prediction sequence are as follows: S111: Obtain the set of disturbance parameters detected in real time in the drying section of the particleboard raw material, and record the moisture content detection sequence as the moisture content detection sequence of the particleboard raw material at the same time. Perform maximum-minimum normalization on the set of disturbance parameters to generate a normalized disturbance parameter sequence. Among them, the set of disturbance parameters includes the hot air temperature change rate, the heating source power fluctuation amplitude, and the instantaneous offset of the humidity gradient in the drying channel; When obtaining the set of disturbance parameters detected in real time in the drying section of the particleboard raw material, it is necessary to monitor and collect the environmental and equipment parameters in the drying section in real time during the drying process of the particleboard raw material, including the hot air temperature change rate, the heating source power fluctuation amplitude, and the instantaneous offset of the humidity gradient in the drying channel. The hot air temperature change rate can be measured by a thermocouple to obtain the temperature change per minute of the hot air. For example, by real-time sampling of temperature data T1 = 150 °C, T2 = 152 °C, T3 = 151.5 °C, calculate the average temperature change rate within one minute as (152 - 150) / 1 = 2 °C / min. The heating source power fluctuation amplitude is calculated by sampling current and voltage and calculating the power. Assume that the maximum power within one minute is 15 kW and the minimum power is 12 kW, then the fluctuation amplitude is 15 - 12 = 3 kW. The instantaneous offset of the humidity gradient in the drying channel is measured by humidity sensors installed at different positions in the drying channel. For example, the inlet humidity is 45%, the outlet humidity is 20%, at a certain moment the inlet humidity changes to 46%, and the outlet humidity is 19%, then the instantaneous offset is (46 - 19) - (45 - 20) = 2%. After filling in the missing values of the above disturbance parameters by linear interpolation and performing denoising processing, a unified disturbance parameter data is formed.

[0019] S112: Perform time-step splicing on the normalized disturbance parameter sequence and the moisture content detection sequence and use them as the input features of XGBoost. Use the moisture content detection value at the next time step as the supervision label, calculate the feature gain through error backpropagation and complete the training iteration to generate a disturbance and moisture content feature gain sequence; During the process of performing time-step splicing on the normalized disturbance parameter sequence and the moisture content detection sequence, first, it is necessary to perform normalization on the disturbance parameters. For example, the original value range of the hot air temperature change rate is 0 to 10 and the current value is 2 , which is normalized to 0.2. The heating source power fluctuation amplitude value range is 0 to 5 kW, the current value is 3 kW, which is normalized to 0.6. The instantaneous offset of the humidity gradient value range is 0 to 5%, the current value is 2%, which is normalized to 0.4. The moisture content detection value does not need to be normalized. For example the moisture content at a certain moment is 6%, at a certain moment is 5.8%, at a certain moment is 5.6%. The input feature sequence formed by splicing is 、 、 , set the water content detection value of the next time step as the supervision label. For example, If the water content at a certain moment is 5.5%, then the label .

[0020] During the training process, the error backpropagation mechanism is adopted, and the loss function is the mean square error (MSE). The formula is: ; Where: : The total number of samples. For example, if the number of time steps is 5, then , determined by counting; : The sample number index, indicating the th sample, taking integer values from 1 to ; : The actual water content value of the th sample, obtained from the measurement of the water content detector. For example, , ; : The predicted water content value of the th sample, predicted by the XGBoost model. For example, , ; : Represents the square of the prediction error of the th sample. For example, .

[0021] In the XGBoost model, the core during the training process is the calculation of feature gain. The gain formula is: ; Where: : Represents the gain value when the current feature is used to split nodes. The larger the value, the more suitable the feature is as a splitting condition; : The sum of first-order gradients, calculated as , where: : The set of samples included in the current node. For example, indicates that the node contains the 1st and 2nd samples; : The gradient of the th sample, the first-order partial derivative, calculated as , where is the form of the loss function MSE , then , for example , , then ; : The sum of second-order gradients, calculated as , where: : The The Hessian value of a sample, under the MSE loss , is a constant value; : Regularization term, used to control the model complexity, usually taking values between 0.1 and 1.0. For example, set it here .

[0022] Suppose the current node contains two samples, , , then , , substituting into the formula gives , this value represents the information gain generated by splitting the node using this feature.

[0023] After calculating the gains of all features, record the gain results of each feature to form a perturbation and moisture content feature gain sequence. For example, the feature gain of the hot air temperature change rate is 0.5, the feature gain of the heating source power fluctuation amplitude is 0.6, the feature gain of the humidity gradient instantaneous offset is 0.4, and the feature gain of the moisture content is 0.3. Then the sequence is .

[0024] S113: Perform weighted calculation on the normalized perturbation parameter sequence and the moisture content detection sequence record according to the perturbation and moisture content feature gain sequence to generate an initial moisture content prediction sequence; When performing weighted calculation on the normalized perturbation parameter sequence and the moisture content detection sequence record according to the perturbation and moisture content feature gain sequence, it is first necessary to clarify the meaning of the perturbation and moisture content feature gain sequence. Each value in this sequence represents the weight contribution of the corresponding feature in the prediction. For example, the feature gain of the hot air temperature change rate is 0.5, the feature gain of the heating source power fluctuation amplitude is 0.6, the feature gain of the humidity gradient instantaneous offset is 0.4, and the feature gain of the moisture content is 0.3, which respectively represent the relative weights of these four features. Use these gain values as weighting coefficients to perform weighted calculation on the normalized perturbation parameter sequence and the moisture content detection sequence record step by step in time.

[0025] For example, for the first time step, the normalized perturbation parameter is , and the moisture content detection value is 6.0%. Then the weighted calculation process is as follows: The weighted result of the hot air temperature change rate is , the weighted result of the heating source power fluctuation amplitude is , the weighted result of the humidity gradient instantaneous offset is , the weighted result of the moisture content detection value is , and the sum of the weighted results is . This 2.42 is the first value of the initial moisture content prediction sequence after weighted calculation.

[0026] For the second time step, the normalized perturbation parameter is , the measured moisture content is 5.8%, and the weighted result of the hot air temperature change rate is , the weighted result of the heating source power fluctuation amplitude is , the weighted result of the instantaneous offset of the humidity gradient is , the weighted result of the measured moisture content is , the total sum after weighting is .

[0027] Similarly, for the third time step, the normalized perturbation parameter is , the measured moisture content is 5.6%, and the weighted result of the hot air temperature change rate is , the weighted result of the heating source power fluctuation amplitude is , the weighted result of the instantaneous offset of the humidity gradient is , the weighted result of the measured moisture content is , the total sum after weighting is .

[0028] Please refer to Figure 3 , the steps for obtaining the set of perturbation direction and influence intensity flags are specifically as follows: S211: Obtain the increase or decrease difference between the predicted value of each time step and the predicted value of the previous time step from the initial moisture content prediction sequence, and calculate the sign mark of each difference to generate a moisture content prediction change sign sequence; When obtaining the increase or decrease difference between the predicted value of each time step and the predicted value of the previous time step from the initial moisture content prediction sequence, list the arithmetic expressions step by step. For example, if the initial moisture content prediction sequence is [2.42, 2.48, 2.42, 2.50], then the difference at the second time step is 2.48 - 2.42 = 0.06, the difference at the third time step is 2.42 - 2.48 = -0.06, and the difference at the fourth time step is 2.50 - 2.42 = 0.08. The sign mark of each difference is obtained by judging its positive or negative. For example, 0.06 > 0, then it is marked as +1; -0.06 < 0, then it is marked as -1; 0.08 > 0, then it is marked as +1. The finally generated moisture content prediction change sign sequence is [+1, -1, +1], and this sequence represents the prediction change direction of each time step.

[0029] S212: Based on the moisture content prediction change sign sequence and the normalized perturbation parameter sequence, perform a sign comparison element by element, judge that the same sign is positive and different signs are negative, and mark the direction flag for each perturbation parameter to obtain a perturbation direction flag sequence; When the sign sequences of the predicted changes in water content and the normalized perturbation parameter sequences are compared element by element, first obtain the signs of each perturbation parameter. For example, if the normalized perturbation parameter sequence is [0.2, -0.6, 0.4], the sign labels are [+1, -1, +1]. The sign sequence of the predicted change in water content is [+1, -1, +1]. When the two sequences are compared element by element, the first +1 and +1 have the same sign and are marked as positive. The second -1 and +1 have different signs and are marked as negative. The third +1 and +1 have the same sign and are marked as positive. Therefore, the direction flag sequence for this time step is [positive, negative, positive]. Each perturbation parameter obtains a direction flag based on whether the signs are the same or different, and the directions of the perturbation parameters at each time step are marked in turn.

[0030] S213: Sort the absolute values of the normalized perturbation parameter sequences according to the perturbation direction flag sequence, and combine the direction flags to output the perturbation direction and influence intensity flag set; When sorting the absolute values of the normalized perturbation parameter sequences according to the perturbation direction flag sequence, first calculate the absolute values. For example, if the normalized perturbation parameter sequence is [0.2, -0.6, 0.4], the absolute values are [0.2, 0.6, 0.4]. After sorting by size, the sorting result is 0.6 > 0.4 > 0.2, and the corresponding perturbation parameters are the fluctuation amplitude of the heating source power > the instantaneous offset of the humidity gradient > the change rate of the hot air temperature. Combining with the direction flag sequence [positive, negative, positive], the perturbation direction and influence intensity flag set is obtained as the fluctuation amplitude of the heating source power (negative), the instantaneous offset of the humidity gradient (positive), and the change rate of the hot air temperature (positive). The above calculations are repeated for each time step, gradually generating the perturbation direction and influence intensity flag set to form a complete perturbation direction and intensity marking sequence.

[0031] Please refer to Figure 4 , and the steps for obtaining the perturbation correction prediction sequence are specifically as follows: S311: Obtain the perturbation direction flag bits and amplitude sorting weights for each time step based on the perturbation direction and influence intensity flag set, and weight the direction flag bits and sorting weights with the predicted values of the initial water content prediction sequence time steps to generate a perturbation weighted prediction sequence; When obtaining the perturbation direction flag bit and amplitude sorting weight for each time step based on the perturbation direction and influence intensity flag set, it is necessary to clarify the meaning and acquisition method of each element. The perturbation direction flag bit refers to the direction flag of each perturbation parameter at the current time step. For example, "forward" is marked as +1 and "backward" is marked as -1. The amplitude sorting weight comes from the weight value assigned after sorting the absolute values of the perturbation parameters. The weights are assigned in decreasing order according to the sorting order. For example, in the sorting result, the maximum value weight is set to 0.5, the second-largest value weight is set to 0.3, and the minimum value weight is set to 0.2, and so on. Taking a certain time step as an example, the perturbation direction flag bits are [forward, backward, forward], corresponding numerical marks are [+1, -1, +1], and the amplitude sorting result is the heating source power fluctuation amplitude (0.6), humidity gradient instantaneous offset (0.4), hot air temperature change rate (0.2). Then the assigned weights are [0.5, 0.3, 0.2]. The direction flag bit and sorting weight are multiplied element by element to obtain the direction-weighted value of each perturbation parameter. For example, in the second time step, the direction flag bit of the hot air temperature change rate is +1, and the sorting weight is 0.2, and the product is +1×0.2 = +0.2; the direction flag bit of the heating source power fluctuation amplitude is -1, and the sorting weight is 0.5, and the product is -1×0.5 = -0.5; the direction flag bit of the humidity gradient instantaneous offset is +1, and the sorting weight is 0.3, and the product is +1×0.3 = +0.3. The above-mentioned perturbation direction-weighted values are weighted with the time step prediction values of the initial moisture content prediction sequence. For example, the initial moisture content prediction value is 2.48, then the perturbation weighted result is 2.48 + 0.2 - 0.5 + 0.3 = 2.48. The same operation is performed for each time step to generate a perturbation weighted prediction sequence.

[0032] S312: Continuously splice the prediction values of each time step in the perturbation amplitude adjustment sequence and match them with the corresponding time step numbers to form a perturbation correction input sequence; When continuously splicing the prediction values of each time step in the perturbation amplitude adjustment sequence, it is necessary to directly splice the perturbation weighted prediction values of each time step into a long sequence according to the time step order. Taking the perturbation weighted prediction sequence as an example, if the time step prediction values are [2.48, 2.45, 2.50, 2.46], then the continuous splicing result is [2.48, 2.45, 2.50, 2.46], without additional operations. The operation of matching the time step numbers is to append the corresponding numbers at each time step. For example, the first time step number is 1, the second time step number is 2, the third time step number is 3, and the fourth time step number is 4. The prediction values and time step numbers are put in one-to-one correspondence to form the final perturbation correction input sequence, such as [(1, 2.48), (2, 2.45), (3, 2.50), (4, 2.46)]. The whole process is carried out in sequence according to the time step order, and the prediction value and number information are retained for each time step to form a complete perturbation correction input sequence.

[0033] S313: Input the perturbation-corrected input sequence into the LSTM model, calculate the corrected water content prediction value at the current time step based on the time step prediction value of the specified period and the current input vector, and output the perturbation-corrected prediction sequence; When inputting the perturbation-corrected input sequence into the LSTM model, first clarify the composition of the input vector. The input vector at each time step is , where is the time step number, is the perturbation-weighted prediction value. For example, the input sequence is . The LSTM model processes the input sequence step by step in time and calculates the corrected water content prediction value based on the recursive structure. The core calculation formula is: ; Among them, : represents the corrected water content prediction value at time step , which is calculated by the LSTM model and is gradually output at each time step according to the recursive formula. For example, at the current time step , the is determined by the of the previous time step and the current input ; : represents the LSTM output value of the previous time step , which is the corrected water content prediction value of the previous time step and comes from the recursive calculation result of the model. For example ; : represents the perturbation-weighted prediction value, which comes from the perturbation-weighted prediction sequence of S311. For example , and is obtained by weighting the perturbation direction flag bit, sorting weight, and initial water content prediction value; : is the weight matrix of the previous time step state, indicating the weighted influence of the memory ability learned by the model on the previous state. Generally, it is obtained through training. When not trained, the initial value can be set according to experimental experience. For example, according to the historical fluctuation range, set , indicating a strong memory dependence. The numerical range is usually , and in actual use, the parameters can be adjusted through the validation set; : is the weight matrix of the current input vector, indicating the sensitivity of the model to the current perturbation input. Usually, it is obtained through training. The initial experimental value can be set to according to the change amplitude of the perturbation sequence, and the range is recommended to be set within , and adjusted according to the actual perturbation amplitude ratio; : is the bias term, which is used to adjust the output baseline of the model. Usually, it is initialized to a small constant. For example , and the setting basis is to ensure that there is no excessive deviation in the initial state; : is the hyperbolic tangent activation function, which maps the weighted sum to the interval to ensure the smooth convergence of the model output. The activation function is the default choice for the internal structure of LSTM.

[0034] Let the output of the previous time step be , the perturbed weighted prediction value at the current time step be , and the weight matrix be set as , , the bias term be . Then substitute into the formula for calculation: ; Calculate . To map to the actual moisture content range , anti-normalization processing is required. The formula is: The corrected moisture content prediction value ; Substitute to get: .

[0035] Therefore, the corrected moisture content prediction value at the current time step is 9.98%. This process is calculated for each time step. For example, at calculated from and , and a complete perturbed correction prediction sequence is generated in sequence.

[0036] Please refer to Figure 5 . The steps to obtain the abnormal trend residual sorting table are specifically as follows: S411: Based on the continuous increase and decrease change trend of the corrected moisture content prediction value at each time step in the perturbed correction prediction sequence, calculate the positive and negative signs of the difference between the prediction values of adjacent time steps to generate a moisture content increase and decrease change sign sequence; When calculating the continuously increasing or decreasing trend of the corrected water content prediction values at each time step in the prediction sequence based on perturbation correction, it is necessary to calculate the differences between the prediction values of adjacent time steps and extract their positive and negative signs. First, list the perturbation correction prediction sequence step by step in time, for example, [9.98, 9.85, 10.00, 9.90]. Calculate the differences between the prediction values of adjacent time steps as follows: at the second time step, 9.85 - 9.98 = -0.13; at the third time step, 10.00 - 9.85 = 0.15; at the fourth time step, 9.90 - 10.00 = -0.10. Determine the signs based on the positive and negative of the differences. If the difference is greater than zero, mark it as +1; if the difference is less than zero, mark it as -1; if the difference is equal to zero, mark it as 0. In the above example, at the second time step, the difference is -0.13 and the sign is -1; at the third time step, the difference is 0.15 and the sign is +1; at the fourth time step, the difference is -0.10 and the sign is -1. The finally generated sign sequence of the increase and decrease of the water content is [-1, +1, -1], which is used to mark the change direction of the prediction values at each time step.

[0037] S412: Based on the sign sequence of the increase and decrease of the water content, compare the difference between the sign of the current time step and the sign of the previous time step, determine the position where the sign changes from positive to negative or from negative to positive as the sign reversal point, and screen the positions of the time steps with sign reversals to generate a sequence of abnormal response inflection point positions; When comparing the difference between the sign of the current time step and the sign of the previous time step based on the sign sequence of the increase and decrease of the water content, it is necessary to determine whether the sign has reversed. Check step by step in time. For example, the sign sequence of the increase and decrease of the water content is [-1, +1, -1]. Starting from the second time step, compare the sign +1 at the second time step with the sign -1 at the first time step. Since the signs are different, the second time step is the sign reversal point; compare the sign -1 at the third time step with the sign +1 at the second time step. Since the signs are different, the third time step is also the sign reversal point. Screen the positions of the time steps with sign reversals to generate a sequence of abnormal response inflection point positions, for example, [2, 3], indicating that sign reversals are detected at the second and third time steps, which are potential abnormal inflection point positions.

[0038] S413: Based on the sequence of abnormal response inflection point positions, select the perturbation correction prediction sequences within a fixed time window before and after the inflection points, perform linear fitting and calculate the fitting residual values at each time step, and obtain the abnormal trend residual sorting table by sorting according to the residual values; When selecting the disturbance correction prediction sequence within a fixed time window before and after the inflection point based on the abnormal response inflection point position sequence, it is necessary to first set the length of the time window. For example, if the window length is 1, it means selecting one time step before and after the inflection point time step. Taking the abnormal response inflection point position sequence [2, 3] as an example, for the inflection point at the second time step, select the disturbance correction prediction values at the first, second, and third time steps, such as [9.98, 9.85, 10.00]; for the inflection point at the third time step, select the disturbance correction prediction values at the second, third, and fourth time steps, such as [9.85, 10.00, 9.90]. Perform linear fitting within the selected window, and the fitting form is prediction value = slope × time step number + intercept, and calculate the fitting line parameters by the least squares method. Taking the inflection point at the second time step as an example, the time step numbers are [1, 2, 3], and the prediction values are [9.98, 9.85, 10.00]. The fitting line parameters are calculated, and then calculate the fitting residual value of each time step, that is, the difference between the prediction value and the fitting line value. For example, the residual at time step 1 is 9.98 - fitting value, the residual at time step 2 is 9.85 - fitting value, and the residual at time step 3 is 10.00 - fitting value. Sort all the residual values to obtain the abnormal trend residual sorting table. For example, the residual at the first time step is 0.05, the residual at the second time step is -0.10, and the residual at the third time step is 0.07. Sorting by absolute value from largest to smallest is the second time step > the third time step > the first time step, and finally form a complete abnormal trend residual sorting table.

[0039] Please refer to Figure 6 , and the steps for obtaining the particleboard raw material drying instruction are specifically as follows: S511: Based on the time step position of the specified ranking in the abnormal trend residual sorting table, select the corrected moisture content prediction value of the corresponding time step, and use the sorting position as the weight factor to adjust the scaling coefficient and direction symbol of the prediction value to obtain the scaled and direction-adjusted prediction sequence; When selecting the corrected moisture content prediction value at the corresponding time step based on the time step position of the specified rank in the abnormal trend residual sorting table, it is necessary to determine the specified rank according to the sorting table. For example, select the first two ranked time steps as references. Taking the abnormal trend residual sorting table as an example, the first rank is the second time step, and the second rank is the third time step. Then select the corrected moisture content prediction values of these two time steps. For example, the second time step is 9.85 and the third time step is 10.00. Taking the sorting position as the weight factor, for example, setting the weight factor as the ratio value of the reciprocal of the rank. The weight of the time step ranked first is 1, and the weight of the time step ranked second is 0.5. When adjusting the scaling factor, multiply the corrected prediction value by the weight factor. For example, the second time step is adjusted to 9.85×1 = 9.85, and the third time step is adjusted to 10.00×0.5 = 5.00. The direction symbol adjustment is reversed according to the sign of the prediction deviation. For example, if the moisture content prediction value is greater than the target value, the direction symbol is +1, and if it is less than the target value, it is -1. The adjusted symbol acts on the prediction value. For example, the direction symbol of the second time step is +1, and the adjusted value is 9.85×+1 = 9.85. The direction symbol of the third time step is -1, and the adjusted value is 5.00×-1 = -5.00. In this way, a prediction sequence after scaling and direction adjustment is generated.

[0040] S512: Based on the prediction sequence after scaling and direction adjustment, calculate the deviation between the corrected moisture content prediction value at each time step and the target moisture content setting value, and perform amplitude grading according to the deviation amount to generate a moisture content prediction deviation grading result; When calculating the deviation between the corrected moisture content prediction value at each time step and the target moisture content setting value based on the prediction sequence after scaling and direction adjustment, it is first necessary to clarify the setting basis of the target moisture content setting value. The target moisture content setting value is usually determined according to the particleboard production process standard. For example, the national standard requires that the moisture content range of particleboard at the time of leaving the factory is 6% to 10%. To ensure production safety and product performance, the target setting value is often taken as 8.00% as the process setting benchmark.

[0041] When calculating the deviation at each time step, directly list the formula. For example, the corrected prediction value at the second time step is 9.85, and the target value is 8.00, then the deviation is 9.85 - 8.00 = +1.85; the adjusted prediction value at the third time step is -5.00, and the target value is 8.00, then the deviation is -5.00 - 8.00 = -13.00.

[0042] The interval setting basis for the deviation amount amplitude grading is the division of the process sensitivity range and the safety control line. Considering the influence of the particleboard moisture content on product performance, the following grading intervals are set: An absolute deviation value between 0 and 1.0 indicates that the fluctuation is within the allowable range and belongs to "slight deviation". The basis is that a moisture content deviation within 1% has little impact on the mechanical properties of the product; An absolute deviation value greater than 1.0 and not exceeding 3.0 indicates a certain deviation, belonging to "medium deviation". The basis is that after the deviation exceeds 1%, the product performance begins to fluctuate significantly, and the process needs attention; An absolute deviation value greater than 3.0 and not exceeding 5.0 indicates a relatively high degree of deviation, belonging to "significant deviation". The basis is that at this time, the moisture content deviation may cause problems such as surface cracking and warping of the product; An absolute deviation value greater than 5.0 indicates a serious deviation, belonging to "severe deviation". The basis is that at this time, the moisture content far exceeds the normal process control line, and immediate measures need to be taken to adjust the drying process.

[0043] Combined with the example calculation, the deviation at the second time step is +1.85, and the absolute value is 1.85, which is in the range of 1.0 to 3.0, and the classification is "medium deviation"; the deviation at the third time step is -13.00, and the absolute value is 13.00, exceeding 5.0, and the classification is "severe deviation". The final predicted deviation classification result of the moisture content is [(2, medium deviation), (3, severe deviation)].

[0044] S513: Based on the predicted deviation classification result of the moisture content, calculate the deviation amplitude to perform corresponding adjustment amplitudes of the hot air temperature, increase and decrease amplitudes of the heating source power, and set parameters for the humidity control in the drying channel, and generate a drying adjustment instruction for the particleboard raw material; When calculating the adjustment amplitudes of the hot air temperature, increase and decrease amplitudes of the heating source power, and set parameters for the humidity control in the drying channel based on the predicted deviation classification result of the moisture content, first clarify the meaning and setting basis of each adjustment parameter: Adjustment amplitude of the hot air temperature: It represents the amount of change in the hot air temperature, with the unit of degree Celsius, and is mainly determined by the degree of moisture content deviation. The greater the deviation, the greater the adjustment amplitude; Increase and decrease amplitude of the heating source power: It represents the adjusted value of the heating source output power, with the unit of kilowatt (kW), and is set according to the relationship between the drying load and the change in moisture content; Set parameter for the humidity control in the drying channel: It represents the correction value of the set humidity in the drying channel, with the unit of percentage (%), and is achieved by adjusting the moisture discharge amount in the channel.

[0045] The core formula for calculating the adjustment amplitude is a linear proportional adjustment model, which is set as follows: , , .

[0046] Where: : Adjustment amplitude of the hot air temperature (unit: °C), used to adjust the temperature of the drying medium; : Increase and decrease amplitude of the heating source power (unit: kW), used to adjust the output of the heat source; : Adjustment value of the humidity control setting parameter for the drying channel (unit: %), used to modify the channel humidity setting; : Hot air temperature adjustment coefficient, set based on experience, for example 3.0 °C / %, indicating that for every 1% increase in the moisture content deviation, the hot air temperature needs to be adjusted by 3.0 °C; : Heating source power adjustment coefficient, set according to the equipment capacity and process design, for example 1.5 kW / %; : Humidity setting adjustment coefficient, set according to the moisture exhaust capacity and process sensitivity, for example 0.8% / %; : The absolute deviation between the predicted moisture content and the target moisture content setting value, for example, the deviation at the second time step is 1.85%.

[0047] Taking the second time step as an example, , substitute into the formula for calculation: , , .

[0048] Taking the third time step as an example again, , substitute into the formula for calculation: , , .

[0049] The finally generated drying adjustment instructions for the particleboard raw material are: at the second time step, adjust the hot air temperature to increase by 5.55 °C, the heating source power to increase by 2.775 kW, and the drying channel humidity setting to increase by 1.48%; at the third time step, adjust the hot air temperature to increase by 39.00 °C, the heating source power to increase by 19.50 kW, and the drying channel humidity setting to increase by 10.40%. Through this calculation method, for the deviation at each time step, combined with the preset adjustment coefficients, accurate drying adjustment parameters are calculated to achieve precise control.

[0050] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed technical content to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for predicting and adjusting the moisture content of raw materials for particleboard production, characterized in that, It includes the following steps: S1: Obtain the disturbance parameter set detected in real time in the drying section of the particleboard raw material and the moisture content detection sequence record for a specified period. Construct a moisture content prediction model through the XGBoost algorithm. The model makes predictions based on the disturbance parameter set and the moisture content detection sequence record, and outputs the initial moisture content prediction sequence; S2: Perform symbol matching and amplitude sorting on the initial moisture content prediction sequence and the disturbance parameter set, and output the disturbance direction and influence intensity flag set; S3: Adjust the amplitude of the initial moisture content prediction sequence according to the disturbance direction and influence intensity flag set, and perform moisture content prediction correction through the LSTM algorithm, and output the disturbance correction prediction sequence; S4: Detect the reversal of the sign of the increase and decrease change of the moisture content prediction value in the disturbance correction prediction sequence, extract the abnormal response inflection point and calculate the local trend residual, and construct an abnormal trend residual sorting table; S5: Perform amplitude scaling and direction adjustment on the disturbance correction prediction sequence according to the abnormal trend residual sorting table, and generate a particleboard raw material drying instruction according to the adjustment result.

2. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 1, characterized in that The initial moisture content prediction sequence includes a time step prediction value sequence, a prediction change trend, and a prediction time axis. The disturbance direction and influence intensity flag set includes a disturbance direction flag, an influence intensity flag, and a disturbance parameter weight. The disturbance correction prediction sequence is specifically a predicted moisture content value sequence, a disturbance correction amplitude, and a prediction correction time node. The abnormal trend residual sorting table specifically refers to the abnormal response inflection point position, the local trend residual, and the residual sorting weight. The adjustment of the particleboard raw material drying process instruction includes the adjustment amplitude of the hot air temperature, the adjustment amplitude of the heating source power, and the setting of the drying channel humidity adjustment.

3. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 1, wherein, The specific steps for obtaining the initial moisture content prediction sequence are as follows: S111: Obtain the disturbance parameter set detected in real time in the drying section of the particleboard raw material, and the moisture content detection sequence record is the moisture content detection sequence of the particleboard raw material at the same time. Perform maximum-minimum normalization processing on the disturbance parameter set to generate a normalized disturbance parameter sequence. Among them, the disturbance parameter set includes the hot air temperature change rate, the heating source power fluctuation amplitude, and the instantaneous offset of the drying channel humidity gradient; S112: Perform time step splicing on the normalized disturbance parameter sequence and the moisture content detection sequence and use them as the input features of XGBoost. Use the moisture content detection value of the next time step as the supervision label, calculate the feature gain through error backpropagation and complete the training iteration to generate a disturbance and moisture content feature gain sequence; S113: Perform weighted calculation on the normalized disturbance parameter sequence and the moisture content detection sequence record according to the disturbance and moisture content feature gain sequence to generate the initial moisture content prediction sequence.

4. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 3, characterized in that, The specific steps for obtaining the disturbance direction and influence intensity flag set are as follows: S211: Obtain the increase and decrease difference between the prediction value of each time step and the prediction value of the previous time step from the initial moisture content prediction sequence, and calculate the sign mark of each difference to generate a moisture content prediction change sign sequence; S212: Based on the predicted moisture content change sign sequence and the normalized perturbation parameter sequence, perform element-by-element sign comparison, determine that the same sign is positive and different signs are negative, and mark a direction flag for each perturbation parameter to obtain a perturbation direction flag sequence; S213: According to the perturbation direction flag sequence, sort the absolute values of the normalized perturbation parameter sequence by magnitude, and combine the direction flags to output a perturbation direction and influence intensity flag set.

5. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 4, wherein, The steps for obtaining the perturbation correction prediction sequence are specifically as follows: S311: Based on the perturbation direction and influence intensity flag set, obtain the perturbation direction flag bit and amplitude sorting weight for each time step, and weight the direction flag bit and sorting weight with the predicted value of the initial moisture content prediction sequence at the time step to generate a perturbation weighted prediction sequence; S312: Continuously splice the predicted values of each time step in the perturbation amplitude adjustment sequence and match them with the corresponding time step numbers to form a perturbation correction input sequence; S313: Input the perturbation correction input sequence into the LSTM model, calculate the corrected predicted moisture content value at the current time step based on the predicted value of the time step in the specified period and the current input vector, and output the perturbation correction prediction sequence.

6. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 5, characterized in that, The steps for obtaining the abnormal trend residual sorting table are specifically as follows: S411: Based on the continuous increase and decrease change trend of the corrected predicted moisture content value at each time step in the perturbation correction prediction sequence, calculate the positive and negative signs of the difference between the predicted values of adjacent time steps to generate a moisture content increase and decrease change sign sequence; S412: Based on the moisture content increase and decrease change sign sequence, compare the difference between the sign at the current time step and the sign at the previous time step, determine that the position where the sign changes from positive to negative or from negative to positive is a sign reversal point, and screen the position of the sign reversal time step to generate an abnormal response inflection point position sequence; S413: Based on the abnormal response inflection point position sequence, select the perturbation correction prediction sequence within a fixed time window before and after the inflection point, perform linear fitting and calculate the fitting residual value at each time step, and sort according to the residual value to obtain the abnormal trend residual sorting table.

7. The method for predicting and adjusting the moisture content of raw materials for particleboard production according to claim 6, characterized in that, The steps for obtaining the drying instruction for the particleboard raw material are specifically as follows: S511: Based on the time step position of the specified ranking in the abnormal trend residual sorting table, select the corrected predicted moisture content value at the corresponding time step, and use the sorting position as a weight factor to adjust the scaling coefficient and direction sign of the predicted value to obtain a scaled and direction-adjusted prediction sequence; S512: Based on the scaled and direction-adjusted prediction sequence, calculate the deviation between the corrected predicted moisture content value at each time step and the target moisture content set value, and perform amplitude grading according to the deviation amount to generate a moisture content prediction deviation grading result; S513: Based on the moisture content prediction deviation grading result, calculate the deviation amplitude to adjust the corresponding hot air temperature adjustment amplitude, heating source power increase and decrease amplitude, and drying channel humidity control set parameters to generate a drying adjustment instruction for the particleboard raw material.

Images