Quartz stone slab manufacturing method

Abstract

The embodiments of the present invention relate to the technical field of artificial quartz stone slabs and disclose a method for manufacturing quartz stone slabs. At least two materials required for manufacturing the quartz stone slabs are added to a stirring tank according to a preset ratio, and the at least two materials are stirred by a stirrer to obtain a mixture. During the stirring process, stirring parameters are collected according to a preset collection frequency. Target pressure-vibration control parameters for pressing a quartz stone slab that meets the target product quality are determined based on the stirring parameters. The mixture poured into a mold is vibrated and pressed based on the target pressure-vibration control parameters to obtain an initial quartz stone slab. The initial quartz stone slab is heated and cured, demolded, polished, and cut to obtain a quartz stone slab. Target pressure-vibration control parameters suitable for pressing a quartz stone slab that meets the target product quality are determined based on the actual stirring state of the mixture. The mixture is pressed based on the target pressure-vibration control parameters, thereby ensuring the product quality of the manufactured quartz stone slab.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of artificial quartz stone slabs, and in particular to a method for manufacturing a quartz stone slab. Background Art

[0002] When manufacturing quartz slabs, there are certain errors in the input amount of materials, and different batches of materials may have subtle differences in basic properties, such as particle size, material hardness, and material ratio. These differences in basic material properties and / or input amount can result in different mixing states under the same mixing conditions. When vibrating and compacting these mixtures with different mixing states using the same pressure and vibration control parameters, the resulting quartz slabs have varying product quality. Therefore, a solution to this problem is urgently needed. Summary of the Invention

[0003] The purpose of the present invention is to provide at least a method for manufacturing quartz stone slabs, which can at least solve the technical problem that the product quality of the quartz stone slabs obtained by vibrating and pressing mixed materials in different stirring states with the same pressure-vibration control parameters is uneven, and at least achieve the technical effect of ensuring the product quality of the manufactured quartz stone slabs.

[0004] To solve the above technical problems, at least one embodiment of the present application provides a method for manufacturing a quartz stone slab, comprising: adding at least two materials required for manufacturing the quartz stone slab into a stirring tank according to a preset ratio, and stirring the at least two materials in the stirring tank with a stirrer to obtain a mixture; during the stirring process, collecting stirring parameters according to a preset collection frequency, the stirring parameters including torque data of the stirrer, power data of the stirrer, and infrared temperature image data of the stirring tank; based on the stirring parameters, determining target pressure-vibration control parameters for pressing a quartz stone slab that meets the target product quality; pouring the mixture into a mold, and vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters to obtain an initial quartz stone slab; heating and curing the initial quartz stone slab, demolding, grinding and polishing, and cutting the initial quartz stone slab to obtain a quartz stone slab.

[0005] This solution obtains the stirring parameters during the stirring process, and based on the stirring parameters, determines the target pressure-vibration control parameters suitable for pressing out quartz stone slabs that meet the target product quality. The mixture is pressed based on the target pressure-vibration control parameters, thereby solving the problem of uneven product quality of the pressed quartz stone slabs caused by pressing the stirring states of different mixtures based on the same pressure-vibration control parameters. This ensures the product quality of the manufactured quartz stone slabs. In addition, the stirring state is determined by using the stirring parameters instead of directly using the stirring state data because the stirring parameters do not require a sampling device to be set up in the tank and have better sampleability than the stirring state data. Moreover, the stirring parameters can accurately determine the stirring state through the real-time stirring torque data and power data of the stirrer, avoiding the problem of inaccurate stirring state data collection due to insufficient and uneven stirring of the mixture after stirring and unreasonable selection of sampling points.

[0006] In some examples, based on the stirring parameters, target pressure-vibration control parameters for pressing quartz stone slabs that meet the target product quality are determined, including: generating a stirring feature vector based on the stirring parameters; determining whether the stirring feature vector meets the expected standard stirring feature vector; if not, obtaining the target pressure-vibration control parameters based on a preset pressure-vibration control parameter prediction model and the stirring parameter prediction, the preset pressure-vibration control parameter prediction model is used to predict the pressure-vibration control parameters that can be used to press quartz stone slabs that meet the target product quality; if it meets the standard, obtaining the preset pressure-vibration control parameters corresponding to the expected standard stirring feature vector, and using them as the target pressure-vibration control parameters.

[0007] In some examples, generating a stirring feature vector based on stirring parameters includes: performing feature processing on the stirring parameters based on a low-latency feature encoding method to generate a stirring feature vector; determining whether the stirring feature vector meets the expected standard stirring feature vector, including: calculating the similarity between the stirring feature vector and the expected standard stirring feature vector based on a preset similarity calculation method; if the similarity is less than a preset similarity threshold, it is deemed to be non-compliant.

[0008] In some examples, the preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2, and an LSTM time series sub-model M3. The target pressure-vibration control parameter is obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: inputting the stirring parameter into M1, M2, and M3, respectively, to obtain a first predicted pressure-vibration control parameter, a second predicted pressure-vibration control parameter, and a third predicted pressure-vibration control parameter, respectively; inputting the stirring parameter into the fusion sub-model to obtain a first predicted weight for the first predicted pressure-vibration control parameter, a second predicted weight for the second predicted pressure-vibration control parameter, and a third predicted weight for the third predicted pressure-vibration control parameter; and using the fusion sub-model to generate the target pressure-vibration control parameter based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the first prediction weight, the second prediction weight, and the third prediction weight.

[0009] In some examples, the preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2, and an LSTM time series sub-model M3. The target pressure-vibration control parameter is obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: inputting the stirring parameter into M1, M2, and M3, respectively, to obtain a first predicted pressure-vibration control parameter, a second predicted pressure-vibration control parameter, and a third predicted pressure-vibration control parameter, respectively; inputting the stirring feature vector into the fusion sub-model to obtain a fourth predicted weight W1 for the first predicted pressure-vibration control parameter, a fifth predicted weight W2 for the second predicted pressure-vibration control parameter, and a sixth predicted weight W3 for the third predicted pressure-vibration control parameter; and using the fusion sub-model to generate the target pressure-vibration control parameter based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the fourth predicted weight W1, the fifth predicted weight W2, and the sixth predicted weight W3.

[0010] In some examples, the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, and the third predicted pressure-vibration control parameter are multidimensional parameter data, and the fourth predicted weight W1, the fifth predicted weight W2, and the sixth predicted weight W3 respectively include a number of fourth predicted sub-weights, a number of fifth predicted sub-weights, and a number of sixth predicted sub-weights that correspond one-to-one to the multidimensional parameter data, and the number of the fourth predicted sub-weights, the number of the fifth predicted sub-weights, and the number of the sixth predicted sub-weights satisfy the second weight constraint relationship.

[0011] In some examples, the fusion sub-model includes a K-means model and a processing module, and the inputting of the stirring feature vector into the fusion sub-model to obtain a fourth prediction weight W1 for the first predicted pressure-vibration control parameter, a fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and a sixth prediction weight W3 for the third predicted pressure-vibration control parameter includes: inputting the stirring feature vector into the K-means model to obtain a classification result G output by the K-means model; G is the first label G1, the second label G2, and the third label G3 corresponding to M1, M2, and M3, respectively. one of them; if G=Gi, it means that the credibility of Mi is greater than the credibility of the other two of M1, M2 and M3 except Mi; i=1, 2, 3; using the processing module to determine the weight group corresponding to the classification result G according to G and the preset weight mapping relationship; the weight group includes a fourth prediction weight W1, a fifth prediction weight W2 and a sixth prediction weight W3 corresponding to the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter respectively; and in the weight group corresponding to Gi, Wi is greater than the other two of W1, W2 and W3 except Wi.

[0012] In some examples, the fusion sub-model includes a K-means model and a processing module, and the inputting of the stirring feature vector into the fusion sub-model to obtain the fourth prediction weight W1 for the first predicted pressure-vibration control parameter, the fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and the sixth prediction weight W3 for the third predicted pressure-vibration control parameter includes: inputting the stirring feature vector into the K-means model to obtain a confidence list Z=(Z1, Z2, Z3) output by the K-means model; wherein Zi is the confidence corresponding to Mi; Zi represents the credibility of Mi; i=1, 2, 3; and using the processing module to determine the fourth prediction weight W1, the fifth prediction weight W2, and the sixth prediction weight W3 corresponding to Z1, Z2, and Z3, respectively, according to the mapping relationship between Z and the preset weight.

[0013] In some examples, the method also includes: obtaining the stirring state of the mixture, and determining the PID control parameters and filtering parameters during the vibration pressing process, as well as the target pressing state corresponding to the target pressure-vibration control parameters based on the stirring state and the target pressure-vibration control parameters; during the vibration pressing process, obtaining the actual pressing state of the mixture in real time, and filtering the actual pressing state using the filtering parameters to obtain a corrected pressing state; based on the corrected pressing state, the target pressing state and the PID control parameters, correcting the above-mentioned target pressure-vibration control parameters to obtain the corrected target pressure-vibration control parameters; and using the corrected target pressure-vibration control parameters to implement vibration pressing on the mixture.

[0014] In some examples, based on the stirring state and target pressure-vibration control parameters, PID control parameters and filtering parameters are determined during the vibration pressing process, including: obtaining the target stirring state corresponding to the target pressure-vibration control parameters; inputting the stirring state and the target stirring state into a PID control parameter prediction model to obtain PID control parameters; inputting the stirring state and the target stirring state into a filtering control parameter prediction model to obtain filtering control parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] One or more embodiments are exemplarily described by the figures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments.

[0016] Figure 1 This is a flow chart of a method for manufacturing a quartz stone slab provided by one embodiment of the present application;

[0017] Figure 2 is a schematic diagram of a flow chart for determining target pressure-vibration control parameters provided by an embodiment of the present application;

[0018] Figure 3 This is a schematic diagram of a structure for determining target pressure-vibration control parameters based on a model provided by an embodiment of the present application;

[0019] Figure 4 This is a schematic diagram of another structure for determining target pressure-vibration control parameters based on a model provided by an embodiment of the present application;

[0020] Figure 5 It is a structural diagram of a quartz stone slab manufacturing system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, each embodiment of the present application will be described in detail below with reference to the accompanying drawings. However, it will be understood by those skilled in the art that in each embodiment of the present application, many technical details are proposed to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is for convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined and referenced with each other under the premise of no contradiction.

[0022] It should be noted that the acquisition or use of data in the embodiments of this application requires the user's consent. The relevant data can only be obtained after the user's authorization and permission, and the acquisition or use of the data complies with the provisions of relevant laws and regulations.

[0023] In order to facilitate understanding of the embodiments of the present application, relevant content about the quartz stone slab manufacturing method is first introduced here.

[0024] When manufacturing quartz slabs, there are certain errors in the input amount of materials, and different batches of materials may have subtle differences in basic properties, such as particle size, material hardness, and material ratio. These differences in basic material properties and / or input amount can result in different mixing states under the same mixing conditions. When vibrating and compacting these mixtures with different mixing states using the same pressure and vibration control parameters, the resulting quartz slabs have varying product quality. Therefore, a solution to this problem is urgently needed.

[0025] In order to solve the above-mentioned technical problem that the quality of the quartz stone slabs obtained by vibrating and pressing the mixtures in different stirring states with the same pressure-vibration control parameters is uneven, the present invention proposes a method for manufacturing quartz stone slabs. The implementation details of the quartz stone slab manufacturing method of this embodiment are described in detail below. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this solution.

[0026] Example 1:

[0027] The specific process of the quartz stone plate manufacturing method of this embodiment can be as follows: Figure 1 As shown, including:

[0028] Step 110 : adding at least two materials required for manufacturing the quartz stone slab into a stirring tank according to a preset ratio, and stirring the at least two materials in the stirring tank with a stirrer to obtain a mixture.

[0029] Specifically, quartz stone slabs are a kind of artificial composite material made of quartz sand as the main component, with a small amount of resin and other additives added, under vacuum conditions and high pressure.

[0030] Specifically, the at least two materials include at least quartz sand and resin, and may also include other additives such as pigment, curing agent, nano titanium dioxide, silane coupling agent, glass powder, sodium tripolyphosphate, etc.

[0031] Specifically, the preset ratio refers to the proportion of each material in the at least two materials added to the stirring tank. Specifically, the preset ratio is set according to the actual needs of manufacturing quartz stone slabs.

[0032] Specifically, the order of adding at least two materials to the stirring tank can be determined according to actual needs. The at least two materials can be added to the stirring tank at the same time, or the at least two materials can be added to the stirring tank in a specified order. For example, quartz sand can be added to the stirring tank first, and then the remaining materials are added to the stirring tank.

[0033] Specifically, a mixing tank refers to a container or tank that holds at least two materials.

[0034] Specifically, an agitator refers to a device installed inside a mixing tank for mixing at least two materials.

[0035] Specifically, the mixed material refers to a material obtained by uniformly or unevenly mixing at least two materials added to a stirring tank.

[0036] Step 120 : During the stirring process, stirring parameters are collected according to a preset collection frequency. The stirring parameters include torque data of the stirrer, power data of the stirrer, and infrared temperature image data of the stirring tank.

[0037] Specifically, the infrared temperature image data refers to a temperature image of the temperature in the stirring pool collected by an infrared thermal imager, which is used to characterize the temperature conditions in the stirring pool using the image, that is, to determine the temperature data in the stirring pool based on the infrared temperature image data.

[0038] Specifically, the mixer stirs with fixed preset parameters, such as a fixed preset speed. However, during the stirring process, due to differences in the materials being stirred, such as differences in the basic properties of the materials, the mixer is subject to different resistance during the stirring process, and the fusion effect of the materials after the same / or slightly different stirring is also different, resulting in different torque, power, and heating states of the mixer (such as resin). For example, when the viscosity of the mixture is high, the torque increases, the power decreases, and the heating increases. Based on this, the sampling device collects the stirring parameters at a preset frequency, especially the torque data, power data in the stirring parameters, and the infrared temperature image data in the stirring tank, which can effectively reflect the properties of the mixture after stirring, and provide an accurate and effective data basis for the subsequent selection of appropriate pressure and vibration control parameters.

[0039] In some examples, in order to ensure the accuracy of reflecting the stirring effect and fully understand the changes in the state properties of the mixture during the stirring process, this solution is set to obtain stirring parameters multiple times, specifically: collecting stirring parameters in real time according to a preset acquisition frequency; or collecting stirring parameters according to a preset time interval and a preset acquisition frequency. It should be understood that when obtaining stirring parameters, multiple stirring parameters can be obtained at the same time, such as obtaining torque data, power data and infrared temperature data at the same time according to a preset acquisition frequency, or obtaining torque data, power data and infrared data at the same time according to a preset time interval. Thus, based on the stirring parameters at different time points, the dynamic changes in the stirring process can be accurately captured, making the determined stirring effect more accurate, avoiding the problem that the data at a single time point may not fully reflect the entire stirring process and cannot accurately determine the stirring state. The stirring parameters collected at a certain time point are obtained by obtaining the torque data of the stirrer, the power data of the stirrer and the infrared temperature image data of the stirring tank at multiple time points.

[0040] Step 130: Determine target pressure-vibration control parameters for pressing a quartz stone plate that meets target product quality based on the stirring parameters.

[0041] Specifically, the target pressure-vibration control parameters include the target pressure value, target increase rate, target holding time, target vibration frequency, and target vibration amplitude. The target increase rate refers to the speed at which the pressure is gradually increased from the initial value to the set target pressure value during the pressing process; the target holding time refers to the duration of the pressing of the mixed material, which can be specifically described as a period or a duration, where the duration or period is equal to the pressing duration.

[0042] It should be understood that in the actual vibration pressing process of manufacturing quartz stone slabs, in order to ensure the quality of the manufactured quartz stone slabs, it is often necessary to perform multiple vibration pressing on the quartz stone slabs, that is, the vibration pressing includes multiple pressing stages, and the target pressure-vibration control parameters include the stage pressure-vibration control parameters of each pressing stage in the multiple pressing stages.

[0043] For example, if vibration compaction includes three stages, the target pressure-vibration control parameters include stage-by-stage pressure-vibration control parameters for each of the three stages. That is, the target pressure-vibration control parameters include three stage-by-stage pressure-vibration control parameters. Each stage-by-stage pressure-vibration control parameter includes a stage-by-stage pressure value, a stage-by-stage pressure increase rate, a stage-by-stage pressure dwell time, a stage-by-stage vibration frequency, and a stage-by-stage vibration amplitude. Parameters of the same dimension within the pressure-vibration control parameters for different stages can have the same or different values. Taking the stage-by-stage pressure value as an example, the parameter values ​​can be the same or different across the three stages of vibration compaction. To better understand the parameters for each stage, this solution uses the example of each stage-by-stage pressure-vibration control parameter including a stage-by-stage pressure value and a stage-by-stage pressure dwell time. For example, when manufacturing a 20mm thick quartz slab, the target pressure-vibration control parameters determined include the three stage-by-stage pressure-vibration control parameters. The stage-by-stage pressure-vibration control parameters for the first stage include a stage-by-stage pressure dwell time of 120 seconds (i.e., starting at zero seconds and ranging from 0 to 120 seconds), a stage-by-stage pressure value of 300 tons, and a vacuum device activated in the first stage to expel a large amount of air. The second stage pressure-vibration control parameters include a holding time of 25 seconds (from 120 to 145 seconds) and a pressure of 800 tons. The vibration device is activated to rearrange the particles and reduce the space between them. The third stage pressure-vibration control parameters include a holding time of 200 seconds (from 145 to 345 seconds) and a pressure of 800 tons. Vibration is continued to completely fill the voids with resin.

[0044] For some examples, see Figure 2 In the aforementioned step 130, the target pressure-vibration control parameters for pressing the quartz stone plate meeting the target product quality are determined, including steps 1301 to 1304:

[0045] Step 1301: Generate a stirring feature vector based on stirring parameters.

[0046] In some examples, to improve data acquisition efficiency, generating a stirring feature vector based on stirring parameters in step 1301 includes performing feature processing on the stirring parameters using a low-latency feature encoding method to generate the stirring feature vector. As can be seen, the low-latency feature encoding method can quickly acquire the stirring feature vector, thereby improving data acquisition efficiency.

[0047] Specifically, low-latency feature encoding methods include statistical feature methods, energy feature methods, peak detection methods, etc.

[0048] Specifically, the stirring feature vector includes a multi-dimensional feature vector, and the multi-dimensionality can be set based on actual needs. For example, the mean, standard deviation, sum of squares, maximum and minimum values ​​of torque data, power data and temperature data are set, and the stirring feature vector includes the mean, standard deviation, sum of squares, maximum and minimum values. Among them, the multiple dimensions of the stirring feature vector correspond one to one with the multiple dimensions of the expected standard stirring feature vector. For relevant content on how to obtain the stirring feature vector based on the feature encoding method, please refer to the prior art, and no further details will be given here.

[0049] Step 1302 : Determine whether the stirring feature vector meets the expected standard stirring feature vector.

[0050] In some examples, in the aforementioned step 1302, determining whether the stirring feature vector meets the expected standard stirring feature vector includes: calculating the similarity between the stirring feature vector and the expected standard stirring feature vector based on a preset similarity calculation method; if the similarity is less than a preset similarity threshold, it is considered not to meet the requirements.

[0051] Specifically, the expected standard stirring feature vector is used to characterize the expected standard stirring state of the mixture after stirring in a quantitative form.

[0052] Specifically, the preset similarity calculation method can be any one of the following methods: cosine similarity, Euclidean distance, Manhattan distance, and Mahalanobis distance. The preset similarity calculation method can also be other methods that can calculate the similarity between the stirring feature vector and the expected standard stirring feature vector. No more details are given here.

[0053] Step 1303: If not, target pressure-vibration control parameters are obtained based on a preset pressure-vibration control parameter prediction model and agitation parameter prediction. The preset pressure-vibration control parameter prediction model is used to predict pressure-vibration control parameters that can produce a quartz stone plate that meets the target product quality.

[0054] As can be seen, by encoding the stirring parameters, it is possible to quickly determine whether the stirring state of the mixture after stirring meets the expected stirring state. The stirring parameters of the mixture that do not meet the expected stirring state are input into the preset pressure-vibration control parameter prediction model to analyze the stirring parameters and obtain the target pressure-vibration control parameters. By reusing the stirring parameters multiple times, the computing resources that would be unnecessary to analyze the stirring parameters using the model are saved, improving the efficiency of quartz stone slab manufacturing. Furthermore, the stirring parameters that require model analysis are accurately determined, thereby improving the product quality of the manufactured quartz stone slabs.

[0055] It should be understood that to improve the accuracy of the target pressure-vibration control parameters, this solution sets the preset pressure-vibration control parameter prediction model as a hybrid model. Multiple prediction values ​​are derived based on the strengths of each model in the hybrid model. These different prediction values ​​are then integrated to obtain the target pressure-vibration control parameters, thereby improving the accuracy of the target pressure-vibration control parameters. Specifically, there are two methods for obtaining the target pressure-vibration control parameters based on the preset pressure-vibration control parameter prediction model: Method 1 and Method 2.

[0056] Method 1: In the above step 1303, please refer to Figure 3 The preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2 and an LSTM time series sub-model M3. The target pressure-vibration control parameters are obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: inputting the stirring parameters into M1, M2 and M3 respectively to obtain the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter respectively; inputting the stirring parameters into the fusion sub-model to obtain the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter; and using the fusion sub-model to generate the target pressure-vibration control parameters based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the first prediction weight, the second prediction weight and the third prediction weight.

[0057] Method 2: In the above step 1303, please refer to Figure 4 The preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2 and an LSTM time series sub-model M3. The target pressure-vibration control parameter is obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: inputting the stirring parameter into M1, M2 and M3 respectively to obtain a first predicted pressure-vibration control parameter, a second predicted pressure-vibration control parameter and a third predicted pressure-vibration control parameter respectively; inputting the stirring feature vector into the fusion sub-model to obtain a fourth predicted weight W1 for the first predicted pressure-vibration control parameter, a fifth predicted weight W2 for the second predicted pressure-vibration control parameter and a sixth predicted weight W3 for the third predicted pressure-vibration control parameter; and using the fusion sub-model to generate the target pressure-vibration control parameter based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the fourth predicted weight W1, the fifth predicted weight W2 and the sixth predicted weight W3.

[0058] Specifically, the fusion sub-model is used to predict the prediction weights of each regression sub-model and to fuse the predicted pressure-vibration control parameters of different sub-models with the predicted weights. The weight prediction function of the fusion sub-model can be trained based on a classification model, which can be a K-means model. The output of the K-means model is a classification identifier (for this solution, the classification identifier can include a first identifier, a second identifier, and a third identifier) ​​and / or the confidence level of the first identifier, the second identifier, and the third identifier. Specifically, the confidence level of a particular identifier indicates how well the corresponding regression sub-model predicts the mixture pressure-vibration control parameters.

[0059] In a feasible embodiment, the fusion sub-model includes a K-means model and a processing module, and the stirring feature vector is input into the fusion sub-model to obtain the fourth prediction weight W1 for the first predicted pressure-vibration control parameter, the fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and the sixth prediction weight W3 for the third predicted pressure-vibration control parameter, including: inputting the stirring feature vector into the K-means model to obtain the classification result G output by the K-means model; G is the first label G1, the second label G2, and the third label G3 corresponding to M1, M2, and M3 respectively. One of them; if G=Gi, it means that the credibility of Mi is greater than the credibility of the other two of M1, M2 and M3 except Mi; i=1, 2, 3; the processing module is used to determine the weight group corresponding to the classification result G according to the mapping relationship between G and the preset weight; the weight group includes the fourth prediction weight W1, the fifth prediction weight W2 and the sixth prediction weight W3 corresponding to the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter respectively; and in the weight group corresponding to Gi, Wi is greater than the other two of W1, W2 and W3 except Wi.

[0060] In another feasible embodiment, the fusion sub-model includes a K-means model and a processing module, and the stirring feature vector is input into the fusion sub-model to obtain the fourth prediction weight W1 for the first predicted pressure-vibration control parameter, the fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and the sixth prediction weight W3 for the third predicted pressure-vibration control parameter, including: inputting the stirring feature vector into the K-means model to obtain the confidence list Z=(Z1, Z2, Z3) output by the K-means model; wherein Zi is the confidence corresponding to Mi; Zi represents the credibility of Mi; i=1, 2, 3; using the processing module according to the mapping relationship between Z and the preset weight, respectively determine the fourth prediction weight W1, the fifth prediction weight W2 and the sixth prediction weight W3 corresponding to Z1, Z2, and Z3.

[0061] Taking the output of a K-means model as a classification result as an example, during the training process of the K-means model, multiple sample data containing a first identifier corresponding to the random forest regression sub-model M1, a second identifier corresponding to the GBDT regression sub-model M2, and a third identifier corresponding to the LSTM time series sub-model M3 as sample labels are used to train a K-means model that can be used as a classification in the fusion sub-model. In this case, each prediction result obtained based on the K-means model includes one of the first identifier, the second identifier, and the third identifier. Furthermore, each identifier corresponds to a preset weight set, and the mapping relationship between the weight set and the identifier can be determined using a preset weight mapping relationship (e.g., a preset configuration table). Therefore, based on each identifier, the weight set corresponding to each identifier can be determined from the preset configuration table to determine the prediction weights corresponding to the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3, ultimately achieving fusion.

[0062] It should be understood that the classification result output by the K-means model is used to indicate which regression sub-model should be more trusted in this fusion. Therefore, the prediction weight corresponding to the more trusted regression sub-model is greater than the prediction weights corresponding to the other two regression sub-models. Exemplarily, the K-means model classifies and predicts the predicted pressure-vibration control parameters obtained by the three regression sub-models, and the identification Gi of the regression sub-model with the more credible classification result is obtained, where if G=Gi, it means that the credibility of Mi is greater than the credibility of the other two of M1, M2, and M3 except Mi, i=1, 2, 3. For example, when the random forest regression sub-model M1 is more credible, the classification result G can be the first label G1 corresponding to the random forest regression sub-model M1, that is, G=G1. When the GBDT regression sub-model M2 is more credible, the classification result G can be the second label G2 corresponding to the GBDT regression sub-model M2, that is, G=G2. When the LSTM time series sub-model M3 is more credible, the classification result G can be the third label G3 corresponding to the LSTM time series sub-model M3, that is, G=G3. Then, the processing module determines the weight group corresponding to the classification result based on the classification result G (the label output by the K-means model) and the preset weight mapping relationship (such as a preset configuration table), that is, the weight group corresponding to the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, and the third predicted pressure-vibration control parameter. Specifically, the weight group includes three weights, such as the fourth prediction weight W1, the fifth prediction weight W2, and the sixth prediction weight W3, which are the prediction weights of the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3 for predicting the stirring feature vector. And because the classification result represents a more credible regression sub-model, the prediction weight corresponding to this regression sub-model is greater than the prediction weights corresponding to the other two regression sub-models, that is, in the same weight group, Wi is greater than the other two of W1, W2, and W3 except Wi. For example, when the classification result G = G1, the fourth prediction weight W1 for the first predicted pressure-vibration control parameter is greater than the fifth prediction weight W2 for the second predicted pressure-vibration control parameter and the sixth prediction weight W3 for the third predicted pressure-vibration control parameter, respectively. That is, W1 > W2, and W1 > W3. In one specific embodiment, the three weights may be W1 = 0.5, W2 = 0.25, and W3 = 0.25, but this application does not impose any specific limitations thereon.

[0063] It should be understood that when the pressure-vibration control parameter is multi-dimensional parameter data, the weight corresponding to the classification result is 3 n weights, where n is the number of dimensions corresponding to the pressure-vibration control parameters.

[0064] Taking the output of a K-means model as confidence, this example illustrates that during K-means model training, multiple sample data containing a first identifier corresponding to a random forest regression sub-model M1, a second identifier corresponding to a GBDT regression sub-model M2, and a third identifier corresponding to an LSTM time series sub-model M3 are used as sample labels. This training results in a K-means model that can be used for classification in a fusion sub-model. Each prediction result obtained based on this K-means model includes a confidence list Z = (Z1, Z2, Z3) corresponding to the first, second, and third identifiers. Here, Zi represents the confidence corresponding to Mi, indicating the degree of trustworthiness of Mi; i = 1, 2, or 3. Furthermore, each confidence list corresponds to a preset weight set, and the mapping between the weight set and the identifier is determined using a preset weight mapping relationship (e.g., a preset configuration table). Therefore, based on each identifier, the weight set corresponding to each identifier can be determined from the preset configuration table. This determines the weights corresponding to the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3, ultimately achieving fusion.

[0065] In a feasible embodiment, when the three confidence levels satisfy the weight constraint relationship, the three confidence levels can be directly used as corresponding weight groups. For example, the confidence levels Z1, Z2, and Z3 corresponding to the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM timing sub-model M3 are 70%, 20%, and 10%, respectively. At this time, the three confidence levels satisfy the weight constraint relationship, and the three confidence levels can be directly used as weight groups, that is, the preset weight mapping relationship is Zi=Wi. Therefore, it can be directly determined that the prediction weight W1 corresponding to the random forest regression sub-model M1 is 0.7, the prediction weight W2 corresponding to the GBDT regression sub-model M2 is 0.2, and the prediction weight W3 corresponding to the LSTM timing sub-model M3 is 0.1. Alternatively, when the three confidence levels do not satisfy the weight constraint relationship, the weight group is recalculated according to the preset mapping relationship. For example, if the confidence levels corresponding to the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3 are 70%, 30%, and 20%, respectively, then if the preset weight groups are 0.5, 0.25, and 0.25, the weights corresponding to the confidence levels greater than the preset confidence threshold can be further mapped to the maximum value in the preset weight group, that is, the prediction weight W1 corresponding to the random forest regression sub-model M1 is determined to be 0.5, and the weights corresponding to the other two confidence levels are mapped to the other two values ​​in the preset weight group, that is, the prediction weight W2 corresponding to the GBDT regression sub-model M2 is 0.25, and the prediction weight W3 corresponding to the LSTM time series sub-model M3 is 0.25. Alternatively, Z1, Z2, and Z3 are calculated to obtain W1, W2, and W3, respectively, according to the preset weight mapping relationship f(Zi)=Wi.

[0066] It should be understood that the three confidence levels in the confidence list Z may be the same or different, and this application does not impose any specific limitation.

[0067] Specifically, the random forest regression sub-model M1 is trained based on the random forest regression model. The random forest regression model is good at processing high-dimensional features and has good resistance to noise, such as sensor noise, and can prevent overfitting.

[0068] Specifically, the GBDT regression sub-model M2 is trained based on the GBDT regression model. The GBDT regression model excels at capturing nonlinear relationships and can automatically learn feature importance. For example, it can automatically capture the effect of temperature on viscosity from features of different dimensions, that is, the importance of temperature on viscosity.

[0069] Specifically, the LSTM time series submodel M3 is trained based on the LSTM time series model. The LSTM time series model excels at processing time series data, analyzing temporal changes in data and learning the dynamic evolution of features. For example, it can learn the delayed effect of rising temperature on the fluidity of a mixture.

[0070] Specifically, the prediction weight is used to measure the credibility of the prediction results of each model.

[0071] It can be seen that by inputting the mixing parameters into the random forest regression sub-model M1, the GBDT regression sub-model M2 and the LSTM time series sub-model M3 respectively, the prediction results (the first, second and third predicted pressure-vibration control parameters) under three different modeling perspectives are obtained, and then the prediction weights of each sub-model are dynamically calculated in combination with a fusion sub-model, and finally the weighted fusion is used to generate high-precision target pressure-vibration control parameters. Specifically, the present application predicts the pressure-vibration control parameters by using three different sub-models respectively, and can fully consider and predict the different properties of the mixture using the three different sub-models. For example, the random forest regression sub-model M1 is good at processing nonlinear static features to capture the structured data in the mixing data, the high precision and interpretability of the GBDT regression sub-model M2 are used to refine the key features in the mixing data, and the LSTM time series sub-model M3 is good at processing time series data to capture the dynamic change trend in the mixing data.

[0072] Furthermore, when the fusion sub-models fuse the predicted pressure-vibration control parameters, they can determine the confidence level of each sub-model's predicted pressure-vibration control parameters based on the mixing parameters, given the material's blended state as represented by the current mixing parameters, thereby performing the fusion. This ensures highly reliable predicted pressure-vibration control parameters for pressure-vibration control while fully preserving the impact of the components of interest from the other two sub-models on pressure-vibration control. Furthermore, in some cases, when mixing materials with different basic properties, a regression sub-model may have extremely high matching prediction capabilities. This effectively avoids the problem of using other single-type sub-models failing to meet the prediction requirements, thereby ensuring that the pre-set pressure-vibration control parameter prediction model possesses excellent adaptability and robustness. Compared to a single model, this multi-model fusion strategy effectively improves model prediction accuracy and generalization, reducing model bias and the risk of overfitting. Furthermore, the sub-models in this pre-set pressure-vibration control parameter prediction model are independent of each other, facilitating independent training, updating, or replacement, significantly enhancing the model's scalability. It also facilitates independent maintenance of each model.

[0073] It can be seen that by reusing the stirring feature vector generated in the aforementioned step 1301 and directly inputting the stirring feature vector into the fusion sub-model, the stirring data is input into the fusion sub-model, and the model is avoided from performing feature extraction again. This reduces the data processing process of the model, reduces the model calculation complexity, improves the data processing efficiency of the model, and saves the model's data processing resources. Optionally, during the implementation process, the stirring feature vector extraction module can also be used as the input layer of the preset pressure-vibration control parameter prediction model, that is, the preset pressure-vibration control parameter prediction model completes the process of stirring feature vector extraction, expected standard judgment, and determination of target pressure-vibration control parameters as a whole, thereby improving the efficiency of determining the target pressure-vibration control parameters and reducing the complexity of multi-module interaction. In addition, the stirring feature vector in the aforementioned step 1301 is a feature vector extracted based on specific domain knowledge. Compared with the feature vector obtained by using the fusion sub-model to extract features from the stirring data, the stirring feature vector in the aforementioned step 1301 is more accurate, thereby making the fourth prediction weight W1, fifth prediction weight W2 and sixth prediction weight W3 obtained by the fusion sub-model based on the more accurate stirring feature vector more accurate. Furthermore, the accuracy of the target pressure-vibration control parameters obtained based on the highly accurate fourth prediction weight W1, fifth prediction weight W2 and sixth prediction weight W3 is also higher.

[0074] In some examples, the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, and the third predicted pressure-vibration control parameter are multidimensional parameter data, and the first predicted weight, the second predicted weight, and the third predicted weight respectively include a number of first predicted sub-weights, a number of second predicted sub-weights, and a number of third predicted sub-weights that correspond one-to-one to the multidimensional parameter data, and the number of first predicted sub-weights, the number of second predicted sub-weights, and the number of third predicted sub-weights satisfy a first weight constraint relationship.

[0075] In other examples, the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, and the third predicted pressure-vibration control parameter are multidimensional parameter data, and the fourth predicted weight W1, the fifth predicted weight W2, and the sixth predicted weight W3 respectively include a number of fourth prediction sub-weights, a number of fifth prediction sub-weights, and a number of sixth prediction sub-weights that correspond one-to-one to the multidimensional parameter data, and the number of fourth prediction sub-weights, the number of fifth prediction sub-weights, and the number of sixth prediction sub-weights satisfy the second weight constraint relationship.

[0076] The pressure-vibration control parameters include a target pressure-vibration control parameter, a first predicted pressure-vibration control parameter, a second predicted pressure-vibration control parameter, a third predicted pressure-vibration control parameter, a first-stage predicted pressure-vibration control parameter, a second-stage predicted pressure-vibration control parameter, and a third-stage predicted pressure-vibration control parameter. Specifically, the multidimensional parameter data of the pressure-vibration control parameters includes pressure value, increase rate, dwell time, vibration frequency, and vibration amplitude.

[0077] Specifically, the first weight constraint relationship is equal to 1, that is, for the same dimensional parameter, the sum of the three prediction weights corresponding to the three different regression sub-models is equal to 1. In other words, the sum of the weights corresponding to the parameter data of the same dimension in the prediction results of each model in the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3 is equal to 1. Exemplarily, taking the multidimensional parameter data as pressure value, increase rate, holding time, vibration frequency and vibration amplitude as an example, the prediction weight includes a pressure weight for the pressure value, a rate weight for the increase rate, a time weight for the holding time, a frequency weight for the vibration frequency, and an amplitude weight for the vibration amplitude number; the pressure weight for the pressure value included in the first prediction weight is denoted as a11, the rate weight for the increase rate is denoted as a12, the time weight for the holding time is denoted as a13, the frequency weight for the vibration frequency is denoted as a14, and the amplitude weight for the vibration amplitude number is denoted as a15. The second prediction weight includes the pressure weight for the pressure value as a21, the rate weight for the increase rate as a22, the time weight for the holding time as a23, the frequency weight for the vibration frequency as a24, and the amplitude weight for the vibration amplitude as a25. The third prediction weight includes the pressure weight for the pressure value as a31, the rate weight for the increase rate as a32, the time weight for the holding time as a33, the frequency weight for the vibration frequency as a34, and the amplitude weight for the vibration amplitude as a35. The prediction sub-weights included in the first prediction weight, the second prediction weight, and the third prediction weight are expressed in the form of a matrix as follows:

[0078] , (i=1, 2, 3; j=1, 2, 3, 4, 5)

[0079] Among them, i=1 means that the corresponding weight is the weight for the prediction result of the random forest regression sub-model M1, i=2 means that the corresponding weight is the weight for the prediction result of the GBDT regression sub-model M2, and i=3 means that the corresponding weight is the weight for the prediction result of the LSTM timing sub-model M3. The weight corresponding to j=1 is the weight for the pressure value, the weight corresponding to j=2 is the weight for the increase rate, the weight corresponding to j=3 is the weight for the holding time, the weight corresponding to j=4 is the weight for the vibration frequency, and the weight corresponding to j=5 is the weight for the vibration amplitude number. Specifically, the sum of the weights of each column in Aij satisfies the first weight constraint, that is, the sum of the weights of each column is equal to 1. Specifically, it is expressed as Ai1=a21+a22+a31=1; Ai2=a12+a22+a32=1; Ai3=a13+a23+a33=1; Ai4=a14+a24+a34=1; Ai5=a15+a25+a35=1.

[0080] Specifically, the second weight constraint can be 1 or the difference between the stirring feature vector and the expected standard stirring feature vector. That is, for the same dimension parameter, the sum of the three prediction weights corresponding to the three different regression sub-models is equal to 1 or equal to the difference. In other words, the sum of the weights corresponding to the parameter data of the same dimension in the prediction results of each model in the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3 is equal to 1 or equal to the difference.

[0081] In some examples, the compression-vibration control parameters include stage compression-vibration control parameters for multiple compression stages. For any dimension of parameter data, the stage parameter data for multiple compression stages are included. For any dimension of parameter data, the weight of each stage parameter data is equal to the predictor weight corresponding to the parameter data.

[0082] In some examples, regarding the aforementioned method one, vibration suppression includes multiple suppression stages, the first predicted pressure-vibration control parameter includes the first predicted pressure-vibration control parameter of the multiple suppression stages, the second predicted pressure-vibration control parameter includes the second predicted pressure-vibration control parameter of the multiple suppression stages, and the third predicted pressure-vibration control parameter includes the third predicted pressure-vibration control parameter of the multiple suppression stages. A fusion sub-model is used to generate target pressure-vibration control parameters based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the first predicted weight, the second predicted weight, and the third predicted weight, including: for each suppression stage in the multiple suppression stages, a fusion sub-model is used to generate a stage target pressure-vibration control parameter based on the first predicted pressure-vibration control parameter of the stage, the second predicted pressure-vibration control parameter of the stage, the third predicted pressure-vibration control parameter, the first predicted weight, the second predicted weight, and the third predicted weight; and the target pressure-vibration control parameters of each stage are combined into a target pressure-vibration control parameter.

[0083] Among them, the first predicted pressure-vibration control parameters of each stage are predicted based on the random forest regression sub-model M1, and the first predicted pressure-vibration control parameters of multiple pressing stages constitute the first predicted pressure-vibration control parameters.

[0084] The second predicted pressure-vibration control parameters of each stage are predicted based on the GBDT regression sub-model M2, and the second predicted pressure-vibration control parameters of multiple pressing stages constitute the second predicted pressure-vibration control parameters.

[0085] The third predicted pressure-vibration control parameters of each stage are predicted based on the LSTM time series sub-model M3, and the third predicted pressure-vibration control parameters of multiple pressing stages constitute the third predicted pressure-vibration control parameters.

[0086] In some examples, regarding the aforementioned method 2, vibration suppression includes multiple suppression stages, the first predicted pressure-vibration control parameter includes the first predicted pressure-vibration control parameter of the multiple suppression stages, the second predicted pressure-vibration control parameter includes the second predicted pressure-vibration control parameter of the multiple suppression stages, and the third predicted pressure-vibration control parameter includes the third predicted pressure-vibration control parameter of the multiple suppression stages. The target pressure-vibration control parameter is generated based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the fourth predicted weight, the fifth predicted weight, and the sixth predicted weight using the fusion sub-model, including: for each suppression stage in the multiple suppression stages, the stage target pressure-vibration control parameter is generated based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the fourth predicted weight, the fifth predicted weight, and the sixth predicted weight using the fusion sub-model; and the target pressure-vibration control parameters of each stage are combined into the target pressure-vibration control parameter.

[0087] In some examples, the pressure-vibration control parameter is multi-dimensional parameter data, and the prediction weight includes prediction sub-weights corresponding one-to-one to the multi-dimensional parameter data, as well as prediction sub-weights for any dimensional parameter data.

[0088] The prediction weights include a first prediction weight, a second prediction weight, a third prediction weight, a fourth prediction weight, a fifth prediction weight, and a sixth prediction weight. The aforementioned method of using the fusion sub-model to generate the stage target pressure-vibration control parameter based on the first stage predicted pressure-vibration control parameter, the second stage predicted pressure-vibration control parameter, the third stage predicted pressure-vibration control parameter, the first prediction weight, the second prediction weight, and the third prediction weight includes: for each dimension of the multidimensional parameter data, obtaining the first product of the parameter data of the dimension in the first stage predicted pressure-vibration control parameter and the first prediction sub-weight corresponding to the dimension; obtaining the second product of the parameter data of the dimension in the second stage predicted pressure-vibration control parameter and the second prediction sub-weight corresponding to the dimension; obtaining the third product of the parameter data of the dimension in the third stage predicted pressure-vibration control parameter and the third prediction sub-weight corresponding to the dimension; taking the sum of the first product, the second product, and the third product as the stage target sub-pressure-vibration control parameter of the dimension; and combining the stage target sub-pressure-vibration control parameters of each dimension into the stage target pressure-vibration control parameter.

[0089] For example, taking the calculation of the target pressure value for the first pressing stage of vibration suppression as an example, the parameter data for the pressure dimension in the first predicted pressure-vibration control parameter corresponding to the first pressing stage is 60, the parameter data for the pressure dimension in the second predicted pressure-vibration control parameter corresponding to the first pressing stage is 80, and the parameter data for the pressure dimension in the third predicted pressure-vibration control parameter corresponding to the first pressing stage is 70. The first prediction sub-weight corresponding to the pressure dimension in the first prediction weight is 0.2, the second prediction sub-weight corresponding to the pressure dimension in the second prediction weight is 0.4, and the third prediction sub-weight corresponding to the pressure dimension in the third prediction weight is 0.4. The calculated value of 60×0.2+80×0.4+70×0.4 is 72, so the target pressure value is 72.

[0090] The above-mentioned fusion sub-model is used to generate the stage target pressure and vibration control parameters based on the first stage predicted pressure and vibration control parameters, the second stage predicted pressure and vibration control parameters, the third stage predicted pressure and vibration control parameters, the fourth predicted weight, the fifth predicted weight and the sixth predicted weight, including: for each dimension parameter data in the multidimensional parameter data, obtaining the fourth product of the parameter data of the dimension in the first stage predicted pressure and vibration control parameters and the fourth predicted sub-weight corresponding to the dimension; obtaining the fifth product of the parameter data of the dimension in the second stage predicted pressure and vibration control parameters and the fifth predicted sub-weight corresponding to the dimension; obtaining the sixth product of the parameter data of the dimension in the third stage predicted pressure and vibration control parameters and the sixth predicted sub-weight corresponding to the dimension; taking the sum of the fourth product, the fifth product and the sixth product as the stage target sub-pressure and vibration control parameter of the dimension; and combining the stage target sub-pressure and vibration control parameters of each dimension into the stage target pressure and vibration control parameter.

[0091] In some examples, in order to improve the readability and model processing efficiency of the model, the present solution sets a stirring parameter preprocessing method, specifically: with respect to the aforementioned method one and method two, before the stirring parameters are respectively input into the random forest regression sub-model M1, the GBDT regression sub-model M2 and the LSTM timing sub-model M3, the method also includes: for the infrared temperature image data in the stirring parameters, extracting the temperature characteristic parameter data of the infrared temperature image data based on the CNN feature extraction method; splicing the temperature characteristic parameter data with the parameter data in the stirring parameters to obtain spliced ​​data; and using the spliced ​​data as the stirring parameters respectively input into the random forest regression model, the GBDT regression model and the LSTM timing model.

[0092] Specifically, parameter data refers to numerical data with clear physical meaning collected through sensors, instruments, control systems, etc., which are usually quantifiable engineering parameters.

[0093] It can be seen that by converting the data in the stirring parameters into parameter data, the model can more easily recognize the information in the stirring parameters, which simplifies the model processing process and improves the model processing efficiency.

[0094] Step 1304: If the condition is met, the preset pressure-vibration control parameter corresponding to the expected standard stirring eigenvector is obtained and used as the target pressure-vibration control parameter.

[0095] Specifically, the preset pressure-vibration control parameters include stage preset pressure-vibration control parameters of multiple pressing stages, and the preset parameter data of each stage includes multi-dimensional parameter data.

[0096] For example, taking three pressing stages as an example, the preset pressure-vibration control parameters include the target pressure-vibration control parameters of stage one, the target pressure-vibration control parameters of stage two, and the target pressure-vibration control parameters of stage three. The target pressure-vibration control parameters are composed based on the target pressure-vibration control parameters of the three stages, wherein for the same dimension, the parameter data of each stage may be the same or different.

[0097] Step 140 : pouring the mixed material into a mold, and vibrating and pressing the mixed material in the mold based on target pressure-vibration control parameters to obtain an initial quartz stone plate.

[0098] Specifically, vibration compaction is a process for compacting powdered or granular materials. The material is placed in a mold, with a pressure plate above and a vibrator below. This method combines vibration and pressure, improving the uniformity of the mixture and the removal of air, shortening the compaction time and enhancing the compaction effect.

[0099] In some examples, the mixture is vibrated and compacted under vacuum.

[0100] In some examples, in order to ensure the accuracy of vibration suppression, the method further provides a feedback mechanism, which includes two feedback modes, see Mode 3 and Mode 4.

[0101] Method three: Before vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters, and during the process of vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters, the method also includes: obtaining the stirring state of the mixture, and determining the PID control parameters and filtering parameters in the vibration pressing process, as well as the target pressing state corresponding to the target pressure-vibration control parameters based on the stirring state and the target pressure-vibration control parameters; during the vibration pressing process, obtaining the actual pressing state of the mixture in real time, and filtering the actual pressing state using the filtering parameters to obtain a corrected pressing state; based on the corrected pressing state, the target pressing state and the PID control parameters, correcting the above-mentioned target pressure-vibration control parameters to obtain the corrected target pressure-vibration control parameters; and using the corrected target pressure-vibration control parameters to implement vibration pressing on the mixture.

[0102] Specifically, the stirring state refers to the state of the mixture after being poured into the mold, including viscosity, fluidity, and uniformity. The stirring state data includes the viscosity, uniformity, and fluidity of the mixture after stirring.

[0103] Specifically, the target pressing state refers to an ideal pressing state after the mixed material in a stirring state is pressed using target pressure-vibration control parameters.

[0104] Specifically, the filtering parameters are used to remove sensor information noise, eliminate noise in the real-time pressing state, and obtain a corrected pressing state that can more truly reflect the actual pressing state of the mixture.

[0105] In some examples, based on the revised pressing state, the target pressing state, and the PID control parameter, the target pressure-vibration control parameter is revised to obtain the revised target pressure-vibration control parameter, including: analyzing the revised pressing state and the target pressing state using a PID controller to obtain an error between the revised pressing state and the target pressing state, determining a first pressure-vibration control parameter difference based on the error, and adding the first pressure-vibration control parameter difference to the target pressure-vibration control parameter to obtain the revised target pressure-vibration control parameter. The first pressure-vibration control parameter difference includes parameters for multiple pressing stages, and the parameters for each pressing stage include parameters for multiple dimensions, and the dimensions and stages of the data included in the first pressure-vibration control parameter difference correspond one-to-one to the dimensions and stages of the data included in the target pressure-vibration control parameter.

[0106] Method 4: Before vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters, and during the process of vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters, the method also includes: obtaining the stirring state of the mixture, and determining the PID control parameters and filtering during the vibration pressing process based on the stirring state and the target pressure-vibration control parameters; during the vibration pressing process, obtaining the actual pressure-vibration control parameters of the mixture in real time, and filtering the actual pressure-vibration control parameters using the filtering parameters to obtain corrected pressure-vibration control parameters; correcting the above-mentioned target pressure-vibration control parameters based on the corrected pressure-vibration control parameters, the target pressure-vibration control parameters and the PID control parameters to obtain corrected target pressure-vibration control parameters; and using the corrected target pressure-vibration control parameters to implement vibration pressing on the mixture.

[0107] In some examples, based on the revised pressure-vibration control parameter, the target pressure-vibration control parameter, and the PID control parameter, the target pressure-vibration control parameter is revised to obtain the revised target pressure-vibration control parameter, including: determining the difference between the revised pressure-vibration control parameter and the target pressure-vibration control parameter using a PID controller to obtain a second pressure-vibration control parameter difference; and adding the second pressure-vibration control parameter difference to the target pressure-vibration control parameter to obtain the revised target pressure-vibration control parameter. The second pressure-vibration control parameter difference includes parameters for multiple pressing stages, and the parameters for each pressing stage include parameters for multiple dimensions, and the dimensions and stages of the data included in the second pressure-vibration control parameter difference correspond one-to-one to the dimensions and stages of the data included in the target pressure-vibration control parameter.

[0108] In some examples, regarding the aforementioned method three and the aforementioned method four, the PID control parameters and filtering parameters in the vibration pressing process are determined based on the stirring state and the target pressure-vibration control parameters, including: obtaining the target stirring state corresponding to the target pressure-vibration control parameters; inputting the stirring state and the target stirring state into the PID control parameter prediction model to obtain the PID control parameters; inputting the stirring state and the target stirring state into the filtering control parameter prediction model to obtain the filtering control parameters.

[0109] Among them, if the stirring state is closer to the target stirring state, the PID control parameter is closer to the preset standard value, and the correction amount of the PID control parameter is smaller.

[0110] As can be seen, PID control parameters are obtained based on the real-time actual pressing state (actual pressure-vibration control parameters) and the target pressing state (target pressure-vibration control parameters), and the target pressure-vibration control parameters are promptly corrected, gradually approaching the actual pressing state (actual pressure-vibration control parameters) and the target pressing state (target pressure-vibration control parameters), thereby improving the product quality of the pressed quartz slabs. The real-time collected pressing state (pressure-vibration control parameters) is filtered using filtering parameters to eliminate noise interference, making the actual pressing state (actual pressure-vibration control parameters) more accurate. Furthermore, before pressing, the PID control parameters and filtering used during the vibration pressing process are determined based on the mixing state of the mixture and the target pressure-vibration control parameters. This fully considers the actual mixing state of the mixture and reduces the deviation between the actual pressing state (actual pressure-vibration control parameters) and the target pressing state (target pressure-vibration control parameters).

[0111] Step 150 : The initial quartz stone plate is subjected to heating and curing, demoulding, grinding and polishing, and cutting to obtain a quartz stone plate.

[0112] In some examples, in the aforementioned step 150, the initial quartz stone slab is subjected to heating and curing, demolding, grinding and polishing, and cutting to obtain a quartz stone slab, including: heating and curing the initial quartz stone slab using preset heating and curing conditions to obtain a cured quartz stone slab; demolding the cured quartz stone slab to obtain a demolded quartz stone slab; grinding and polishing the surface smoothness of the demolded quartz stone slab based on a preset smoothness requirement to obtain a smooth quartz stone slab; cutting the smooth quartz stone slab according to a preset size to obtain a quartz stone slab.

[0113] In summary, this solution is to add at least two materials required for manufacturing quartz stone slabs into a stirring tank according to a preset ratio, and use a stirrer to stir at least two materials in the stirring tank to obtain a mixture; during the stirring process, stirring parameters are collected according to a preset collection frequency, and the stirring parameters include torque data of the stirrer, power data of the stirrer, and infrared temperature image data of the stirring tank; based on the stirring parameters, target pressure-vibration control parameters for pressing quartz stone slabs that meet the target product quality are determined; the mixture is poured into a mold, and the mixture in the mold is vibrated and pressed based on the target pressure-vibration control parameters to obtain an initial quartz stone slab; the initial quartz stone slab is heated and cured, demolded, polished, and cut to obtain a quartz stone slab. By obtaining the stirring parameters during the stirring process, target pressure-vibration control parameters suitable for pressing out quartz stone slabs that meet the target product quality are determined based on the stirring state. Specifically, the stirring parameters that need to be predicted are predicted based on a preset pressure-vibration control parameter prediction model, so that the mixture is pressed based on the accurate target pressure-vibration control parameters. This solves the problem of uneven product quality of the pressed quartz stone slabs caused by pressing the stirring states of different mixtures based on the same pressure-vibration control parameters, thereby ensuring the product quality of the manufactured quartz stone slabs. In addition, the stirring parameters that need to be predicted are predicted based on the preset pressure-vibration control parameter prediction model to obtain the target pressure-vibration control parameters, fully utilizing the learning ability and accuracy of the model to achieve adaptive and rapid acquisition of the target pressure-vibration control parameters that match the stirring state corresponding to the stirring parameters. The stirring state is determined using the stirring parameters instead of directly using the stirring state data because the stirring parameters do not require a sampling device to be set up in the tank and have better sampleability than the stirring state data. Moreover, the stirring parameters can accurately determine the stirring state through the real-time stirring torque data and power data of the stirrer, avoiding the problem of inaccurate stirring state data collection due to insufficient and uneven stirring of the mixture after stirring and unreasonable selection of sampling points.

[0114] Furthermore, the preset pressure-vibration control parameter prediction model predicts the required mixing parameters and obtains the target pressure-vibration control parameters. The model leverages the data modeling capabilities of the random forest regression sub-model M1, the GBDT regression sub-model M2, and the LSTM time series sub-model M3. The model predicts the pressure-vibration control parameters in different dimensions. The fusion sub-model dynamically assigns prediction weights to the prediction values ​​of each model, ultimately outputting the fused, high-precision target pressure-vibration control parameters, enabling intelligent and adaptive control of the pressing process. This overcomes the limitations of a single model and improves the predictive robustness and generalization capabilities of the preset pressure-vibration control parameter prediction model. By encoding the mixing parameters, the system quickly determines whether the mixing state of the mixture after mixing meets the expected state. The mixing parameters of mixtures that do not meet the expected state are then fed into the preset pressure-vibration control parameter prediction model for analysis to obtain the target pressure-vibration control parameters. By reusing the mixing parameters, the system saves computational resources that would otherwise be spent on model analysis, improving the efficiency of quartz stone slab manufacturing. The system accurately determines the mixing parameters required for model analysis, thereby improving the quality of quartz slabs. PID control parameters are derived based on the real-time acquired actual pressing state (actual pressure-vibration control parameters) and the target pressing state (target pressure-vibration control parameters). The target pressure-vibration control parameters are then promptly corrected, gradually aligning the actual pressing state (actual pressure-vibration control parameters) with the target pressing state (target pressure-vibration control parameters), thereby improving the quality of the pressed quartz slabs. The real-time acquired pressing state (pressure-vibration control parameters) is filtered using filtering parameters to eliminate noise interference and achieve a more accurate actual pressing state (actual pressure-vibration control parameters). Before pressing, the PID control parameters and filtering used during the vibration pressing process are determined based on the mixing state of the mixture and the target pressure-vibration control parameters. This fully considers the actual mixing state of the mixture and reduces the deviation between the actual pressing state (actual pressure-vibration control parameters) and the target pressing state (target pressure-vibration control parameters).

[0115] This solution utilizes mixing data to determine the target pressure-vibration control parameters during the pressing phase, and uses the mixing state of the mixture to determine the PID control parameters and filtering during the pressing phase. This allows for the selection of different sampling data at different stages, fully considering the characteristics of different stages in the quartz stone slab manufacturing process. This ensures both the reliability of the target pressure-vibration control parameters determined before pressing and the reliability of the pressing results during the pressing process.

[0116] Example 2:

[0117] The quartz stone plate manufacturing method of this scheme is applicable to the quartz stone plate manufacturing system, which can be found in Figure 5The quartz stone slab manufacturing system 500 includes a collection device 51, a control device 52, a stirring device 53, a vibration pressing device 54, a heating and curing device 55, a demoulding device 56, a grinding and polishing device 57 and a cutting device 58, wherein the control device is respectively connected to the collection device 51, the stirring device 53, the vibration pressing device 54, the heating and curing device 55, the demoulding device 56, the grinding and polishing device 57 and the cutting device 58.

[0118] The control device 52 is configured to receive at least two materials collected by the collection device 51 according to a preset ratio. The control device 52 is configured to add the at least two materials to the stirring tank and control the stirring device 53 to stir the at least two materials. While controlling the stirring device 53 to stir the at least two materials, the control device 52 is also configured to receive stirring parameters collected by the collection device 51 according to a preset frequency. The control device 52 is configured to determine target pressure-vibration control parameters based on the received stirring parameters. After determining the target pressure-vibration control parameters, the control device 52 controls the vibration pressing device 54 to vibrate and press the mixture in the mold based on the target pressure-vibration control parameters to produce an initial quartz stone slab. Before the vibration pressing device 54 vibrates and presses the mixture in the mold based on the target pressure-vibration control parameters, the control device 52 is also configured to receive the stirring state of the mixture collected by the collection device 51 and, based on the stirring state and the target pressure-vibration control parameters, determine the PID control parameters and filtering parameters for the vibration pressing process, as well as the target pressing state corresponding to the target pressure-vibration control parameters. While controlling the vibration pressing device 54 to vibrate and press the mixed material in the mold based on the target pressure-vibration control parameters, the control device 52 receives the actual pressing state of the mixed material collected in real time by the acquisition device 51. The control device 52 filters the actual pressing state of the mixed material using the filter parameters to obtain a corrected pressing state. Based on the corrected pressing state, the target pressing state, and the PID control parameters, the control device 52 corrects the target pressure-vibration control parameters to obtain the corrected target pressure-vibration control parameters. The control device 52 controls the vibration pressing device 54 to vibrate and press the mixed material in the mold based on the corrected target pressure-vibration control parameters to obtain an initial quartz stone slab. The control device 52 controls the heating and curing device 55 to heat and cure the initial quartz stone slab to obtain a cured quartz stone slab. The control device 52 controls the demolding device 56 to demold the cured quartz stone slab to obtain a demolded quartz stone slab. The control device 52 controls the grinding and polishing device 57 to grind and polish the demolded quartz stone slab to obtain a smooth quartz stone slab. The control device 52 controls the cutting device 58 to cut the smooth quartz stone plate to obtain a quartz stone plate.

[0119] The control device 52 is used to implement the contents of the aforementioned quartz stone plate manufacturing method. The acquisition device 51 is used to acquire stirring parameters, the stirring state of the mixture, the actual pressing state, the actual pressure vibration control parameters, and other acquired data required by the control device.

[0120] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A method for manufacturing a quartz stone plate, characterized in that: include: Adding at least two materials required for manufacturing the quartz stone plate into a stirring tank according to a preset ratio, and stirring the at least two materials in the stirring tank with a stirrer to obtain a mixture; During the stirring process, stirring parameters are collected according to a preset collection frequency, wherein the stirring parameters include torque data of the stirrer, power data of the stirrer, and infrared temperature image data of the stirring tank; Based on the stirring parameters, determining target pressure-vibration control parameters for pressing a quartz stone plate that meets target product quality; pouring the mixture into a mold, and vibrating and pressing the mixture in the mold based on the target pressure-vibration control parameters to obtain an initial quartz stone plate; The initial quartz stone plate is subjected to heating and curing, demoulding, grinding and polishing, and cutting processes to obtain the quartz stone plate; Wherein, the target pressure-vibration control parameters for pressing the quartz stone plate meeting the target product quality are determined based on the stirring parameters, including: generating a stirring feature vector based on the stirring parameters; determining whether the stirring characteristic vector meets an expected standard stirring characteristic vector; If not, the target pressure-vibration control parameters are obtained based on a preset pressure-vibration control parameter prediction model and the stirring parameter prediction, wherein the preset pressure-vibration control parameter prediction model is used to predict the pressure-vibration control parameters that can be used to press out a quartz stone plate that meets the target product quality; If so, the preset pressure-vibration control parameter corresponding to the expected standard stirring eigenvector is obtained and used as the target pressure-vibration control parameter.

2. The method for manufacturing a quartz stone plate according to claim 1, characterized in that: Generating a stirring feature vector based on the stirring parameter includes: Performing feature processing on the stirring parameters based on a low-latency feature encoding method to generate a stirring feature vector; Determining whether the stirring feature vector meets the expected standard stirring feature vector includes: Calculating the similarity between the stirring feature vector and the expected standard stirring feature vector based on a preset similarity calculation method; If the similarity is less than the preset similarity threshold, it is considered as not meeting the requirements.

3. The method for manufacturing a quartz stone plate according to claim 1, wherein: The preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2, and an LSTM timing sub-model M3. The target pressure-vibration control parameter is obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: Input the stirring parameters into M1, M2 and M3 respectively to obtain the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter; Inputting the stirring parameter into the fusion sub-model to obtain a first prediction weight for a first predicted pressure-vibration control parameter, a second prediction weight for a second predicted pressure-vibration control parameter, and a third prediction weight for a third predicted pressure-vibration control parameter; The target pressure-vibration control parameter is generated by using the fusion sub-model based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the first predicted weight, the second predicted weight, and the third predicted weight.

4. The method for manufacturing a quartz stone plate according to claim 1, wherein: The preset pressure-vibration control parameter prediction model includes a fusion sub-model, a random forest regression sub-model M1, a GBDT regression sub-model M2, and an LSTM timing sub-model M3. The target pressure-vibration control parameter is obtained based on the preset pressure-vibration control parameter prediction model and the stirring parameter prediction, including: Input the stirring parameters into M1, M2 and M3 respectively to obtain the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter; Inputting the stirring feature vector into the fusion sub-model to obtain a fourth prediction weight W1 for the first predicted pressure-vibration control parameter, a fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and a sixth prediction weight W3 for the third predicted pressure-vibration control parameter; The target pressure-vibration control parameter is generated by utilizing the fusion sub-model based on the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, the third predicted pressure-vibration control parameter, the fourth predicted weight W1, the fifth predicted weight W2, and the sixth predicted weight W3.

5. The method for manufacturing a quartz stone plate according to claim 4, characterized in that: The first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter and the third predicted pressure-vibration control parameter are multidimensional parameter data, and the fourth predicted weight W1, the fifth predicted weight W2 and the sixth predicted weight W3 respectively include a number of fourth predicted sub-weights, a number of fifth predicted sub-weights and a number of sixth predicted sub-weights corresponding one-to-one to the multidimensional parameter data, and the number of the fourth predicted sub-weights, the number of the fifth predicted sub-weights and the number of the sixth predicted sub-weights satisfy the second weight constraint relationship.

6. The method for manufacturing a quartz stone plate according to claim 4, characterized in that: The fusion sub-model includes a K-means model and a processing module, and the stirring feature vector is input into the fusion sub-model to obtain a fourth prediction weight W1 for the first predicted pressure-vibration control parameter, a fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and a sixth prediction weight W3 for the third predicted pressure-vibration control parameter, including: The stirring feature vector is input into the K-means model to obtain the classification result G output by the K-means model; G is one of the first label G1, the second label G2, and the third label G3 corresponding to M1, M2, and M3 respectively; if G=Gi, it means that the credibility of Mi is greater than the credibility of the other two of M1, M2, and M3 except Mi; i=1, 2, 3; The processing module is used to determine a weight group corresponding to the classification result G according to a mapping relationship between G and a preset weight; the weight group includes a fourth prediction weight W1, a fifth prediction weight W2, and a sixth prediction weight W3 corresponding to the first predicted pressure-vibration control parameter, the second predicted pressure-vibration control parameter, and the third predicted pressure-vibration control parameter, respectively; and in the weight group corresponding to Gi, Wi is greater than the other two of W1, W2, and W3 except Wi.

7. The method for manufacturing a quartz stone plate according to claim 4, characterized in that: The fusion sub-model includes a K-means model and a processing module. The stirring feature vector is input into the fusion sub-model to obtain a fourth prediction weight W1 for the first predicted pressure-vibration control parameter, a fifth prediction weight W2 for the second predicted pressure-vibration control parameter, and a sixth prediction weight W3 for the third predicted pressure-vibration control parameter, including: Input the stirring feature vector into the K-means model to obtain the confidence list Z=(Z1, Z2, Z3) output by the K-means model; where Zi is the confidence corresponding to Mi; Zi represents the credibility of Mi; i=1, 2, 3; The processing module is used to determine the fourth prediction weight W1, the fifth prediction weight W2 and the sixth prediction weight W3 corresponding to Z1, Z2 and Z3 respectively according to the mapping relationship between Z and the preset weights.

8. The method for manufacturing a quartz stone plate according to claim 1, characterized in that: The method further comprises: Acquiring a stirring state of the mixture, and determining, based on the stirring state and the target pressure-vibration control parameter, PID control parameters and filter parameters during the vibration pressing process, as well as a target pressing state corresponding to the target pressure-vibration control parameter; During the vibration pressing process, the actual pressing state of the mixture is obtained in real time, and the actual pressing state is filtered using the filtering parameters to obtain a corrected pressing state; Based on the modified compression state, the target compression state and the PID control parameter, the target pressure-vibration control parameter is modified to obtain the modified target pressure-vibration control parameter; The mixed material is subjected to vibration compaction using the corrected target compression and vibration control parameters.

9. The method for manufacturing a quartz stone plate according to claim 8, characterized in that: The determining of the PID control parameters and the filtering parameters during the vibration pressing process based on the stirring state and the target pressure-vibration control parameters includes: Obtaining a target stirring state corresponding to the target pressure-vibration control parameter; Inputting the stirring state and the target stirring state into a PID control parameter prediction model to obtain PID control parameters; The stirring state and the target stirring state are input into a filtering control parameter prediction model to obtain filtering control parameters.

Images