A method and apparatus for predicting the digestibility of a compressed ration

Abstract

The application discloses a kind of compressed dry food digestion performance prediction method and device, the method comprises: the component information set of compressed dry food is obtained by acquisition;The component information set includes protein content information, mineral content information, starch content information and fat content information;The preset digestion performance prediction model is trained and handled, and the digestion performance prediction model that training is completed is obtained;Using the digestion performance prediction model that training is completed, the component information set of compressed dry food obtained by acquisition is fused and predicted and handled, and the digestion performance prediction information set is obtained;The digestion performance prediction information set includes protein content prediction value, amino acid content prediction value, apparent viscosity prediction value, protein digestion rate prediction value.

Description

Technical Field

[0001] This invention relates to the field of data modeling and processing in the field of food evaluation and processing, and specifically to a method and apparatus for predicting the digestibility of compressed dry food. Background Technology

[0002] Compressed rations, as an important emergency food and essential supplies for outdoor activities, are widely used in military, rescue, and field exploration due to their high energy density and long shelf life. However, the digestibility of compressed rations directly affects their nutrient absorption efficiency and the gastrointestinal health of consumers. Accurately predicting their digestibility is crucial for optimizing product formulation and improving product quality. Currently, traditional assessments of the digestibility of compressed rations largely rely on animal experiments or human trials. These methods are not only time-consuming and costly but also highly susceptible to individual differences, making it difficult to quickly and accurately obtain digestibility indicators. Furthermore, existing prediction methods based on single components or simple models cannot fully consider the interactions between the complex components of compressed rations, resulting in low prediction accuracy that fails to meet the needs of actual production and research and development. Therefore, there is an urgent need for an efficient and accurate method for predicting the digestibility of compressed rations to promote the optimization and upgrading of compressed ration products. Summary of the Invention

[0003] This invention primarily addresses the problem of how to achieve efficient and accurate prediction of the digestibility of compressed dry food. This invention discloses a method and apparatus for predicting the digestibility of compressed dry food.

[0004] In a first aspect, this invention discloses a method for predicting the digestibility of compressed dry rations, comprising: S1, Collect the set of component information of the compressed dry food; the set of component information includes protein content information, mineral content information, starch content information and fat content information; S2, train the preset digestive performance prediction model to obtain the trained digestive performance prediction model; S3. Using the trained digestibility prediction model, the collected composition information set of compressed dry food is fused and predicted to obtain a digestibility prediction information set; the digestibility prediction information set includes predicted values ​​of protein content, amino acid content, apparent viscosity, and protein digestibility.

[0005] The preset digestion performance prediction model includes: a first input module, a first feature extraction module, a second feature extraction module, a second input module, a dimension transformation module, a selected feature fusion module, and a fully connected layer module; The input terminal of the first input module is used to input the protein content information and mineral content information from the component information set; the input terminal of the second input module is used to input the starch content information and fat content information from the component information set; the output terminal of the first input module is connected to the input terminal of the first feature extraction module; the output terminal of the second input module is connected to the input terminal of the second feature extraction module; the output terminal of the first feature extraction module is connected to the first input terminal of the dimension transformation module; the output terminal of the second feature extraction module is connected to the second input terminal of the dimension transformation module; the output terminal of the dimension transformation module is connected to the input terminal of the selected feature fusion module; the output terminal of the selected feature fusion module is connected to the input terminal of the fully connected layer module. The dimension transformation module is used to transform the output vectors of the first feature extraction module and the second feature extraction module into matrices that match the input dimension of the feature selection fusion module, and then concatenate the two matrices to obtain the input tensor of the feature selection fusion module. The output of the fully connected layer module is used to output a digestibility prediction information vector; the digestibility prediction information vector includes: predicted protein content, predicted amino acid content, predicted apparent viscosity, and predicted protein digestibility.

[0006] The first feature extraction module and the second feature extraction module both include an input layer, a first convolutional neural network, a second convolutional neural network, a first max pooling layer, a third convolutional neural network, a second max pooling layer, a fourth convolutional neural network, a global average pooling layer, a random dropout layer, and a normalized exponential function layer; The normalized exponential function layer is used to implement the calculation of the normalized exponential function; The first, second, third, and fourth convolutional neural networks are all implemented using one-dimensional convolutional neural networks; The output of the input layer is connected to the input of the first convolutional neural network; the output of the first convolutional neural network is connected to the input of the second convolutional neural network; the output of the second convolutional neural network is connected to the input of the first max pooling layer; the output of the first max pooling layer is connected to the input of the third convolutional neural network; the output of the third convolutional neural network is connected to the input of the second max pooling layer; the output of the second max pooling layer is connected to the input of the fourth convolutional neural network; the output of the fourth convolutional neural network is connected to the input of the global average pooling layer; the output of the global average pooling layer is connected to the input of the random dropout layer; and the output of the random dropout layer is connected to the input of the normalized exponential function layer.

[0007] The selected feature fusion module includes a first fusion module and a second fusion module; the input terminal of the first fusion module serves as the input terminal of the selected feature fusion module; the output terminal of the second fusion module serves as the output terminal of the selected feature fusion module. The first fusion module includes an input layer, a first average pooling layer, a first max pooling layer, a first fully connected layer, an addition module, a first activation function module, a first multiplication module, and an output layer; The input layer is used to receive the input from the first fusion module; the first output of the input layer is connected to the input of the first average pooling layer; the second output of the input layer is connected to the input of the first max pooling layer; the output of the first average pooling layer is connected to the first input of the first fully connected layer; the output of the first max pooling layer is connected to the second input of the first fully connected layer; the output of the first fully connected layer is connected to the input of the addition module; the output of the addition module is connected to the input of the first activation function module; the output of the first activation function module is connected to the first input of the first multiplication module; the third output of the input layer is connected to the second input of the first multiplication module; and the output of the first multiplication module is connected to the input of the output layer. The first multiplication module is used to multiply the input quantity at its first input terminal and the input quantity at its second input terminal; the first activation function module is used to perform activation function operation on its input quantity; the addition module is used to sum the two output quantities of the first fully connected layer. The output quantities of the three output terminals of the input layer of the first fusion module are obtained by copying the input quantity received by the input layer into three copies and then outputting them from the three output terminals respectively.

[0008] The second fusion module includes an input layer, a second average pooling layer, a second max pooling layer, a splicing layer, a fifth convolutional neural network, a second fully connected layer, and an output layer; The input layer is used to receive the input from the second fusion module; the first output of the input layer is connected to the input of the second average pooling layer; the second output of the input layer is connected to the input of the second max pooling layer; the output of the second average pooling layer is connected to the first input of the concatenation layer; the output of the second max pooling layer is connected to the second input of the concatenation layer; the concatenation layer is used to concatenate the vector input from the first input and the vector input from the second input to obtain a concatenation matrix, and output the concatenation matrix; the output of the concatenation layer... The input layer is connected to the input of the fifth convolutional neural network; the output of the fifth convolutional neural network is connected to the second input of the second fully connected layer; the third output of the input layer is connected to the first input of the second fully connected layer; the second fully connected layer is used to multiply the input of its first input and the input of its second input; the outputs of the three outputs of the input layer of the second fusion module are obtained by copying the input received by the input layer three times and then outputting them from the three outputs respectively; the output of the second fully connected layer is connected to the input of the output layer.

[0009] The training process of the preset digestibility performance prediction model includes: S21, Obtain the training dataset; the training dataset includes training data and corresponding label information; the training data includes protein content information, mineral content information, starch content information, and fat content information; the label information is a digestibility prediction information vector corresponding to the training data; the digestibility prediction information vector includes predicted protein content, predicted amino acid content, predicted apparent viscosity, and predicted protein digestibility. S22, The training data in the training dataset is used as input data and input into the digestive performance prediction model; S23, The input data is processed using the digestibility prediction model to obtain a prediction vector; S24, perform difference calculation processing on the obtained prediction vector and the label information corresponding to the input data to obtain the difference value; S25, determine whether the difference value satisfies the convergence condition, and obtain the first judgment result; S26, when the first judgment result is negative, determine whether the training iteration count value is equal to the training count threshold to obtain the second judgment result; When the second judgment result is negative, the model training state is determined to be that the termination training condition is not met. When the second judgment result is yes, it is determined that the model training state meets the termination training condition; When the first judgment result is yes, it is determined that the model training state meets the termination training condition; When the model training state does not meet the termination training condition, the parameters of the digestive performance prediction model are updated using the parameter update model, the training iteration number is increased by 1, and the training data in the training dataset is used as input data to the digestive performance prediction model. When the model training state meets the termination training condition, the training process of the digestive performance prediction model is completed, and the trained digestive performance prediction model is obtained.

[0010] , In the formula, For the i-th training data in the training dataset, The label information for the i-th training data in the training dataset. For loss function, Update the parameter value. The parameters of the digestive performance prediction model are... The initial parameter is the learning rate. For momentum angle parameters, Indicates the variable Find the partial derivative. This represents the predicted value obtained by the digestibility performance prediction model for the i-th training data in the training dataset. This is the calculation function corresponding to the digestibility performance prediction model; exp represents the power operation of the constant e; and This is the default value; and This is a preset weighting factor.

[0011] A second aspect of the present invention discloses a device for predicting the digestibility of compressed dry food, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for predicting the digestibility of compressed dry food.

[0012] In a third aspect of this invention, a computer-storable medium is disclosed, the computer-storable medium storing computer instructions, which, when invoked by a computer, are used to execute the method for predicting the digestibility of compressed dry food.

[0013] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the method for predicting the digestibility of compressed dry food.

[0014] The beneficial effects of this invention are as follows: The digestibility prediction method for compressed dry food disclosed in this invention has significant technical advantages and application value. First, this method collects information on core components such as protein, minerals, starch, and fat, and uses a constructed multi-module deep neural network model for fusion prediction, overcoming the limitations of traditional single-component prediction. It fully considers the synergistic and antagonistic effects between various components of compressed dry food, significantly improving the accuracy and comprehensiveness of digestibility prediction. Second, the model design employs a feature extraction module containing multiple convolutional and pooling layers, effectively extracting deep-level features from complex component data. Combined with a dimensionality transformation module and a feature fusion module, it achieves efficient fusion of features from different types of components, further enhancing the model's predictive performance. Furthermore, this method does not rely on extensive animal or human experiments; it can quickly complete digestibility prediction through a data-driven approach, significantly shortening the R&D cycle and reducing experimental costs. It provides strong technical support for the formulation optimization and product development of compressed dry food, contributing to the intelligent and efficient development of the compressed dry food industry. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention. Detailed Implementation

[0016] To better understand the content of this invention, an embodiment is provided here.

[0017] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0018] In a first aspect, this invention discloses a method for predicting the digestibility of compressed dry rations, comprising: S1, Collect the set of component information of the compressed dry food; the set of component information includes protein content information, mineral content information, starch content information and fat content information; S2, train the preset digestive performance prediction model to obtain the trained digestive performance prediction model; S3. Using the trained digestibility prediction model, the collected composition information set of compressed dry food is fused and predicted to obtain a digestibility prediction information set; the digestibility prediction information set includes predicted values ​​of protein content, amino acid content, apparent viscosity, and protein digestibility.

[0019] The preset digestibility performance prediction model includes: a first input module, a first feature extraction module, a second feature extraction module, a second input module, a dimension transformation module, a selected feature fusion module, and a fully connected layer module; The input terminal of the first input module is used to input the protein content information and mineral content information from the component information set; the input terminal of the second input module is used to input the starch content information and fat content information from the component information set; the output terminal of the first input module is connected to the input terminal of the first feature extraction module; the output terminal of the second input module is connected to the input terminal of the second feature extraction module; the output terminal of the first feature extraction module is connected to the first input terminal of the dimension transformation module; the output terminal of the second feature extraction module is connected to the second input terminal of the dimension transformation module; the output terminal of the dimension transformation module is connected to the input terminal of the selected feature fusion module; the output terminal of the selected feature fusion module is connected to the input terminal of the fully connected layer module. The dimension transformation module is used to transform the output vectors of the first feature extraction module and the second feature extraction module into matrices that match the input dimension of the feature selection fusion module, and then concatenate the two matrices to obtain the input tensor of the feature selection fusion module. The output of the fully connected layer module is used to output a digestibility prediction information vector; the digestibility prediction information vector includes: predicted protein content, predicted amino acid content, predicted apparent viscosity, and predicted protein digestibility.

[0020] The first feature extraction module and the second feature extraction module both include an input layer, a first convolutional neural network, a second convolutional neural network, a first max pooling layer, a third convolutional neural network, a second max pooling layer, a fourth convolutional neural network, a global average pooling layer, a random dropout layer, and a normalized exponential function layer; The normalized exponential function layer is used to implement the calculation of the normalized exponential function; The first, second, third, and fourth convolutional neural networks are all implemented using one-dimensional convolutional neural networks; The output of the input layer is connected to the input of the first convolutional neural network; the output of the first convolutional neural network is connected to the input of the second convolutional neural network; the output of the second convolutional neural network is connected to the input of the first max pooling layer; the output of the first max pooling layer is connected to the input of the third convolutional neural network; the output of the third convolutional neural network is connected to the input of the second max pooling layer; the output of the second max pooling layer is connected to the input of the fourth convolutional neural network; the output of the fourth convolutional neural network is connected to the input of the global average pooling layer; the output of the global average pooling layer is connected to the input of the random dropout layer; and the output of the random dropout layer is connected to the input of the normalized exponential function layer.

[0021] The selected feature fusion module includes a first fusion module and a second fusion module; the input terminal of the first fusion module serves as the input terminal of the selected feature fusion module; the output terminal of the second fusion module serves as the output terminal of the selected feature fusion module. The first fusion module includes an input layer, a first average pooling layer, a first max pooling layer, a first fully connected layer, an addition module, a first activation function module, a first multiplication module, and an output layer; The input layer is used to receive the input from the first fusion module; the first output of the input layer is connected to the input of the first average pooling layer; the second output of the input layer is connected to the input of the first max pooling layer; the output of the first average pooling layer is connected to the first input of the first fully connected layer; the output of the first max pooling layer is connected to the second input of the first fully connected layer; the output of the first fully connected layer is connected to the input of the addition module; the output of the addition module is connected to the input of the first activation function module; the output of the first activation function module is connected to the first input of the first multiplication module; the third output of the input layer is connected to the second input of the first multiplication module; and the output of the first multiplication module is connected to the input of the output layer. The first multiplication module is used to multiply the signal at its first input terminal and the signal at its second input terminal; the first activation function module is used to perform activation function operation on its input signal; the addition module is used to sum the two signals output by the first fully connected layer. The output signals of the three output terminals of the input layer of the first fusion module are obtained by copying the signal received by the input layer into three copies and then outputting them from the three output terminals respectively.

[0022] The second fusion module includes an input layer, a second average pooling layer, a second max pooling layer, a splicing layer, a fifth convolutional neural network, a second fully connected layer, and an output layer; The input layer is used to receive the input from the second fusion module; the first output of the input layer is connected to the input of the second average pooling layer; the second output of the input layer is connected to the input of the second max pooling layer; the output of the second average pooling layer is connected to the first input of the splicing layer; the output of the second max pooling layer is connected to the second input of the splicing layer; the splicing layer is used to splice the signal vectors input from the first input and the second input to obtain a splicing matrix, and output the splicing matrix; the output of the splicing layer is connected to the input of the fifth convolutional neural network; the output of the fifth convolutional neural network is connected to the second input of the second fully connected layer; the third output of the input layer is connected to the first input of the second fully connected layer; the second fully connected layer is used to multiply the signals from its first input and the signals from its second input; the output signals of the three outputs of the input layer of the second fusion module are obtained by copying the signals received by the input layer three times and then outputting them from the three outputs respectively. The output of the second fully connected layer is connected to the input of the output layer.

[0023] The training process of the preset digestibility performance prediction model includes: S21, Obtain the training dataset; the training dataset includes training data and corresponding label information; the training data includes protein content information, mineral content information, starch content information, and fat content information; the label information is a digestive performance prediction information vector corresponding to the training data; S22, The training data in the training dataset is used as input data and input into the digestive performance prediction model; S23, The input data is processed using the digestibility prediction model to obtain a prediction vector; S24, perform difference calculation processing on the obtained prediction vector and the label information corresponding to the input data to obtain the difference value; S25, determine whether the difference value satisfies the convergence condition, and obtain the first judgment result; S26, when the first judgment result is negative, determine whether the training iteration count value is equal to the training count threshold to obtain the second judgment result; When the second judgment result is negative, the model training state is determined to be that the termination training condition is not met. When the second judgment result is yes, it is determined that the model training state meets the termination training condition; When the first judgment result is yes, it is determined that the model training state meets the termination training condition; When the model training state does not meet the termination training condition, the parameters of the digestive performance prediction model are updated using the parameter update model, the training iteration number is increased by 1, and the training data in the training dataset is used as input data to the digestive performance prediction model. When the model training state meets the termination training condition, the training process of the digestive performance prediction model is completed, and the trained digestive performance prediction model is obtained.

[0024] The difference value satisfies the convergence condition when it is less than a preset convergence threshold; the difference value does not satisfy the convergence condition when it is not less than a preset convergence threshold.

[0025] The difference calculation process can be implemented using a loss function.

[0026] The loss function can be the cross-entropy loss function.

[0027] The expression for the parameter update model is: , In the formula, For the i-th training data in the training dataset, The label information for the i-th training data in the training dataset. For loss function, Update the parameter value. The parameters of the digestive performance prediction model are... The initial parameter is the learning rate. For momentum angle parameters, Indicates the variable Find the partial derivative. This represents the predicted value obtained by the digestibility performance prediction model for the i-th training data in the training dataset. This is the calculation function corresponding to the digestibility performance prediction model; exp represents the power operation of the constant e; and This is the default value; and This is a preset weighting factor.

[0028] The trained digestibility prediction model is used to fuse and predict the collected composition information set of compressed dry food to obtain a digestibility prediction information set, including: Using the trained digestibility prediction model, the set of composition information of compressed dry food collected at each time moment is processed to obtain the digestibility prediction information vector at that time moment. The digestibility performance prediction information vectors at all times are fused and processed to obtain the digestibility performance prediction information vector.

[0029] The expression for the fusion calculation process is: , in, Let j be the j-th element in the digestibility performance prediction information vector, where j ranges from 1 to 4, representing the predicted values ​​for protein content, amino acid content, apparent viscosity, and protein digestibility, respectively. This represents the j-th element of the digestive performance prediction information vector at time i, where N is the total number of time points. This represents the mean of the j-th element of the digestibility performance prediction information vector at all times.

[0030] The aforementioned fusion calculation process further enhances the technical advantages and application value of this invention. First, by processing the compressed dry food composition information collected at each time point separately, the dynamic impact of compositional changes on digestibility at different time dimensions can be captured. Compared to prediction methods relying solely on data from a single time point, this fully utilizes the potential information in time-series data, avoids prediction bias caused by compositional fluctuations, and greatly improves the fit between the prediction results and actual digestibility. Second, the fusion calculation expression uses a mean calculation method, effectively integrating prediction results from multiple time points. This reduces the interference of individual time-point data anomalies or measurement errors on the final prediction value, making indicators such as protein content prediction, amino acid content prediction, apparent viscosity prediction, and protein digestibility prediction in the digestibility prediction information vector more representative and stable, thus improving the reliability of the prediction results. This combination of dynamic prediction and fusion calculation provides a more scientific and comprehensive technical means for the accurate assessment of compressed dry food digestibility, helping manufacturers to more accurately control product quality, optimize production processes, and provide consumers with compressed dry food products that better meet their nutritional needs and digestibility characteristics.

[0031] The first to fourth elements of the digestibility prediction information vector are the predicted values ​​for protein content, amino acid content, apparent viscosity, and protein digestibility, respectively.

[0032] The apparent viscosity values ​​in the training data can be obtained by measuring the digestion fluid using an MCR102 rheometer.

[0033] The protein digestibility was determined by mixing 10% trichloroacetic acid and the vortexed digestion solution at a ratio of 1:6, followed by centrifugation at 503g for 20 minutes. The supernatant was diluted to an appropriate concentration, and the protein content was measured using a BCA kit. The protein digestibility was calculated based on the total protein content and the supernatant protein content.

[0034] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions have been dimensionlessized before computation.

[0035] In all embodiments of the present invention, the values ​​of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.

[0036] A second aspect of the present invention discloses a device for predicting the digestibility of compressed dry food, the device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for predicting the digestibility of compressed dry food.

[0037] In a third aspect of this invention, a computer-storable medium is disclosed, the computer-storable medium storing computer instructions, which, when invoked by a computer, are used to execute the method for predicting the digestibility of compressed dry food.

[0038] In a fourth aspect of this invention, an information data processing terminal is disclosed, which is used to implement the method for predicting the digestibility of compressed dry food.

[0039] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for predicting the digestibility of compressed dry rations, characterized in that, include: S1, Collect the set of component information of the compressed dry food; the set of component information includes protein content information, mineral content information, starch content information and fat content information; S2, train the preset digestive performance prediction model to obtain the trained digestive performance prediction model; S3. Using the trained digestibility prediction model, the collected composition information set of compressed dry food is fused and predicted to obtain a digestibility prediction information set; the digestibility prediction information set includes predicted values ​​of protein content, amino acid content, apparent viscosity, and protein digestibility. The preset digestion performance prediction model includes: a first input module, a first feature extraction module, a second feature extraction module, a second input module, a dimension transformation module, a selected feature fusion module, and a fully connected layer module; The input terminal of the first input module is used to input the protein content information and mineral content information from the component information set; the input terminal of the second input module is used to input the starch content information and fat content information from the component information set; the output terminal of the first input module is connected to the input terminal of the first feature extraction module; the output terminal of the second input module is connected to the input terminal of the second feature extraction module; the output terminal of the first feature extraction module is connected to the first input terminal of the dimension transformation module; the output terminal of the second feature extraction module is connected to the second input terminal of the dimension transformation module; the output terminal of the dimension transformation module is connected to the input terminal of the selected feature fusion module; the output terminal of the selected feature fusion module is connected to the input terminal of the fully connected layer module. The dimension transformation module is used to transform the output vectors of the first feature extraction module and the second feature extraction module into matrices that match the input dimension of the feature selection fusion module, and then concatenate the two matrices to obtain the input tensor of the feature selection fusion module. The output of the fully connected layer module is used to output a digestibility prediction information vector; the digestibility prediction information vector includes: predicted protein content, predicted amino acid content, predicted apparent viscosity, and predicted protein digestibility.

2. The method for predicting the digestibility of compressed dry rations as described in claim 1, characterized in that, The first feature extraction module and the second feature extraction module both include an input layer, a first convolutional neural network, a second convolutional neural network, a first max pooling layer, a third convolutional neural network, a second max pooling layer, a fourth convolutional neural network, a global average pooling layer, a random dropout layer, and a normalized exponential function layer; The normalized exponential function layer is used to implement the calculation of the normalized exponential function; The first, second, third, and fourth convolutional neural networks are all implemented using one-dimensional convolutional neural networks; The output of the input layer is connected to the input of the first convolutional neural network; the output of the first convolutional neural network is connected to the input of the second convolutional neural network; the output of the second convolutional neural network is connected to the input of the first max pooling layer; the output of the first max pooling layer is connected to the input of the third convolutional neural network; the output of the third convolutional neural network is connected to the input of the second max pooling layer; the output of the second max pooling layer is connected to the input of the fourth convolutional neural network; the output of the fourth convolutional neural network is connected to the input of the global average pooling layer; the output of the global average pooling layer is connected to the input of the random dropout layer; and the output of the random dropout layer is connected to the input of the normalized exponential function layer.

3. The method for predicting the digestibility of compressed dry rations as described in claim 2, characterized in that, The selected feature fusion module includes a first fusion module and a second fusion module; the input terminal of the first fusion module serves as the input terminal of the selected feature fusion module; the output terminal of the second fusion module serves as the output terminal of the selected feature fusion module. The first fusion module includes an input layer, a first average pooling layer, a first max pooling layer, a first fully connected layer, an addition module, a first activation function module, a first multiplication module, and an output layer; The input layer is used to receive the input from the first fusion module; the first output of the input layer is connected to the input of the first average pooling layer; the second output of the input layer is connected to the input of the first max pooling layer; and the output of the first average pooling layer is connected to the first input of the first fully connected layer. The output of the first max-pooling layer is connected to the second input of the first fully connected layer; the output of the first fully connected layer is connected to the input of the addition module; the output of the addition module is connected to the input of the first activation function module; the output of the first activation function module is connected to the first input of the first multiplication module; the third output of the input layer is connected to the second input of the first multiplication module; and the output of the first multiplication module is connected to the input of the output layer. The first multiplication module is used to multiply the input quantity at its first input terminal and the input quantity at its second input terminal; the first activation function module is used to perform activation function operation on its input quantity; the addition module is used to sum the two output quantities of the first fully connected layer. The output quantities of the three output terminals of the input layer of the first fusion module are obtained by copying the input quantity received by the input layer into three copies and then outputting them from the three output terminals respectively.

4. The method for predicting the digestibility of compressed dry rations as described in claim 3, characterized in that, The second fusion module includes an input layer, a second average pooling layer, a second max pooling layer, a splicing layer, a fifth convolutional neural network, a second fully connected layer, and an output layer; The input layer is used to receive the input from the second fusion module; the first output of the input layer is connected to the input of the second average pooling layer; the second output of the input layer is connected to the input of the second max pooling layer; and the output of the second average pooling layer is connected to the first input of the splicing layer. The output of the second max-pooling layer is connected to the second input of the splicing layer; The concatenation layer is used to concatenate the vectors input from the first input terminal and the vectors input from the second input terminal to obtain a concatenation matrix, and then output the concatenation matrix. The output terminal of the concatenation layer is connected to the input terminal of the fifth convolutional neural network. The output terminal of the fifth convolutional neural network is connected to the second input terminal of the second fully connected layer. The third output terminal of the input layer is connected to the first input terminal of the second fully connected layer. The second fully connected layer is used to multiply the input values ​​from its first input terminal and the input values ​​from its second input terminal. The output values ​​of the three output terminals of the input layer of the second fusion module are obtained by copying the input values ​​received by the input layer three times and then outputting them from the three output terminals respectively. The output terminal of the second fully connected layer is connected to the input terminal of the output layer.

5. The method for predicting the digestibility of compressed dry rations as described in claim 1, characterized in that, The training process of the preset digestibility performance prediction model includes: S21, Obtain the training dataset; the training dataset includes training data and corresponding label information; the training data includes protein content information, mineral content information, starch content information, and fat content information; the label information is a digestibility prediction information vector corresponding to the training data; the digestibility prediction information vector includes predicted protein content, predicted amino acid content, predicted apparent viscosity, and predicted protein digestibility. S22, The training data in the training dataset is used as input data and input into the digestive performance prediction model; S23, The input data is processed using the digestibility prediction model to obtain a prediction vector; S24, perform difference calculation processing on the obtained prediction vector and the label information corresponding to the input data to obtain the difference value; S25, determine whether the difference value satisfies the convergence condition, and obtain the first judgment result; S26, when the first judgment result is negative, determine whether the training iteration count value is equal to the training count threshold, and obtain the second judgment result; When the second judgment result is negative, the model training state is determined to be that the termination training condition is not met. When the second judgment result is yes, it is determined that the model training state meets the termination training condition; When the first judgment result is yes, it is determined that the model training state meets the termination training condition; When the model training state does not meet the termination training condition, the parameters of the digestive performance prediction model are updated using the parameter update model, the training iteration number is increased by 1, and the training data in the training dataset is used as input data to the digestive performance prediction model. When the model training state meets the termination training condition, the training process of the digestive performance prediction model is completed, and the trained digestive performance prediction model is obtained.

6. The method for predicting the digestibility of compressed dry rations as described in claim 5, characterized in that, The expression for the parameter update model is: ， In the formula, For the i-th training data in the training dataset, The label information for the i-th training data in the training dataset. For loss function, Update the parameter value. The parameters of the digestive performance prediction model are... The initial parameter is the learning rate. For momentum angle parameters, Indicates the variable Find the partial derivative. This represents the predicted value obtained by the digestibility performance prediction model for the i-th training data in the training dataset. This is the calculation function corresponding to the digestibility performance prediction model; exp represents the power operation of the constant e; and This is the default value; and This is a preset weighting factor.

7. A device for predicting the digestibility of compressed dry rations, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the method for predicting the digestibility of compressed dry food as described in any one of claims 1 to 6.

8. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by the computer, are used to execute the method for predicting the digestibility of compressed dry food as described in any one of claims 1 to 6.

9. An information data processing terminal, characterized in that, The information data processing terminal is used to implement the method for predicting the digestibility of compressed dry food as described in any one of claims 1 to 6.

Images