A method and device for predicting the remaining degradation days of a film bag, and a storage medium

Abstract

The application discloses a kind of film bag remaining degradation days prediction method, device and storage medium, belong to the field of degradable film bag thermal degradation, the method includes: obtaining the real-time infrared spectrum data of the degradable film bag to be measured in thermal degradation process, and the real-time characteristic peak intensity data in real-time infrared spectrum data is extracted;Real-time characteristic peak intensity data is input into pre-trained combination estimation model, and the degraded day estimation result is obtained;According to the degraded day estimation result and the thermal degradation critical day obtained in advance, the remaining degradation days are obtained;Wherein, the combination estimation model is obtained by the weight proportion of each combination according to multiple single estimation models;The single estimation model includes linear regression estimation model, decision tree estimation model, multivariate perception machine estimation model and support vector machine estimation model, realizes the accurate estimation of degraded day, so as to obtain accurate remaining degradation days in combination with degradable day estimation result.

Description

Technical Field

[0001] This invention belongs to the field of thermal degradation technology of biodegradable film bags, and particularly relates to a method, device and storage medium for predicting the remaining degradation days of film bags. Background Technology

[0002] Biodegradable film bags (mainly composed of polybutylene adipate terephthalate, polylactic acid, and talc) serve as an environmentally friendly alternative to traditional plastics. Their thermal stability is a core indicator determining product lifespan. During storage and transportation (especially by sea), high temperatures can trigger thermal degradation behaviors such as molecular chain breakage and changes in crystallinity, leading to a sudden drop in mechanical properties (e.g., tensile strength decreasing by more than 70% after the critical number of days), causing product failure. Therefore, accurately estimating the number of days of thermal degradation is crucial for ensuring product quality and establishing reasonable warranty periods. However, existing technologies face the following bottlenecks:

[0003] 1. Limitations of single characteristic peak modeling

[0004] Traditional methods typically rely on a single characteristic peak (such as the 730 cm⁻¹ peak of polybutylene adipate terephthalate). -1 Aromatic ring peak or 1758 cm⁻¹ of polylactic acid -1 The absolute intensity of the ester bond peak is used for modeling. However, a single peak is susceptible to environmental interference (such as baseline drift caused by temperature and humidity fluctuations), resulting in large fluctuations in estimation error. In addition, a single peak cannot fully reflect the synergistic degradation mechanism of multi-component materials (such as the interaction between polybutylene adipate terephthalate and polylactic acid), resulting in poor model generalization ability.

[0005] 2. Lack of highly robust estimation models

[0006] Traditional methods are sensitive to noise, especially in complex environments (such as temperature-changing scenarios in maritime transport) where they are unstable. For example, support vector machine regression has an estimation error of over 20% under noise interference, while decision trees, although more robust to noise, have poor model interpretability and are difficult to guide process optimization.

[0007] 3. The static weight combination estimation model lacks adaptability.

[0008] Existing combined estimation models mostly use fixed weights (such as equal-weighted averages or empirical allocations), which cannot adapt to the dynamic changes in the contribution of different characteristic peaks during degradation. For example, in the early stage of degradation, the ester bond peak of polylactic acid (1758 cm⁻¹) -1 The changes were significant, while the methylene peak of polybutylene adipate terephthalate (1457 cm⁻¹) in the later stage was observed. -1 The static weighted model dominates the degradation signal, but it cannot dynamically adjust the weights of each model, resulting in a decrease in estimation accuracy as the degradation stage progresses.

[0009] The aforementioned problems demonstrate that existing methods have significant shortcomings in terms of data robustness, dynamic adaptability, and multi-feature synergy. A new method is urgently needed to fill this technological gap and provide a high-precision, highly robust solution for predicting the remaining thermal degradation days of biodegradable membrane bags. Summary of the Invention

[0010] The purpose of this invention is to provide a method, device, and storage medium for predicting the remaining degradation days of biodegradable bags. By training a combined estimation model with optimal weight ratios using historical infrared spectral data of biodegradable bags during the thermal degradation process, the remaining degradation days of biodegradable bags are estimated using this combined estimation model. The estimation results are then combined with the pre-calculated remaining degradation days, which greatly improves the accuracy and efficiency of predicting the remaining degradation days.

[0011] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0012] In a first aspect, the present invention provides a method for predicting the remaining degradation days of a membrane bag, comprising:

[0013] Real-time infrared spectral data of the biodegradable membrane bag to be tested during the thermal degradation process are obtained, and the real-time characteristic peak intensity data in the real-time infrared spectral data are extracted.

[0014] The real-time characteristic peak intensity data is input into the pre-trained combined estimation model to obtain the estimated number of days of degradation.

[0015] Based on the estimated number of days of degradation already completed and the pre-obtained critical number of days for thermal degradation, the remaining number of days for degradation is obtained.

[0016] The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model.

[0017] Optionally, the training method for the combined estimation model includes:

[0018] Obtain historical infrared spectrum data of biodegradable membrane bags during the thermal degradation process;

[0019] Obtain the characteristic peak intensity data of each historical infrared spectral data in the historical infrared spectral dataset;

[0020] The characteristic peak intensity data and the corresponding actual number of days of degradation were combined to form a historical infrared spectrum sample set;

[0021] The model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model were adjusted using historical infrared spectral sample sets to obtain a combined estimation model with optimal weight ratios after training.

[0022] Optionally, obtaining the historical infrared spectral dataset of the biodegradable membrane bag during the thermal degradation process includes:

[0023] A biodegradable film bag with consistent composition and thickness is placed in a constant temperature environment to accelerate the thermal degradation process, wherein the composition of the biodegradable film bag includes talc.

[0024] Daily full-band infrared spectral data were collected during the thermal degradation process, and the corresponding degradation days of the infrared spectral data were marked to obtain a historical infrared spectral dataset.

[0025] Optionally, obtaining the characteristic peak intensity data of each historical infrared spectral data in the historical infrared spectral dataset includes:

[0026] The absolute intensity of characteristic peaks was extracted from the daily infrared spectral data collected.

[0027] 669 cm of talc powder -1 The characteristic peaks are used as a reference to obtain the relative intensities of the characteristic peaks in the daily infrared spectral data;

[0028] The absolute intensity and the corresponding relative intensity are combined to form the characteristic peak intensity data.

[0029] Optionally, the step of adjusting the model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model using historical infrared spectral sample sets to obtain a trained combined estimation model with optimal weight ratios includes:

[0030] The historical infrared spectral sample set is divided into a training set and a test set, and the training set data is further divided into K-fold partitions.

[0031] In one iteration: the parameters of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are tuned using random K-1 fold training set data. The weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are optimized using linear programming. The sum of the weight ratios of each individual estimation model is 1, and the initial weight ratios of each individual estimation model are equal weight ratios.

[0032] After multiple iterations, the performance of the combined estimation model obtained in each iteration is verified using the remaining 1-fold training set data during each iteration. The results of each verification are compared to obtain the combined estimation model with the optimal weight ratio after training. The number of iterations is no greater than K.

[0033] The performance of the trained combined estimation model with the optimal weight ratio was tested using test set data.

[0034] Optionally, the step of using the remaining 1-fold training set data in each iteration to perform performance verification on the combined estimation model obtained in that iteration, and comparing the results of each verification to obtain the best-trained combined estimation model with the optimal weight ratio, includes:

[0035] The remaining 1-fold training set data in each iteration is used as the input data of the combined estimation model in the corresponding iteration round. The sum of the estimated number of degraded days of the training set data in that fold is obtained, and the total absolute difference is obtained based on the sum of the estimated number of degraded days and the sum of the actual number of degraded days corresponding to the training set data in that fold.

[0036] By comparing the total absolute difference in each iteration process, the combined estimation model with the smallest total absolute difference is selected as the combined estimation model with the optimal weight ratio.

[0037] Optionally, after obtaining the trained combined estimation model with the optimal weight ratio, the method further includes:

[0038] The test set data is input into each trained single estimation model to obtain the root mean square error of the test set for each single estimation model and the correlation coefficient between the actual number of degraded days and the estimated number of degraded days.

[0039] The test set data is input into the combined estimation model with the optimal weight ratio to obtain the root mean square error of the test set of the combined estimation model and the correlation coefficient between the actual number of degraded days and the estimated number of degraded days.

[0040] The root mean square error and correlation coefficient of the combined estimation model with the optimal weight ratio are compared with those of each individual estimation model to determine whether the combined estimation model with the optimal weight ratio outperforms each individual estimation model.

[0041] Optionally, the method for obtaining the critical number of days for thermal degradation includes:

[0042] Based on the infrared spectral data of the critical point of mechanical properties of the membrane bag obtained in advance, historical infrared spectral data of the membrane bag during thermal degradation process that are consistent with it are obtained by comparison.

[0043] Based on the historical infrared spectral data obtained through comparison, the corresponding number of days since degradation was determined.

[0044] In a second aspect, the present invention provides a device for predicting the remaining degradation days of a membrane bag, comprising:

[0045] Real-time characteristic peak intensity data extraction module: used to acquire real-time infrared spectral data of the biodegradable film bag under test during the thermal degradation process, and extract real-time characteristic peak intensity data from the real-time infrared spectral data;

[0046] Degradation Days Estimation Result Acquisition Module: Used to input the real-time characteristic peak intensity data into a pre-trained combined estimation model to obtain the degradation days estimation result;

[0047] The module for obtaining remaining degradation days is used to obtain the remaining degradation days based on the estimated degradation days and the pre-obtained critical thermal degradation days.

[0048] The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model.

[0049] Thirdly, the present invention provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the remaining degradation days of the membrane bag as described in any of the first aspects.

[0050] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: By combining multiple single estimation models and dynamically optimizing the weights of each single model based on the historical thermal degradation stages of the biodegradable membrane bag during the training process of the single estimation models, adapting to changes in the contribution of characteristic peaks, a combined estimation model with optimal weight ratios is obtained. This solves the problems of insufficient accuracy of single models and unsuitable static weights. The combined estimation model with optimal weight ratios is used to estimate the number of days the biodegradable membrane bag has degraded. Combined with the calculation results of the remaining days of membrane bag degradation, a highly robust prediction of the remaining thermal degradation days is achieved. By combining the absolute and relative intensities of historical characteristic bands, the true signal can be preserved while suppressing environmental interference, improving the data stability of the training data. Furthermore, the thermally stable component talc powder 669 cm⁻¹ is utilized. -1 Using the peak as a benchmark, data standardization is achieved, reducing human error correction. Attached Figure Description

[0051] Figure 1 The diagram shown is a flowchart of the method for predicting the remaining degradation days of the membrane bag in Embodiment 1 of the present invention.

[0052] Figure 2 The diagram shown is a flowchart of the method for predicting the remaining degradation days of the membrane bag in Embodiment 2 of the present invention.

[0053] Figure 3 The image shown is a schematic diagram of the infrared spectra of biodegradable film bags taken on different sampling dates in one embodiment of the present invention;

[0054] Figure 4 The figure shown is a schematic diagram of the fitting distribution of the actual number of days of degradation and the estimated number of days of degradation in an embodiment of the present invention;

[0055] Figure 5 The figure shown is a schematic diagram of the fitting distribution of the actual number of days CART has degraded and the estimated number of days degraded in one embodiment of the present invention;

[0056] Figure 6 The figure shown is a schematic diagram of the fitting distribution of the actual number of days of degradation and the estimated number of days of degradation of SVM in one embodiment of the present invention;

[0057] Figure 7 The diagram shows the fitting distribution of the actual number of days degraded and the estimated number of days degraded in a combined estimation model with the optimal weight ratio in one embodiment of the present invention. Detailed Implementation

[0058] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, this embodiment provides a method for predicting the remaining degradation days of a membrane bag, including:

[0061] Real-time infrared spectral data of the biodegradable membrane bag to be tested during the thermal degradation process are obtained, and the real-time characteristic peak intensity data in the real-time infrared spectral data are extracted.

[0062] The real-time characteristic peak intensity data is input into the pre-trained combined estimation model to obtain the estimated number of days of degradation.

[0063] Based on the estimated number of days of degradation already completed and the pre-obtained critical number of days for thermal degradation, the remaining number of days for degradation is obtained.

[0064] The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model.

[0065] By using linear programming to adjust the weights of a single estimation model in real time during training, adapting to changes in the contribution of feature peaks during historical degradation, the accuracy loss caused by manual parameter tuning is avoided. This solves both the problem of insufficient accuracy of a single model and the problem of static weights not being suitable for different degradation stages, resulting in a highly robust estimation model that achieves high-precision estimation of the number of days of degradation. Combined with the pre-calculated number of days of degradation, it can accurately predict the remaining degradation days of the biodegradable membrane bag.

[0066] Example 2

[0067] Based on Example 1, this example also incorporates the following design.

[0068] like Figure 2 As shown, in this embodiment, the method for predicting the remaining degradation days of the membrane bag is achieved through the following steps:

[0069] Step 1: Data Acquisition and Preprocessing.

[0070] A batch of composite film bags with a thickness of 0.004 mm and a composition of 70% polybutylene adipate terephthalate (PBAT), 20% talc, and 10% polylactic acid (PLA) were used. Before the experiment, the film bags were cut into strips of 10cm*30cm. To avoid the influence of factors other than temperature, the test materials were placed in sealed bags, which were then placed in a sealed box. Finally, the sealed box was placed in a constant temperature oven at 65 degrees Celsius. One sample was taken out each day for infrared spectroscopy testing, and the testing was carried out continuously for 100 days, resulting in a total of 100 sets of infrared spectral data.

[0071] In this embodiment, a Bruker V80V Fourier transform infrared spectrometer with a spectral range of 400-4000 cm⁻¹ is used. -1 The test was conducted at room temperature. For each sample, the spectrum was collected three times at three test points, and the average value was taken to eliminate random errors. Figure 3 The image shows infrared spectra at different degradation days, including a 669 cm⁻¹ section. -1 Talc peaks. As can be seen from the figure, with the increase of degradation days, the molecular weight of the film decreases, the crystallinity increases, the crystallinity increases, the glass transition temperature increases, and the mechanical properties decrease. The absorption capacity of some characteristic peaks gradually increases, while the absorption capacity of others shows a decreasing trend.

[0072] Step 2: Calculation of characteristic peak intensity.

[0073] The absolute intensities of 14 characteristic peaks were extracted from the infrared spectra collected during the thermal degradation process of the biodegradable film bags. Among them, 14 characteristic peaks include: 730 cm⁻¹ -1 752 cm -1 873 cm -1 939 cm -1 1209 cm -1 1320cm -1 1361 cm -1 1409 cm -1 1457 cm -1 1504 cm -1 1577 cm -1 1758 cm -1 2873 cm -1 2958 cm -1 The peak height method is used to calculate the difference between the characteristic peak and the baseline.

[0074] With the heat-stabilizing component talc powder 669 cm -1 Characteristic peak intensity For reference, calculate the relative intensity. : ,in, Talc powder at different thermal degradation stages is located at 669 cm⁻¹ -1 The intensity of the nearby characteristic peaks, Data on the intensity changes of 14 characteristic peaks other than talc during different thermal degradation periods.

[0075] The absolute and relative intensities of each characteristic peak are combined to obtain the intensity data of the corresponding characteristic peak.

[0076] Step 3: Training the single estimation model and optimizing the variable weight combination.

[0077] The intensity data of each characteristic peak in each set of infrared spectral data is grouped into an array, and the number of days of thermal degradation corresponding to the intensity data of each characteristic peak is added to the corresponding array. The arrays are combined to obtain the historical infrared spectral dataset. The 100 sets of data in the historical infrared spectral dataset are divided into training set and test set in a 7:3 ratio, with each set containing 14 samples.

[0078] In this embodiment, multiple single estimation models are selected for combination. The training set data is divided into 10 folds. During each iteration of training, 9 folds of the training set data are randomly used to perform hyperparameter tuning and weight ratio tuning for each single estimation model. Then, the remaining 1 fold of the training set data is used to verify the performance of the combined estimation model obtained in this iteration.

[0079] The steps for training each individual estimation model using the training set data and for testing the estimation accuracy of the trained individual estimation models using the test set data are as follows:

[0080] Linear Regression (LR): The linear regression model fits the linear relationship between the intensity of the characteristic peak and the number of degradation days using the least squares method. The specific steps include: first, standardizing the training set data to eliminate the influence of dimensions; then, constructing a linear equation and solving for the optimal coefficients using gradient descent or the normal equation; finally, calculating the mean squared error (MSE) between the estimated and true values ​​as the loss function, and minimizing the loss through iterative optimization. This model is simple and efficient, suitable for capturing linear degradation trends, and the correlation coefficient between the actual number of degradation days and the estimated number of degradation days on the test set is [data missing]. Root mean square error of the test set However, it cannot handle non-linear feature interactions.

[0081] Multi-layer Perceptron (MLP): The MLP is a feedforward neural network with a single hidden layer (10 neurons), using ReLU as the activation function and mean squared error (MSE) as the loss function. During training, weights are adjusted using backpropagation: first, the estimated values ​​are calculated forward, then the error is propagated backward and the parameters are updated using the Adam optimizer; the learning rate is set to 0.001, the batch size to 32, and the iteration count to 100. To avoid overfitting, an L2 regularization term is added. MLP can fit complex nonlinear degradation patterns, but it requires a large amount of data and parameter tuning. The correlation coefficient between the actual number of days degraded and the estimated number of days degraded is shown. Root mean square error of the test set .like Figure 4 As shown, the regression equation for MLP is: .

[0082] Classification and Regression Tree (CART): CART constructs a tree structure by recursively splitting data, using feature peak intensity as nodes. During training, Gini impurity or mean squared error (MSE) is used as the splitting criterion to select the optimal feature and threshold for data splitting until a preset tree depth or leaf node sample number is reached. To avoid overfitting, pruning strategies (such as cost complexity pruning) are employed to optimize the tree structure. This model can automatically capture nonlinear relationships but is sensitive to noise; robustness needs to be improved through talc feature peak standardization. The correlation coefficient between the actual number of days degraded and the estimated number of days degraded is also discussed. Root mean square error of the test set .like Figure 5 As shown, the regression equation for CART is: .

[0083] Support Vector Machine (SVM): SVM uses a radial basis function (RBF) kernel to map low-dimensional features to a high-dimensional space in order to find the optimal hyperplane. A penalty parameter is set during training. The kernel parameters were determined through grid search; a convex quadratic programming problem was solved using the Sequential Minimal Optimization (SMO) algorithm to maximize the classification margin. This model performed stably with small samples, but its estimation error was high when subjected to noise; therefore, it had the lowest weight (0.01) in the combined model and was only used as an auxiliary model to supplement global features. The correlation coefficient between the actual number of degraded days and the estimated number of degraded days is shown. Root mean square error of the test set .like Figure 6 As shown, the regression equation for SVM is: .

[0084] Based on the training process of each individual estimation model described above, the steps for optimizing the weight ratio of each individual estimation model in each iteration of the model parameters are as follows:

[0085] 1. Optimize the construction of the objective function.

[0086] 1) Define variables: The combined estimation model consists of 4 individual estimation models, each with a weighting coefficient of [value missing]. And satisfy .

[0087] In this embodiment, the initial weights of each single estimation model are equally weighted, that is... During the training of each single estimation model using random 9-fold training set data, linear programming is used to dynamically optimize the weight coefficients of the single estimation model on the remaining 1-fold training set data to adapt to the changes in feature contribution of data at different degradation stages.

[0088] 2) Estimated value calculation: For the first value in the 1-fold training set data... One sample, Degradation days estimation results from combined estimation models The weighted sum of the estimates from each individual estimation model, i.e.:

[0089] ，

[0090] in, For the combined estimation model, the first A single estimation model is used to evaluate samples from the 1-fold training set data. The estimated value, For the combined estimation model, the first Weighting coefficients of a single estimation model , , and These correspond to the MLP, CART, LR, and SVM models respectively, based on the first fold of the training data. Estimates for each sample , , and These correspond to the weight coefficients of the MLP, CART, LR, and SVM models, respectively, and the sum of the weight coefficients is 1.

[0091] 3) Difference calculation: The actual number of days of degradation is The estimated absolute difference is .

[0092] 4) Objective Function: The optimization objective is to minimize the total absolute difference among all samples in the 1-fold training set data. Therefore, the objective function is constructed as follows: for:

[0093] ,

[0094] in, In this embodiment, the total number of samples is [number]. , representing 14 characteristic peaks.

[0095] 2. The process of solving linear programming problems.

[0096] The specific process of dynamically optimizing the weight coefficients of each single estimation model (MLP, CART, LR, SVM) using linear programming includes: constructing a linear programming problem with the objective of minimizing the total absolute difference between the estimated number of degraded days and the actual number of degraded days in the remaining 1-fold training set data, and introducing auxiliary variables. This indicates the first fold of the training set data. The estimation error of each sample is calculated, and the weight coefficients are constrained to be non-negative and normalized. By solving this optimization problem, the optimal weight allocation is obtained, which maximizes the estimation accuracy of the combined estimation model.

[0097] Find the optimal solution The formula is expressed as follows:

[0098] ,

[0099] The constraints are:

[0100] ,

[0101] ,

[0102] ,

[0103] ,

[0104] in, To estimate the number of models, this embodiment... The sum of all weight coefficients is 1, and each weight coefficient is non-negative.

[0105] After iterative optimization, the optimal weighted combination estimation model weights are: MLP: 0.38, CART: 0.52, LR: 0.09, SVM: 0.01. The final estimation result is... for: .

[0106] The performance of the optimal weighted combination estimation model was evaluated using a test set, and the correlation coefficient between the actual number of degraded days and the estimated number of degraded days was obtained. Root mean square error of the test set The standard error of estimation is improved by 10.6% compared to the optimal single estimation model CART, which has a standard error of 2.325%.

[0107] Figure 7 The figure shows the fitted distribution of the estimated values ​​from the combined estimation model with the optimal weight ratio of the measured values. The regression equation is: .

[0108] Step 4: Model Application.

[0109] Tensile tests were conducted on the biodegradable film bags, and infrared spectra were collected for the samples after each tensile test. The infrared spectral data of the film bags at the mechanical property critical point were obtained from the collected infrared spectral data. The mechanical property critical point spectrum was used as a reference and compared with the infrared spectral data collected in the thermal degradation experiment. Consistent thermal degradation spectra were obtained, thus the thermal degradation days corresponding to the mechanical property critical point were found to be approximately 65 days.

[0110] Infrared spectroscopy was used to acquire real-time infrared spectral data of a batch of biodegradable membrane bags to be predicted. Characteristic peaks were extracted from the acquired infrared spectral data, and the absolute intensity and relative intensity of each characteristic peak were calculated. The intensity data of each characteristic peak were then input into a combined estimation model with optimal weighting to estimate the number of days the batch of biodegradable membrane bags had degraded. Subtracting the number of days of degradation from the number of days of thermal degradation corresponding to the mechanical performance critical point yields the remaining degradation days of the membrane bags to be tested.

[0111] Example 3

[0112] This embodiment provides a device for predicting the remaining degradation days of a membrane bag, including:

[0113] Real-time characteristic peak intensity data extraction module: used to acquire real-time infrared spectral data of the biodegradable film bag under test during the thermal degradation process, and extract real-time characteristic peak intensity data from the real-time infrared spectral data;

[0114] Degradation Days Estimation Result Acquisition Module: Used to input the real-time characteristic peak intensity data into a pre-trained combined estimation model to obtain the degradation days estimation result;

[0115] The module for obtaining remaining degradation days is used to obtain the remaining degradation days based on the estimated degradation days and the pre-obtained critical thermal degradation days.

[0116] The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model.

[0117] Example 4

[0118] This embodiment provides a computer storage medium storing a computer program. When the computer program is executed by a processor, it implements the method for predicting the remaining degradation days of the membrane bag as described in any step of Embodiment 1 or Embodiment 2.

[0119] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for predicting the remaining degradation days of a membrane bag, characterized in that, include: Real-time infrared spectral data of the biodegradable membrane bag to be tested during the thermal degradation process are obtained, and the real-time characteristic peak intensity data in the real-time infrared spectral data are extracted. The real-time characteristic peak intensity data is input into the pre-trained combined estimation model to obtain the estimated number of days of degradation. Based on the estimated number of days of degradation already completed and the pre-obtained critical number of days for thermal degradation, the remaining number of days for degradation is obtained. The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model. The training method for the combined estimation model includes: Obtain historical infrared spectrum data of biodegradable membrane bags during the thermal degradation process; Obtain the characteristic peak intensity data of each historical infrared spectral data in the historical infrared spectral dataset; The characteristic peak intensity data and the corresponding actual number of days of degradation were combined to form a historical infrared spectrum sample set; The model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model were adjusted using historical infrared spectral sample sets to obtain a combined estimation model with optimal weight ratios after training. The acquisition of characteristic peak intensity data for each historical infrared spectral data in the historical infrared spectral dataset includes: The absolute intensity of characteristic peaks was extracted from the daily infrared spectral data collected. Using the 669 cm⁻¹ characteristic peak of talc as a reference, the relative intensities of characteristic peaks in daily infrared spectral data were obtained; The absolute intensity and the corresponding relative intensity are combined to form the characteristic peak intensity data; The process involves adjusting the model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model using historical infrared spectral sample sets, respectively, to obtain a well-trained combined estimation model with optimal weight ratios, including: The historical infrared spectral sample set is divided into a training set and a test set, and the training set data is further divided into K-fold partitions. In one iteration: the parameters of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are tuned using random K-1 fold training set data. The weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are optimized using linear programming. The sum of the weight ratios of each individual estimation model is 1, and the initial weight ratios of each individual estimation model are equal weight ratios. After multiple iterations, the performance of the combined estimation model obtained in each iteration is verified using the remaining 1-fold training set data during each iteration. The results of each verification are compared to obtain the combined estimation model with the optimal weight ratio after training. The number of iterations is no greater than K.

2. The method for predicting the remaining degradation days of the membrane bag according to claim 1, characterized in that, The acquisition of historical infrared spectral datasets of the biodegradable membrane bag during the thermal degradation process includes: A biodegradable film bag with consistent composition and thickness is placed in a constant temperature environment to accelerate the thermal degradation process, wherein the composition of the biodegradable film bag includes talc. Daily full-band infrared spectral data were collected during the thermal degradation process, and the corresponding degradation days of the infrared spectral data were marked to obtain a historical infrared spectral dataset.

3. The method for predicting the remaining degradation days of the membrane bag according to claim 1, characterized in that, The process of using the remaining 1-fold training set data from each iteration to perform performance verification on the combined estimation model obtained in that iteration, and comparing the results of each verification to obtain the best-trained combined estimation model with the optimal weight ratio, includes: The remaining 1-fold training set data in each iteration is used as the input data of the combined estimation model in the corresponding iteration round. The sum of the estimated number of degraded days of the training set data in that fold is obtained, and the total absolute difference is obtained based on the sum of the estimated number of degraded days and the sum of the actual number of degraded days corresponding to the training set data in that fold. By comparing the total absolute difference in each iteration, the combined estimation model with the smallest total absolute difference is selected as the combined estimation model with the optimal weight ratio.

4. The method for predicting the remaining degradation days of the membrane bag according to claim 1, characterized in that, After obtaining the optimal weighted combination estimation model, the following steps are also included: The test set data is input into each trained single estimation model to obtain the root mean square error of the test set for each single estimation model and the correlation coefficient between the actual number of degraded days and the estimated number of degraded days. The test set data is input into the combined estimation model with the optimal weight ratio to obtain the root mean square error of the test set of the combined estimation model and the correlation coefficient between the actual number of degraded days and the estimated number of degraded days. The root mean square error and correlation coefficient of the combined estimation model with the optimal weight ratio are compared with those of each individual estimation model to determine whether the combined estimation model with the optimal weight ratio outperforms each individual estimation model.

5. The method for predicting the remaining degradation days of the membrane bag according to claim 1, characterized in that, The method for obtaining the critical number of days for thermal degradation includes: Based on the infrared spectral data of the critical point of mechanical properties of the membrane bag obtained in advance, historical infrared spectral data of the membrane bag during thermal degradation process that are consistent with it are obtained by comparison. Based on the historical infrared spectral data obtained through comparison, the corresponding number of days since degradation was determined.

6. A device for predicting the remaining degradation days of a membrane bag, characterized in that, include: Real-time characteristic peak intensity data extraction module: used to acquire real-time infrared spectral data of the biodegradable film bag under test during the thermal degradation process, and extract real-time characteristic peak intensity data from the real-time infrared spectral data; Degradation Days Estimation Result Acquisition Module: Used to input the real-time characteristic peak intensity data into a pre-trained combined estimation model to obtain the degradation days estimation result; The module for obtaining remaining degradation days is used to obtain the remaining degradation days based on the estimated degradation days and the pre-obtained critical thermal degradation days. The combined estimation model is constructed by combining multiple single estimation models according to their respective weight ratios; the single estimation models include linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model. The training method for the combined estimation model includes: Obtain historical infrared spectrum data of biodegradable membrane bags during the thermal degradation process; Obtain the characteristic peak intensity data of each historical infrared spectral data in the historical infrared spectral dataset; The characteristic peak intensity data and the corresponding actual number of days of degradation were combined to form a historical infrared spectrum sample set; The model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model were adjusted using historical infrared spectral sample sets to obtain a combined estimation model with optimal weight ratios after training. The acquisition of characteristic peak intensity data for each historical infrared spectral data in the historical infrared spectral dataset includes: The absolute intensity of characteristic peaks was extracted from the daily infrared spectral data collected. Using the 669 cm⁻¹ characteristic peak of talc as a reference, the relative intensities of characteristic peaks in daily infrared spectral data were obtained; The absolute intensity and the corresponding relative intensity are combined to form the characteristic peak intensity data; The process involves adjusting the model parameters and weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model, and support vector machine estimation model using historical infrared spectral sample sets, respectively, to obtain a well-trained combined estimation model with optimal weight ratios, including: The historical infrared spectral sample set is divided into a training set and a test set, and the training set data is further divided into K-fold partitions. In one iteration: the parameters of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are tuned using random K-1 fold training set data. The weight ratios of the linear regression estimation model, decision tree estimation model, multivariate perceptron estimation model and support vector machine estimation model are optimized using linear programming. The sum of the weight ratios of each individual estimation model is 1, and the initial weight ratios of each individual estimation model are equal weight ratios. After multiple iterations, the performance of the combined estimation model obtained in each iteration is verified using the remaining 1-fold training set data during each iteration. The results of each verification are compared to obtain the combined estimation model with the optimal weight ratio after training. The number of iterations is no greater than K.

7. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for predicting the remaining degradation days of the membrane bag as described in any one of claims 1-5.

Images