Vacuum drying effect evaluation method and system
By acquiring and pre-treating the temperature, pressure and mass data during the drying process, calculating dynamic adjustment factors, and using LSTM and exponential attenuation models, the problem of traditional evaluation methods neglecting the impact of temperature and pressure is achieved, and a more accurate vacuum drying effect is achieved Evaluate.
Patent Information
- Application Number
- CN202411577694.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-07
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When calculating real-time mass changes and moisture content, the traditional vacuum drying effect evaluation method ignores the impact of temperature and pressure data on the drying process, resulting in deviations and inaccuracies of the evaluation results.
By obtaining temperature, pressure and mass change data during the drying process, pre-processing and calculating dynamic adjustment factors, combining the LSTM model and exponential attenuation model, the real-time changes in moisture content are accurately calculated to improve the accuracy and reliability of the evaluation.
It significantly improves the accuracy and reliability of vacuum drying effect evaluation, can more truly reflect the mass dynamics and moisture content changes during the drying process, and enhances the accuracy and flexibility of the evaluation.
Smart Images

Figure CN119066526A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of data processing, and in particular to a vacuum drying effect evaluation method and system. Background Art
[0002] The traditional vacuum drying effect evaluation method has some defects in the data processing process, which can easily affect the accuracy and reliability of the evaluation results.
[0003] For example, some traditional evaluation methods are based on simplified models or assumptions when calculating real-time mass changes and moisture content. These models or assumptions may not fully reflect the complex dynamic behavior of the evaluation object during the drying process, resulting in deviations in the calculation results.
[0004] For example, some traditional evaluation methods ignore the dynamic effects of temperature and pressure data on moisture content changes. In fact, changes in temperature and pressure will significantly affect the evaporation rate and migration behavior of moisture during the drying process. Therefore, if these dynamic factors are not taken into account and only the initial moisture content curve over time is used to evaluate the drying effect, the comprehensiveness and accuracy of the results will be somewhat biased. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide a vacuum drying effect evaluation method and system, which improves the accuracy and reliability of the evaluation results.
[0006] In order to solve the above technical problems, the technical solution of the present invention is as follows: In a first aspect, a method for evaluating a vacuum drying effect is provided, the method comprising: Step 11, obtaining temperature, pressure and mass change data of the evaluation object during the drying process; Step 12, preprocessing the temperature, pressure and mass change data to obtain preprocessed data; Step 13, based on the preprocessed data, calculate the real-time mass change of the evaluation object during the drying process; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; Step 14, calculating a dynamic adjustment factor based on real-time monitored temperature and pressure data; Step 15, using a dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain a final moisture content change curve; Step 16, comparing the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
[0007] Furthermore, based on the preprocessed data, the real-time quality changes of the evaluation object during the drying process are calculated, including: The preprocessed data is divided into a training set and a test set, where the training set is used to train the LSTM model and the test set is used to verify the accuracy of the LSTM model; The parameters of the LSTM model are encoded as genes to form an initial population, where the parameters of the LSTM model include the number of hidden layers, the number of hidden layer units, and the learning rate; for each generation of the population, the following operations are performed: Decode each individual in the population into the parameter configuration of the LSTM model and build the corresponding LSTM model; Use the training set to train each LSTM model, adjust the weights and bias items of the LSTM model, and iterate the training until the preset stop condition is met; Each trained LSTM model is validated using the test set to calculate the mean squared error; According to the mean square error, the corresponding individuals are selected to enter the next generation, and the selected individuals are cross-operated to generate new individuals. The newly generated individuals are mutated to obtain mutant individuals. The mutant individuals are used to replace some individuals in the original population to form the next generation population. When the preset number of generations is reached, the iterative process is terminated to obtain the final LSTM model. Input the real-time collected temperature and pressure data into the final LSTM model to obtain the predicted real-time quality change value; The real-time mass change of the evaluation object during the drying process is calculated according to the initial mass of the evaluation object and the predicted real-time mass change value.
[0008] Furthermore, the weights and biases of the LSTM model are adjusted, including: Set the initial weights and biases of the LSTM model; The temperature, pressure, and mass change features in the training dataset are input into the LSTM model to calculate the hidden state and output data for each time step; Calculate the mean square error using the mass change or moisture content in the output data and the training data set; Calculate the gradient for each weight and bias term in the LSTM model based on the mean squared error; The Adam algorithm is used to update the weights and biases of the LSTM model according to the gradient.
[0009] Further, according to the initial mass of the evaluation object and the predicted real-time mass change value, the real-time mass change of the evaluation object during the drying process is calculated, including: According to the initial quality of the evaluation object and the predicted real-time quality change value, Calculate the real-time quality change of the evaluation object during the drying process; in, Indicates the target variable at time The value of Indicates the starting value, which is the target variable at the current time The value of Represents the time step The predicted change in represents an exponential decay factor or an exponential growth factor; represents the coefficient; represents the coefficient; represents the sinusoidal wave term, representing periodicity; represents a constant; represents the random disturbance term; Indicates from 0 to the current time The continuous time integral of ; Indicates that over time Function of change; represents the time step; Represents the current time; Indicates temperature; Indicates stress; ; in, represents the initial amplitude of the sine wave; represents the exponential change term; represents an exponential growth rate or decay rate; if k >0, indicating positive exponential growth; if k <0, indicating exponential decay; represents the angular frequency of the sine function; is the initial phase offset of the sine function; Indicates that over time The vertical offset of the function being varied; Represents the base of natural logarithms.
[0010] Furthermore, based on the initial mass of the object to be evaluated, the real-time change of its moisture content is calculated to obtain a preliminary curve of moisture content change over time, including: Determine the initial quality of the evaluation object , continuously collects real-time quality data of the evaluation object during the drying process according to the set frequency , and record the corresponding current time ; Smoothing the collected quality data. ,pass Calculate the current moisture content forecast ; Set the current time The corresponding current moisture content Paired storage to form a data set; According to the changing trend of the data, the exponential decay model is selected as the fitting model, and its form is: ,in, represents the initial moisture content; represents the base of natural logarithms; represents the attenuation constant; The least square method is used, and the error function is defined as the residual sum of squares of all data points; The error function is minimized by the gradient descent algorithm, and the optimization process is iterated until the initial moisture content corresponding to the error function is found. and the decay constant ; Use initial moisture content and the decay constant , calculate the predicted value of the fitted model at each data point; Based on the predicted values, the residuals are calculated and the distribution of the residuals is analyzed to generate a series of time points at certain time intervals within the entire observation time range; Using Fit Model and the optimized initial moisture content and the decay constant , calculate the predicted value of moisture content at a time point; Connect the time points with the corresponding moisture content predictions to form a smooth curve of moisture content changing over time.
[0011] Furthermore, the least squares method is used to define the error function as the residual sum of squares of all data points, including: Initialize the error function to 0; Iterate over each data point in the dataset ,in, is the current time, is the current moisture content; For each data point in the dataset, use the initial moisture content of the current iteration and the decay constant The estimated value of Calculate predicted values; Calculate the residual between the actual observed and predicted values of a data point; The residual square of each data point is added to the error function, and the process continues until all data points in the data set are traversed. The final error function value is the residual square sum of all data points.
[0012] Furthermore, the calculation formula of the dynamic adjustment factor is: ; in, is the dynamic adjustment factor; , , , , and is an empirical parameter obtained by the least squares method; is the current temperature The difference from the initial temperature; is the reference temperature; is the reference pressure; It is the pressure that changes over time; is the power function of the pressure ratio from 0 to the current time The integral of is the rate of change of pressure with time.
[0013] In a second aspect, a vacuum drying effect evaluation system is provided, comprising: An acquisition module is used to acquire the temperature, pressure and mass change data of the evaluation object during the drying process; preprocess the temperature, pressure and mass change data to obtain preprocessed data; calculate the real-time mass change of the evaluation object during the drying process based on the preprocessed data; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; The processing module is used to calculate the dynamic adjustment factor according to the real-time monitored temperature and pressure data; use the dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain the final moisture content change curve; compare the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
[0014] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method described.
[0015] In a fourth aspect, a computer-readable storage medium stores a program, and when the program is executed by a processor, the method described is implemented.
[0016] The above solution of the present invention includes at least the following beneficial effects: By performing meticulous preprocessing on the temperature, pressure and mass change data, the present invention can effectively filter out noise and abnormal values in the original data.
[0017] Based on preprocessed data, the present invention can accurately calculate the real-time quality changes of the evaluation object during the drying process, and further accurately extract the initial quality of the evaluation object. This feature enables the method to more realistically reflect the quality dynamics of the evaluation object during the drying process.
[0018] The present invention innovatively introduces a dynamic adjustment factor, which is calculated based on real-time monitored temperature and pressure data. By using this dynamic adjustment factor to adjust the initial moisture content over time curve, the present invention can more accurately capture the actual changes in moisture content during the drying process, thereby significantly improving the accuracy and reliability of the assessment. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a flow chart of a vacuum drying effect evaluation method provided by an embodiment of the present invention.
[0020] Figure 2 It is a schematic diagram of a vacuum drying effect evaluation system provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0021] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0022] like Figure 1 As shown, an embodiment of the present invention provides a method for evaluating vacuum drying effect, the method comprising the following steps: Step 11, obtaining temperature, pressure and mass change data of the evaluation object during the drying process; Step 12, preprocessing the temperature, pressure and mass change data to obtain preprocessed data; Step 13, based on the preprocessed data, calculate the real-time mass change of the evaluation object during the drying process; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; Step 14, calculating a dynamic adjustment factor based on real-time monitored temperature and pressure data; Step 15, using a dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain a final moisture content change curve; Step 16, comparing the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
[0023] In an embodiment of the present invention, step 11, by directly obtaining the key parameters (temperature, pressure and mass change data) in the drying process, the actual situation of the drying process can be fully and accurately reflected. Step 12, data preprocessing can eliminate noise, outliers and missing values in the original data, and improve the quality and reliability of the data. Step 13, by calculating the mass change and moisture content change in real time, the drying process can be dynamically monitored and abnormal conditions can be discovered in time. At the same time, the extracted initial mass and preliminary moisture content change curves provide important references for subsequent precise adjustments and effect comparisons. Step 14, the introduction of dynamic adjustment factors makes the evaluation method more flexible and accurate. It can dynamically adjust the drying process according to real-time temperature and pressure data, so as to better adapt to different drying conditions and evaluation requirements. Step 15, by applying the dynamic adjustment factor, the accuracy and authenticity of the moisture content change curve can be further improved. Step 16, by comparing with historical data, the difference between the current drying effect and the past situation can be intuitively displayed.
[0024] In a preferred embodiment of the present invention, the above step 11, obtaining the temperature, pressure and mass change data of the evaluation object during the drying process, may include: During the drying process, temperature is a key factor affecting the drying rate and final quality of the resin. In order to obtain accurate temperature data, temperature sensors are installed at key locations of the resin drying equipment. These sensors can monitor and record the temperature changes experienced by the resin at different drying stages in real time. By continuously collecting temperature data, a complete temperature curve can be obtained, reflecting the changes in the thermal state of the resin throughout the drying process.
[0025] During the vacuum drying process, pressure control is crucial to ensure the effective evaporation of moisture inside the resin and prevent adverse reactions such as oxidation. In order to obtain accurate pressure data, pressure sensors are installed in the resin drying equipment. These sensors can monitor and record the changes in the vacuum degree in the drying chamber in real time. By collecting pressure data, we can understand the vacuum environment that the resin is subjected to at different drying stages, and then analyze its impact on the drying effect.
[0026] In order to accurately evaluate the drying effect of the resin, it is necessary to obtain the mass change data during the drying process. Before the resin enters the drying equipment, the initial mass is weighed; during the drying process, the mass change of the resin is recorded regularly or continuously. By comparing the initial mass and real-time mass data, the moisture removal rate of the resin can be calculated, thereby intuitively reflecting its drying degree.
[0027] In a preferred embodiment of the present invention, the above step 12, obtaining the temperature, pressure and mass change data of the evaluation object during the drying process, may include: Check whether there are values in the temperature, pressure and quality data that are obviously inconsistent with the actual process range, such as data points with too high temperature or too low pressure, which may be caused by sensor failure or recording errors. For missing data caused by temporary sensor failure or other reasons, use interpolation methods (such as linear interpolation, polynomial interpolation, etc.) to fill in to ensure data continuity. Apply a sliding average filter or a low-pass filter to reduce noise and fluctuations in the data and make the temperature, pressure and quality change curves smoother.
[0028] Standardize temperature, pressure, and mass data to convert them to a uniform dimension or range. For example, you can convert temperature data to a uniform standard of degrees Celsius, and pressure data to Pascals or millibars, etc.
[0029] In a preferred embodiment of the present invention, the above step 13, calculating the real-time quality change of the evaluation object during the drying process based on the preprocessed data, may include: Step 131, divide the preprocessed data into a training set and a test set, wherein the training set is used to train the LSTM model, and the test set is used to verify the accuracy of the LSTM model, specifically including: randomly shuffling the preprocessed data to ensure the randomness and generalization ability of the data; then, divide the data into a training set and a test set according to a preset ratio (such as 70% training set, 30% test set), and this ratio can be adjusted according to the size of the data and the complexity of the model; according to the input requirements of the LSTM model, format the data in the training set and the test set into time series data. This usually involves organizing the data into a three-dimensional tensor, where the dimensions represent the number of samples, the time step, and the number of features, respectively. For the temperature and pressure data in this case, they can be regarded as features and arranged in chronological order according to the drying process.
[0030] In order to speed up the model training process and improve the stability of the model, the data in the training set and test set can be normalized. This usually means scaling the data to between 0 and 1 or to have zero mean and unit variance. Normalization parameters (such as minimum, maximum or mean, standard deviation) should be calculated only from the training data and applied to the test data to ensure the generalization ability of the model on unknown data.
[0031] Step 132, encode the parameters of the LSTM model into genes to form an initial population, wherein the parameters of the LSTM model include the number of hidden layers, the number of hidden layer units and the learning rate; define a reasonable value range for the parameters of the LSTM model (the number of hidden layers, the number of hidden layer units and the learning rate). For example, the number of hidden layers can vary from 1 to 5, the number of hidden layer units can vary from 50 to 500, and the learning rate can vary from 0.001 to 0.1. Map each parameter to one or more genes, which can be achieved by binary coding, real number coding or other appropriate coding methods. For example, binary coding can be used to represent different levels of the number of hidden layers and the number of hidden layer units, while real number coding is used to represent the learning rate. Randomly generate an initial population based on the defined parameter range and gene coding method. Each individual in the population represents a possible LSTM model configuration in the population. For each generation of the population, perform the following operations: Step 133, decoding each individual in the population into a parameter configuration of the LSTM model, and building a corresponding LSTM model, specifically including: traversing each individual in the current population, decoding its encoded gene sequence into corresponding LSTM model parameters, which include the number of hidden layers, the number of hidden layer units, and the learning rate; decoding can be implemented through a mapping function, such as converting binary codes into integers, or directly using real number codes as parameter values; instantiating the LSTM model using a deep learning framework (such as TensorFlow, PyTorch, etc.) according to the decoded parameters; setting the input and output dimensions of the model, as well as other related configurations such as activation functions and initialization methods; preparing the environment for model training, including configuring GPU acceleration (if available), setting random number seeds to ensure repeatability, and initializing the weights and bias items of the model.
[0032] Step 134, using the training set to train each LSTM model, adjusting the weight and bias of the LSTM model, and iterating the training until a preset stop condition is met; Step 135, using the test set to verify each trained LSTM model to calculate the mean square error, specifically including: loading the preprocessed test set data to verify the performance of the LSTM model; using the test set data to forward propagate the trained LSTM model to obtain the predicted quality change value; calculating the mean square error (MSE) between the predicted value and the true value as an evaluation indicator of the model performance; recording the mean square error of each LSTM model on the test set for subsequent comparison and selection.
[0033] Step 136, according to the mean square error, select the corresponding individuals to enter the next generation, perform a crossover operation on the selected individuals to generate new individuals, perform a mutation operation on the newly generated individuals to obtain mutant individuals; replace some individuals in the original population with the mutant individuals to form the next generation population, and terminate the iterative process when the preset number of generations is reached to obtain the final LSTM model, which specifically includes: According to the mean square error calculated in step 135, individuals with better performance are selected to enter the next generation. This can be achieved through strategies such as roulette selection and tournament selection; crossover operations are performed on the selected individuals to generate new individuals. The crossover operation may include single-point crossover, multi-point crossover or uniform crossover, etc.; mutation operations are performed on the newly generated individuals to increase the diversity of the population. Mutation operations may include bit flipping, real number perturbation, etc., and the specific method also depends on the gene encoding method; some individuals in the original population are replaced with the mutated new individuals to form the next generation population. The replacement strategy can be complete replacement, partial replacement or selective replacement based on fitness, and check whether the preset number of generations is reached. If reached, the iterative process is terminated, and the LSTM model with the best performance in the current population is output as the final result; otherwise, the evolution process of the next generation population is continued.
[0034] Step 137, input the real-time collected temperature and pressure data into the final LSTM model to obtain the predicted real-time quality change value, specifically including: The real-time collected temperature and pressure data undergo the same preprocessing steps as the training data to ensure consistency in data format and range, including data normalization, removal of outliers or noise, etc.; the real-time collected data is formatted into an appropriate form according to the input requirements of the LSTM model. Generally, this means organizing the data into the same time series format as the training data, where the data at each time point includes temperature and pressure values. Load the final LSTM model obtained in step 136; input the formatted real-time temperature and pressure data into the loaded LSTM model, which involves passing the data to the input layer of the model. Forward propagation is performed through the LSTM model, and the output of the model is calculated based on the input temperature and pressure data. This output is the predicted real-time quality change value.
[0035] Step 138, calculating the real-time mass change of the evaluation object during the drying process according to the initial mass of the evaluation object and the predicted real-time mass change value.
[0036] In an embodiment of the present invention, the preprocessed data is learned and trained by the LSTM model, so that the time dependency and complex patterns in the data can be captured. The LSTM model is particularly suitable for processing sequence data, such as the temperature and pressure data in this example, so that the real-time quality changes can be predicted more accurately. By encoding the parameters of the LSTM model as genes and optimizing them using a genetic algorithm, the model configuration (such as the number of hidden layers, the number of hidden layer units, and the learning rate) that best suits the current data can be automatically found. This automated parameter adjustment method avoids the tedious process of manual trial and error and improves the adaptability and generalization ability of the model. By generating new individuals through crossover and mutation operations, and selecting individuals with excellent performance in each generation to enter the next generation, the quality of individuals in the population can be gradually improved. This method helps the model maintain good prediction performance when facing data from different batches or under different conditions, and enhances the robustness of the model. Using the trained LSTM model to predict the temperature and pressure data collected in real time, the quality changes of the evaluation object during the drying process can be fed back in time. This helps production personnel monitor the drying process in real time, discover and adjust abnormal situations in time, and ensure product quality and production efficiency.
[0037] In a preferred embodiment of the present invention, the above step 134, adjusting the weights and bias items of the LSTM model, may include: Step 1341, setting the initial weights and bias items of the LSTM model, specifically including: the weights and bias items of the LSTM model will be randomly initialized before training begins, which can be achieved by using standard normal distribution or uniform distribution to generate initial values; in deep learning frameworks (such as TensorFlow or PyTorch), built-in initialization functions can be used to easily set the initial weights and bias items.
[0038] Step 1342, input the temperature, pressure and mass change features in the training data set into the LSTM model to calculate the hidden state and output data for each time step, specifically including: ensuring that the training data set has been formatted according to the requirements of the LSTM model, usually in the form of a three-dimensional tensor (number of samples, time steps, number of features); in the LSTM model, each time step will calculate a new hidden state and cell state based on the current input, the hidden state and cell state of the previous time step. This process is recursive and proceeds sequentially along the time steps; at the end of each time step, the LSTM model generates output data based on the current hidden state. This usually involves passing the hidden state through a fully connected layer (which may include an activation function) to obtain the output.
[0039] Step 1343, using the output data and the mass change or moisture content in the training data set to calculate the mean square error, specifically including: Mean Squared Error (MSE) is a commonly used loss function in regression problems. It calculates the average of the squares of the difference between the model prediction value and the actual value; at each time step or at the end of the entire sequence, the output data of the LSTM model is compared with the corresponding mass change or moisture content true value in the training data set to calculate the mean square error; for the entire training batch, the errors of all samples are usually accumulated for subsequent gradient calculation and weight update; Step 1344, based on the mean square error, calculate the gradient for each weight and bias item in the LSTM model, specifically including: using the chain rule, starting from the output layer, and calculating the gradient layer by layer toward the input layer. This involves taking partial derivatives for each weight and bias item to understand their contribution to the final error; the calculated gradients will be stored for use in subsequent weight update steps; the deep learning framework provides an automatic differentiation function that can automatically calculate and store gradients, greatly simplifying the implementation of this step.
[0040] Step 1345, using the Adam algorithm to update the weights and bias items of the LSTM model according to the gradient, specifically including: setting the parameters of the Adam algorithm, such as the learning rate, the exponential decay rate of the first-order moment estimate, and the exponential decay rate of the second-order moment estimate; the Adam algorithm adjusts the update amount of each weight and bias item according to the historical gradient and the current gradient, which involves calculating the first-order moment estimate and the second-order moment estimate, and using them to dynamically adjust the learning rate.
[0041] In the embodiment of the present invention, through the adjustment of steps 1341 to 1345, the LSTM model can more accurately capture the complex relationship between temperature, pressure and mass change. In particular, the calculation of the mean square error in step 1343 and the calculation of the gradient in step 1344 enable the model to accurately know which parts need to be improved, so as to optimize the prediction results by adjusting the weights and bias items. The weights and bias items are updated using the Adam algorithm (step 1345), so that the LSTM model has stronger adaptability. The Adam algorithm combines the ideas of momentum gradient and RMSprop, and can dynamically adjust the learning rate according to historical gradient information, so that the model can converge quickly and avoid falling into the local optimal solution during the training process. By fine-tuning each weight and bias item, the LSTM model can better handle noise and outliers in the input data. This improvement in robustness makes the model more reliable in practical applications, and even in the face of real-time data that is not completely accurate or has certain interference, it can maintain good prediction performance. The entire adjustment process in step 134 is iterative, and each iteration adjusts the model parameters according to the last prediction result and error. This iterative approach makes the training process more efficient and can achieve better model performance within a limited number of training rounds.
[0042] In a preferred embodiment of the present invention, the above step 138, calculating the real-time mass change of the evaluation object during the drying process according to the initial mass of the evaluation object and the predicted real-time mass change value, includes: Step 1381, based on the initial quality of the evaluation object and the predicted real-time quality change value, Calculate the real-time quality change of the evaluation object during the drying process; in, Indicates the target variable at time The value of Indicates the starting value, which is the target variable at the current time The value of Represents the time step The predicted change in represents an exponential decay factor or an exponential growth factor; represents the coefficient; represents the coefficient; represents the sinusoidal wave term, representing periodicity; represents a constant; represents the random disturbance term; Indicates from 0 to the current time The continuous time integral of ; Indicates that over time Function of change; represents the time step; Represents the current time; Indicates temperature; Indicates stress; ; in, represents the initial amplitude of the sine wave; represents the exponential change term; represents an exponential growth rate or decay rate; if k >0, indicating positive exponential growth; if k <0, indicating exponential decay; represents the angular frequency of the sine function; is the initial phase offset of the sine function; Indicates that over time The vertical offset of the function being varied; Represents the base of natural logarithms.
[0043] In the embodiment of the present invention, the calculation process comprehensively considers multiple factors, including temperature, pressure, time, periodic fluctuations and random disturbances, etc., so that the prediction of mass change is more accurate. , , , k , The calculation process can be adapted to different drying environments and material properties. This flexibility enhances the versatility and practicality of the model. , the model can take into account the periodic changes that may occur in the drying process, such as periodic fluctuations in equipment operation or periodic effects of environmental factors. This helps to more accurately simulate and predict quality changes in the actual drying process. and continuous time integral This allows the model to dynamically adapt to changes in the drying process. For example, when the drying rate changes, the model can quickly adjust the prediction results to maintain the accuracy of the prediction.
[0044] In a preferred embodiment of the present invention, the above step 13, based on the initial mass of the evaluation object, calculates the real-time change of its moisture content to obtain a preliminary curve of moisture content change over time, and may also include: Step 139, determine the initial quality of the evaluation object , continuously collects real-time quality data of the evaluation object during the drying process according to the set frequency , and record the corresponding current time For example, real-time quality data of the evaluation object is continuously collected during the drying process at a set frequency (such as every second, every minute, etc.). At the same time, use the timestamp to record the current time corresponding to each quality data point .
[0045] Smoothing the collected quality data. ,pass Calculate the current moisture content forecast For example, a smoothing algorithm (such as moving average method, exponential smoothing method, etc.) is used to process the collected real-time quality data to reduce noise and fluctuations in the data; for each collected real-time quality data , according to the formula Calculate the current moisture content forecast.
[0046] Set the current time The corresponding current moisture content Paired storage forms a data set; according to the changing trend of the data, the exponential decay model is selected as the fitting model, which is in the form of: ,in, represents the initial moisture content; represents the base of natural logarithms; represents the decay constant. For example, The corresponding moisture content prediction value Paired storage forms a dataset, which can be achieved in a programming environment by creating data structures such as lists, arrays, or data frames to facilitate subsequent data analysis and model fitting.
[0047] The least squares method is used, and the error function is defined as the residual sum of squares of all data points; The error function is minimized by the gradient descent algorithm, and the optimization process is iterated until the initial moisture content corresponding to the error function is found. and the decay constant , for example, by minimizing the error function through a gradient descent algorithm, and iterating the optimization process until the initial moisture content that minimizes the error function is found and the decay constant , which can be achieved through programming, using numerical computing libraries or optimization libraries to complete parameter optimization.
[0048] Use initial moisture content and the decay constant , calculate the predicted value of the fitted model at each data point; based on the predicted value, calculate the residual and analyze the distribution of the residual, generating a series of time points at a certain time interval within the entire observation time range; using the fitted model and the optimized initial moisture content and the decay constant , calculate the predicted moisture content at the time point, for example, using the initial moisture content obtained by optimization and the decay constant , calculate the predicted value of the fitted model at each data point. Then, calculate the residuals based on the predicted values and the actual observed values, and analyze the distribution of the residuals, which helps to evaluate the fitting effect and reliability of the model.
[0049] Connect the time points with the corresponding predicted moisture content values to form a smooth curve of moisture content changing over time. For example, generate a series of time points at certain time intervals over the entire observation time range. Then, use the fitted model and the optimized parameters to calculate the predicted moisture content at these time points. Finally, connect the time points with the corresponding predicted moisture content values to form a smooth curve of moisture content changing over time.
[0050] In an embodiment of the present invention, by continuously collecting the real-time quality data of the evaluation object during the drying process and recording the corresponding current time, real-time monitoring of the drying process is achieved. This helps to capture the change of moisture content in time. Smoothing the collected quality data can reduce the noise and outliers in the data and improve the quality and reliability of the data. This helps to more accurately reflect the change of moisture content of the evaluation object during the drying process. By calculating the current moisture content prediction value and pairing the current time with the corresponding moisture content for storage, a complete data set is formed. This makes it possible to more accurately track and predict the change of moisture content of the evaluation object during the drying process. The exponential decay model is selected as the fitting model, and the least squares method and gradient descent algorithm are used for parameter optimization, which can effectively fit the change trend of the data. This fitting method not only takes into account the overall trend of the data, but also ensures the accuracy of the fitting by minimizing the residual sum of squares. Using the optimized initial moisture content and decay constant, the predicted value of the fitting model at each data point is calculated, and a smooth curve of moisture content changing over time is generated. This curve can continuously show the change of moisture content of the evaluation object during the entire drying process, which is convenient for observation and analysis.
[0051] In a preferred embodiment of the present invention, the least square method is used to define the error function as the residual sum of squares of all data points, which may include: Initialize the error function to 0; Iterate over each data point in the dataset ,in, is the current time, is the current moisture content; For each data point in the dataset, use the initial moisture content of the current iteration and the decay constant The estimated value of Calculate predicted values; Calculate the residual between the actual observed and predicted values of a data point; The residual square of each data point is added to the error function, and the process continues until all data points in the data set are traversed. The final error function value is the residual square sum of all data points.
[0052] In the embodiment of the present invention, before starting the calculation, the initial value of the error function (expressed as the loss function) is set to 0. This is to ensure that when the residual squares are accumulated later, it can start from a fixed starting point. , and process them one by one; for each data point in the data set, use the initial moisture content of the current iteration and the decay constant The estimated value of Calculate the predicted value, which is the result of the model fitting the data points under the current parameters; by calculating the residual sum of squares of all data points, the model's fitting effect on the actual data can be accurately evaluated. The smaller the residual sum of squares, the closer the model's predicted value is to the actual observed value, and the better the fitting effect. Use the least squares method to define the error function, and by minimizing this error function, the model's parameters (such as initial moisture content) can be optimized. and the decay constant ), thereby improving the prediction accuracy of the model; by accumulating the residual sum of squares for all data points, the model will take into account the distribution of the entire data set during the optimization process, rather than just being limited to individual data points, which helps the model's generalization ability on unknown data, making it more robust and reliable. By defining the error function as the residual sum of squares, the calculation process of the model can be simplified and the calculation efficiency can be improved.
[0053] In a preferred embodiment of the present invention, the calculation formula of the dynamic adjustment factor is: ; in, is the dynamic adjustment factor; , , , , and is an empirical parameter obtained by the least squares method; is the current temperature The difference from the initial temperature; is the reference temperature; is the reference pressure; It is the pressure that changes over time; is the power function of the pressure ratio from 0 to the current time The integral of is the rate of change of pressure with time.
[0054] In the embodiment of the present invention, the dynamic adjustment factor in the formula It can make real-time adjustments based on the temperature and pressure changes in actual situations, so that the system or model can adapt to different working environments more quickly. and pressure change rate The formula can help reduce the fluctuation of the system caused by environmental changes and enhance the stability of the system by combining dynamic factors such as , and fine-tuning them with empirical parameters. , , , , and Ability to ensure dynamic adjustment factor The formula takes into account a variety of factors that may affect the behavior of the system, including nonlinear temperature and pressure effects, and their changes over time. This comprehensive consideration helps improve the accuracy of the system's prediction of future states, especially in complex and dynamically changing environments.
[0055] The specific implementation process of step 15 is: Obtain a preliminary moisture content versus time curve from previous analysis or experimental data. This is usually a series of data points, each of which represents the moisture content at a specific time point; according to the calculation formula of the dynamic adjustment factor, substitute the current environmental parameters (such as temperature difference, pressure change rate, etc.) and empirical parameters (obtained by methods such as least squares); calculate the dynamic adjustment factor corresponding to each time point; apply the calculated dynamic adjustment factor to the preliminary moisture content change curve. Specifically, the moisture content at each time point is multiplied by the dynamic adjustment factor at the corresponding time point, so that each data point on the preliminary curve will be dynamically adjusted according to the current environmental conditions to obtain a moisture content value that is closer to the actual situation. After dynamic adjustment, the obtained data points constitute the final moisture content versus time curve. This curve can be visualized using charting software or programming libraries.
[0056] The specific implementation process of step 16 is: Collect historical data similar to the current vacuum drying process from databases, experimental records or other reliable sources. These data should include moisture content values at different time points, as well as possible other relevant parameters (such as temperature, pressure, etc.). Compare the final moisture content change curve generated in step 15 with the historical data. This can be achieved by plotting two curves on the same chart for intuitive comparison. When comparing, attention should be paid to aspects such as the overall trend of the curve, the moisture content values at key time points, and the fluctuations of the curve. Based on the comparison results, evaluate the effect of the current vacuum drying process. For example, if the final curve shows that the moisture content decreases faster than the historical data, or it takes less time to reach the same moisture content, it can be considered that the current vacuum drying effect is better.
[0057] like Figure 2 As shown, an embodiment of the present invention further provides a vacuum drying effect evaluation system 20, comprising: The acquisition module 21 is used to acquire the temperature, pressure and mass change data of the evaluation object during the drying process; pre-process the temperature, pressure and mass change data to obtain pre-processed data; calculate the real-time mass change of the evaluation object during the drying process based on the pre-processed data; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; The processing module 22 is used to calculate the dynamic adjustment factor according to the real-time monitored temperature and pressure data; use the dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain the final moisture content change curve; compare the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
[0058] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.
[0059] The embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.
[0060] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.
[0061] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A method for evaluating vacuum drying effect, characterized in that: The method comprises: Step 11, obtaining temperature, pressure and mass change data of the evaluation object during the drying process; Step 12, preprocessing the temperature, pressure and mass change data to obtain preprocessed data; Step 13, based on the preprocessed data, calculate the real-time mass change of the evaluation object during the drying process; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; Step 14, calculating a dynamic adjustment factor based on real-time monitored temperature and pressure data; Step 15, using a dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain a final moisture content change curve; Step 16, comparing the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
2. A vacuum drying effect evaluation method according to claim 1, characterized in that: Based on the pre-processed data, the real-time quality changes of the evaluation object during the drying process are calculated, including: The preprocessed data is divided into a training set and a test set, where the training set is used to train the LSTM model and the test set is used to verify the accuracy of the LSTM model; The parameters of the LSTM model are encoded as genes to form an initial population, where the parameters of the LSTM model include the number of hidden layers, the number of hidden layer units, and the learning rate; for each generation of the population, the following operations are performed: Decode each individual in the population into the parameter configuration of the LSTM model and build the corresponding LSTM model; Use the training set to train each LSTM model, adjust the weights and bias items of the LSTM model, and iterate the training until the preset stop condition is met; Each trained LSTM model is validated using the test set to calculate the mean squared error; According to the mean square error, the corresponding individuals are selected to enter the next generation, and the selected individuals are cross-operated to generate new individuals. The newly generated individuals are mutated to obtain mutant individuals. The mutant individuals are used to replace some individuals in the original population to form the next generation population. When the preset number of generations is reached, the iterative process is terminated to obtain the final LSTM model. Input the real-time collected temperature and pressure data into the final LSTM model to obtain the predicted real-time quality change value; The real-time mass change of the evaluation object during the drying process is calculated according to the initial mass of the evaluation object and the predicted real-time mass change value.
3. A vacuum drying effect evaluation method according to claim 2, characterized in that: Adjust the weights and biases of the LSTM model, including: Set the initial weights and biases of the LSTM model; The temperature, pressure, and mass change features in the training dataset are input into the LSTM model to calculate the hidden state and output data for each time step; Calculate the mean square error using the mass change or moisture content in the output data and the training data set; Calculate the gradient for each weight and bias term in the LSTM model based on the mean squared error; The Adam algorithm is used to update the weights and biases of the LSTM model according to the gradient.
4. A vacuum drying effect evaluation method according to claim 3, characterized in that: According to the initial mass of the evaluation object and the predicted real-time mass change value, the real-time mass change of the evaluation object during the drying process is calculated, including: According to the initial quality of the evaluation object and the predicted real-time quality change value, Calculate the real-time quality change of the evaluation object during the drying process; in, Indicates the target variable at time The value of Indicates the starting value, which is the target variable at the current time The value of Represents the time step The predicted change in represents an exponential decay factor or an exponential growth factor; represents the coefficient; represents the coefficient; represents the sinusoidal wave term, representing periodicity; represents a constant; represents the random disturbance term; Indicates from 0 to the current time The continuous time integral of ; Indicates that over time Function of change; represents the time step; Represents the current time; Indicates temperature; Indicates stress; ; in, represents the initial amplitude of the sine wave; represents the exponential change term; represents an exponential growth rate or decay rate; if k >0, indicating positive exponential growth; if k <0, indicating exponential decay; represents the angular frequency of the sine function; is the initial phase offset of the sine function; Indicates that over time The vertical offset of the function being varied; Represents the base of natural logarithms.
5. A vacuum drying effect evaluation method according to claim 4, characterized in that: Based on the initial mass of the object being evaluated, calculate the real-time change in moisture content to obtain a preliminary curve of moisture content changing over time, including: Determine the initial quality of the evaluation object , continuously collects real-time quality data of the evaluation object during the drying process according to the set frequency , and record the corresponding current time ; Smoothing the collected quality data. ,pass Calculate the current moisture content forecast ; Set the current time The corresponding current moisture content Paired storage to form a data set; According to the changing trend of the data, the exponential decay model is selected as the fitting model, and its form is: ,in, represents the initial moisture content; represents the base of natural logarithms; represents the attenuation constant; The least square method is used, and the error function is defined as the residual sum of squares of all data points; The error function is minimized by the gradient descent algorithm, and the optimization process is iterated until the initial moisture content corresponding to the error function is found. and the decay constant ; Use initial moisture content and the decay constant , calculate the predicted value of the fitted model at each data point; Based on the predicted values, the residuals are calculated and the distribution of the residuals is analyzed to generate a series of time points at certain time intervals within the entire observation time range; Using Fit Model and the optimized initial moisture content and the decay constant , calculate the predicted value of moisture content at a time point; Connect the time points with the corresponding moisture content predictions to form a smooth curve of moisture content changing over time.
6. A vacuum drying effect evaluation method according to claim 2, characterized in that: Using the least squares method, the error function is defined as the residual sum of squares of all data points, including: Initialize the error function to 0; Iterate over each data point in the dataset ,in, is the current time, is the current moisture content; For each data point in the dataset, use the initial moisture content of the current iteration and the decay constant The estimated value of Calculate predicted values; Calculate the residual between the actual observed and predicted values of a data point; The residual square of each data point is added to the error function, and the process continues until all data points in the data set are traversed. The final error function value is the residual square sum of all data points.
7. A vacuum drying effect evaluation method according to claim 6, characterized in that: The calculation formula of the dynamic adjustment factor is: ; in, is the dynamic adjustment factor; , , , , and is an empirical parameter obtained by the least squares method; is the current temperature The difference from the initial temperature; is the reference temperature; is the reference pressure; It is the pressure that changes over time; is the power function of the pressure ratio from 0 to the current time The integral of is the rate of change of pressure with time.
8. A vacuum drying effect evaluation system, characterized in that: Applied to the method according to any one of claims 1 to 7, comprising: An acquisition module is used to acquire the temperature, pressure and mass change data of the evaluation object during the drying process; preprocess the temperature, pressure and mass change data to obtain preprocessed data; calculate the real-time mass change of the evaluation object during the drying process based on the preprocessed data; extract the initial mass of the evaluation object according to the real-time mass change of the evaluation object during the drying process; calculate the real-time change of its moisture content according to the initial mass of the evaluation object to obtain a preliminary curve of moisture content change over time; The processing module is used to calculate the dynamic adjustment factor according to the real-time monitored temperature and pressure data; use the dynamic adjustment factor to adjust the preliminary moisture content change curve over time to obtain the final moisture content change curve; compare the final moisture content change curve with historical data to evaluate the effect of vacuum drying.
9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Distributed energy storage robustness evaluation method based on neural network
CN118171099A
Vacuum pump cooling system performance evaluation and optimization system and method
CN118551672A
Cited By
Real-time monitoring and data analysis system based on vacuum filling and sealing equipment
CN119958644A
System for intelligently predicting and automatically adjusting freeze-drying curve
CN120373062A
Preparation method and equipment of instant Chinese yam powder
CN120713232A
