Method for automatically controlling frying temperature and time of potato chips

Abstract

The invention relates to a potato chip frying temperature and time automatic control method. Comprising the steps of data acquisition, and training, evaluation and optimization through a multiple linear regression model, and can automatically output the optimal frying temperature and frying time according to the change of potato varieties so as to achieve consistent taste. And for other factors possibly occurring during continuous production, a subsequent data set expansion method is also added, so that the purpose of continuously adjusting the optimal frying temperature and frying time according to an actual production line is achieved, and the purpose of optimizing the taste is achieved.

Description

Technical Field

[0001] The invention belongs to the field of potato chip production methods, and in particular relates to a method for automatically controlling the frying temperature and time of potato chips. Background Art

[0002] The existing potato chip production line has parameters pre-set by operators. The operators set the thickness of potato slices, blanching temperature, blanching time, frying temperature and frying time according to the final quality requirements of potato chips, such as crispness and thickness of potato chips. However, due to differences in raw material properties, such as different types of potatoes and starch content, the potato chip production line cannot automatically match production parameters according to the different characteristics of raw materials, resulting in a large difference between the actual production parameters and the production parameters required for raw material matching. This difference is difficult for operators to discover during the production process, resulting in fluctuations in product quality, affecting product quality indicators such as taste, color, and crispness.

[0003] For example, a Chinese invention patent: A method for preparing crispy instant flavored French fries; Publication number: CN 116326743 A, this patent discloses a method for preparing crispy instant flavored French fries, comprising the following steps: (1) selecting fresh potatoes without buds and damage, peeling and washing, and evenly cutting into slices with a thickness of 3 to 4 mm; (2) steaming the potato slices, and mashing them into uniform and particle-free potato puree while still hot after steaming; (3) weighing the raw materials and auxiliary materials according to the formula, mixing the auxiliary materials evenly, and then adding the potato puree, and repeatedly kneading the dough to mix the raw materials and auxiliary materials; (4) putting the dough into an extruder for extrusion and forming; (5) baking for preliminary expansion, and requiring the French fries to have a uniform color and no burnt color; (6) sending the French fries after baking into a vacuum low-temperature fryer for frying and de-oiling; (7) packaging the French fries after cooling. The method directly steams and cooks potatoes to make puree, adds auxiliary materials to form, and then vacuum-fryes at low temperature to obtain a finished product. The finished product does not need to be fried again before eating, and the method is convenient and quick, and is a necessary snack for outdoor parties and picnics.

[0004] Most of the existing patents only provide a production method like this patent, and give a rough range. In actual production, due to the different types of potatoes, starch content, etc., the specific temperature and time still need to be manually assessed. When producing according to the method disclosed in this existing patent, using the same formula will still produce a large difference in taste, which is tantamount to having an adverse effect on the overall result of the product when pursuing the taste of potato chips. Summary of the invention

[0005] The present invention is aimed at existing invention applications that mainly protect formulas. The formula ranges given are often not precise values, and the specific potato types targeted by the formulas do not provide data such as starch content, which makes it impossible to fully replicate the effect of the patented formula in actual use. For this reason, this patent uses machine learning methods and self-built data sets to provide a universal method for automatic temperature and time adjustment, so that the formula is universal and the data provided is accurate without manual adjustment.

[0006] To this end, the present invention provides a method for automatically controlling the frying temperature and time of potato chips, comprising the following steps: Step 1: The starch content and slice thickness of various potato varieties are measured in the laboratory, and discrete sampling is performed within the commonly used oil temperature range and frying time range; the taste of the results is evaluated; the data is recorded using a one-hot encoding method and collected into a data set; Step 2, dividing the collected data into a training set, a validation set, and a test set; normalizing the starch content and potato slice thickness in the training set, the validation set, and the test set; Step 3, constructing a multiple linear regression model based on the collected data to form a frying temperature prediction model and a frying time prediction model respectively; Step 4: Train the model, calculate the gradient of the regression coefficient in the model, and use the Adam optimization algorithm to update the regression coefficient of the model according to the calculated gradient; Step 5: Calculate the mean square error, mean absolute error, and coefficient of determination, and use the test set data to evaluate the trained model; Step 6: Apply the trained model evaluated in step 5 to the production equipment. Step 7: Regularly collect potato chip samples actually produced from the production line for taste evaluation. If there is any deviation, retrain and optimize the model.

[0007] Furthermore, in step 2, the collected data is divided into a training set, a validation set, and a test set in a ratio of 70%:15%:15%; The starch content and potato slice thickness in the training set, validation set, and test set were normalized so that their values ​​ranged from 0 to 1. The formula is as follows: in, is the normalized value, is the original value, and are the minimum and maximum values ​​of the feature in the dataset, respectively.

[0008] The step 3 comprises: Step 3-1, set the input feature vector to ,in is the one-hot encoded vector of potato species, is the starch content, is the thickness of the potato slices; For frying temperature prediction, the model is expressed as: in is the frying temperature, is the regression coefficient to be estimated, is the error term; For frying time prediction, the model is: in It's frying time. is the regression coefficient to be estimated, is the error term; Step 3-2, randomly initialize the regression coefficients and , The given random number range is -0.1 to 0.1, and the type is a uniformly distributed random number; Step 3-3, using mean square error as the loss function, we have: Frying temperature prediction model, the loss function calculation formula is: in is the number of samples in the training set, It is The actual frying temperature of the samples, is the frying temperature predicted by the model; Frying time prediction model, the loss function calculation formula is: in yes The actual frying time of the sample, is the frying time predicted by the model.

[0009] In step 4, the number of training rounds is set to 1000 rounds, which specifically includes: Step 4-1: For each round of training, traverse each sample in the training set ,in It is The feature vector of the samples, is the corresponding actual frying temperature, is the corresponding actual frying time; Step 4-2, Substitute into the frying temperature prediction model and calculate the predicted temperature , according to the loss function Calculate the loss of temperature prediction and calculate the gradient of the temperature prediction model with respect to the regression coefficient; Step 4-3, Substitute into the frying time prediction model and calculate the predicted time , according to the loss function Calculate the loss of time prediction and calculate the gradient of the time prediction model with respect to the regression coefficients; Step 4-4, using the Adam optimization algorithm to update the regression coefficients of the frying temperature and frying time prediction model according to the calculated gradient; Step 4-4-1, set the learning rate , hyperparameters ; Step 4-4-2, calculate the first-order moment estimate of the gradient ; Second moment estimate ; Where t is the number of rounds of current training; is the current gradient; Step 4-4-3, Modify the first-order moment estimate ; Second moment estimate ; Step 4-4-4, update parameters ,parameter To insert the regression coefficient and , ; Step 4-5: After every 100 rounds of training, calculate the loss value of the model on the validation set and record it; the loss value includes and .

[0010] The step 5 comprises: Step 5-1, use the test set data to evaluate the trained model; For the frying temperature prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ; ; ; in is the number of samples in the test set, is the actual frying temperature of the i-th test sample, is the frying temperature predicted by the model, is the average of the actual frying temperatures in the test set; For the frying time prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ; ; ; in is the number of samples in the test set, It is The actual frying time of the test samples, is the frying time predicted by the model, is the average of the actual frying time in the test set; Step 5-2, evaluating the model performance according to the calculation in step 5-1; if and Close to 0, and When it is close to 1, it means that the prediction accuracy of the model is high.

[0011] The step 6 comprises: Step 6-1, in the potato chip production line, install a potato starch content detection device; when the potatoes enter the production line entrance after being washed and peeled, the starch content of the potatoes is automatically detected, and at the same time, the potato type information and slice thickness information are recorded; The slice thickness is automatically recorded by the slicer; Step 6-2, the collected potato type, starch content and slice thickness information are combined into a feature vector, and substituted into the frying temperature and frying time prediction model trained in step 5 to obtain the predicted frying temperature and frying time; Step 6-3, transmitting the predicted frying temperature and time parameters to the control system of the fryer, automatically adjusting the temperature setting value and frying time setting value of the fryer, and realizing automatic control of the potato chip frying process.

[0012] In step 7, the taste is evaluated using the evaluation method in step 1 to ensure that it is the same as the evaluation method used during model training; The actual crispness and taste values ​​are compared with the target crispness and taste values. If there is a deviation, these new data samples are added to the original data set and the model is retrained and optimized.

[0013] The step 7 comprises: Step 7-1, calculate the deviation value, and set the taste evaluation result of the potato chip sample actually collected to , and the target taste evaluation result recorded in the data set is , then we get the deviation value , ; Step 7-2, set the threshold of the deviation value to ±0.5, and if it exceeds the threshold, proceed to step 7-3 to record the data; Step 7-3, recording the potato type information, actual frying temperature, and actual frying time in step 7-1 as information data, where the record refers to a new data set; merging the data sets in step 1 of the new data set into an expanded data set; Step 7-4, retrain the model using the merged dataset and repeat steps 2 to 6.

[0014] The substantial effect of the present invention is that, in addition to adjusting the taste, the present application can automatically output the optimal frying temperature and frying time according to the potato variety changes through the self-built data set and machine learning method to achieve a consistent taste. In addition, in view of other factors that may arise during continuous production, a subsequent data set expansion method is also added to continuously adjust the optimal frying temperature and frying time according to the actual production line to achieve the goal of optimizing the taste. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flowchart of the present invention. DETAILED DESCRIPTION

[0016] The technical solution of the present invention is further described in detail below through specific embodiments in conjunction with the accompanying drawings. Example

[0017] like Figure 1 As shown, the present invention provides a method for automatically controlling the frying temperature and time of potato chips, comprising the following steps: 1. Establish a data set.

[0018] The basic formula determines that for temperature , considering the thickness The bigger the potato chips, the higher the temperature is needed to fry them thoroughly. The higher the temperature, the more you need to raise the temperature to make the chips crispier. You can assume that the temperature formula is roughly: ; in is a coefficient related to thickness, starch content and crispness, Is a reference temperature value.

[0019] For example, through experimental and empirical data fitting, The value can be 10 (which means that for every 1mm increase in thickness, the temperature will increase by about 10 degrees Celsius). The value is 5 (for every 1% increase in starch content, the temperature increases by about 5 degrees Celsius). The value is 5 (for every increase of 1 in the brittleness coefficient, the temperature increases by about 5 degrees Celsius), The value is 120 (basic frying temperature).

[0020] For time , considering the thickness The bigger it is, the longer it takes to fry and the starch content The higher the temperature, the longer it may take to keep the chips crispy. We can assume that the time formula is roughly: ; in is the coefficient, is the base time.

[0021] For example, The value is 30 (for every 1mm increase in thickness, the time increases by about 30 seconds). The value is 10 (for every 1% increase in starch content, the time increases by about 10 seconds). The value is 10 (the time increases by about 10 seconds for every increase in the brittleness coefficient). The value is 60 (basic frying time).

[0022] For the coefficient This requires a lot of experimentation to determine and adjust.

[0023] For example, a series of potato slices of different thicknesses and starch contents are prepared, and different combinations of temperature and time are set for frying. The actual crispness of the chips is then determined through sensory evaluation or professional crispness testing equipment. The coefficient is corrected based on the actual results so that the calculated temperature and time can meet the desired crispness as accurately as possible. .

[0024] Collect data on different potato varieties, starch content, potato chip thickness, frying temperature, frying time, and the corresponding crispy taste of potato chips. For a variety of different potato varieties, measure their starch content and make potato slices of different thicknesses. Then, conduct frying experiments under different combinations of frying temperature and frying time to obtain the crisp taste evaluation of potato chips under each combination. The crisp taste evaluation can be scored by professional judges or quantitatively measured using a texture analyzer.

[0025] These data are obtained through a large number of experiments.

[0026] Potato Varieties: One-Hot Encoding can be used to convert different potato varieties into vector form. For example, if there are 5 potato varieties, then a particular variety can be represented as a 5-dimensional vector, where only one element is 1 and the rest are 0.

[0027] Starch content: directly use the original starch content value as a feature. Assume that the starch content of potatoes is Unit: %, value range is 10%-25%.

[0028] Potato slice thickness: Also use the original thickness value as a feature. Suppose the thickness of the potato slice is Unit: mm, value range is 1-5mm.

[0029] Frying temperature: target output value, set the frying temperature to Unit: Celsius.

[0030] Frying time: target output value, set the frying time to Unit: s.

[0031] Desired crunchiness The coefficient range is 1-10, 1 means not brittle and 10 means the best.

[0032] 2. Model selection: select the multivariate linear regression model for model training.

[0033] Suppose our input feature vector is (in is the one-hot encoded vector of potato species, is the starch content, is the thickness of potato slices), the output feature vector is (in is the frying temperature, is the frying time). Then the linear regression model is expressed as ,in is the regression coefficient matrix, and e is the error term. Through training data, we can use methods such as least squares to estimate And it is usually assumed that e has a mean of 0 and a variance of The normal distribution of .

[0034] Define the residual sum of squares: Estimated regression coefficients The corresponding predicted value is , residual .

[0035] The residual sum of squares (RSS) is defined as: ； In the chip context, it is the sum of the squares of the differences between the actual potato chip quality index values ​​and the predicted values ​​calculated by the estimated coefficients under all observed data.

[0036] Find the regression coefficient that minimizes the residual sum of squares: Expand ; ; because is a scalar, and ,so , Then the above formula can be simplified to: ; right about Find the partial derivative: ; Let the partial derivative be 0, that is: ; Solving this equation yields: ; In the context of the potato chip production line, the calculated It is the regression coefficient that can minimize the residual sum of squares when predicting potato chips quality indicators based on frying temperature and frying time, thereby determining the specific quantitative relationship between the impact of frying temperature and frying time on potato chips quality indicators.

[0037] Then the feature vector for the input is , we can get: For frying temperature prediction, the model is expressed as: ; in is the frying temperature, is the regression coefficient to be estimated, is the error term; For frying time prediction, the model is: ; in It's frying time. is the regression coefficient to be estimated, is the error term.

[0038] 3. Model training, use the collected data set for training: Model initialization, Randomly initialize regression coefficients and , a uniformly distributed random number in the range [-0.1, 0.1].

[0039] Frying temperature prediction model, the loss function calculation formula is: ; in is the number of samples in the training set, It is The actual frying temperature of the samples, is the frying temperature predicted by the model; Frying time prediction model, the loss function calculation formula is: ; in yes The actual frying time of the sample, is the frying time predicted by the model.

[0040] Use the Adam optimization algorithm to optimize the model: Setting the learning rate , hyperparameters (These parameters can be adjusted according to the training effect).

[0041] In each round of training, the regression coefficients are updated according to the Adam algorithm: ; ; ; ; ; in and are the first-order moment estimates and second-order moment estimates of the gradient, and are the modified first-order moment estimates and second-order moment estimates, is the current gradient, is the current regression coefficient (including β and γ series coefficients).

[0042] Set the number of training rounds .

[0043] For each round of training: traverse each sample in the training set Substitute into the frying temperature prediction model and calculate the predicted temperature , according to the loss function Calculate the loss for temperature prediction and compute the gradient of the temperature prediction model with respect to the regression coefficients.

[0044] Will Substitute into the frying time prediction model and calculate the predicted time , according to the loss function Compute the loss for temporal predictions and calculate the gradient of the temporal prediction model with respect to the regression coefficients.

[0045] The Adam optimization algorithm is used to update the regression coefficients of the frying temperature and frying time prediction models according to the calculated gradients.

[0046] After every 100 training rounds, the model loss on the validation set is calculated (including and ), and record it.

[0047] 4. Model evaluation and optimization: Use the test set data to evaluate the performance of the model.

[0048] For the multivariate linear regression model, the mean square error (MSE) and mean absolute error (MAE) are used to measure the accuracy of model prediction.

[0049] The mean square error (MSE) measures the average square error between the model's predicted value and the actual value. The smaller the value, the better the model's fitting effect.

[0050] Mean absolute error (MAE) represents the average absolute value of the error between the model prediction value and the actual value. The smaller the value, the better the model.

[0051] Coefficient of determination : The value range is between 0 and 1. The closer it is to 1, the higher the variance ratio of the dependent variable that the model can explain, that is, the better the model fit is.

[0052] For the frying temperature prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ; ; ; in is the number of samples in the test set, is the actual frying temperature of the i-th test sample, is the frying temperature predicted by the model, is the average of the actual frying temperatures in the test set.

[0053] For the frying time prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ; ; ; in is the number of samples in the test set, It is The actual frying time of the test samples, is the frying time predicted by the model, is the average of the actual frying time in the test set.

[0054] if and Close to 0, and When it is close to 1, it means that the prediction accuracy of the model is high.

[0055] If the model performance is not ideal, adjust the model's hyperparameters and train again.

[0056] 5. Put the trained model into production: In the potato chip production line, potato starch content detection equipment (such as near-infrared spectrometer) is installed to automatically detect the starch content of potatoes when they enter the production line after being washed and peeled. At the same time, the potato variety information (which can be identified by a pre-set variety identification system) and slice thickness information (automatically recorded by the slicing equipment) are recorded.

[0057] The collected information on potato types, starch content and slice thickness are combined into a feature vector and substituted into the trained frying temperature and frying time prediction model to obtain the predicted frying temperature and frying time.

[0058] The predicted frying temperature and time parameters are transmitted to the control system of the fryer, and the temperature setting value and frying time setting value of the fryer are automatically adjusted to realize automatic control of the potato chip frying process.

[0059] Regularly collect actual potato chip samples from the production line for crispness and taste evaluation (using the same evaluation method as that used for model training).

[0060] The actual crispness and taste values ​​are compared with the target crispness and taste values. If there is a large deviation, these new data samples are added to the original data set, and the model is retrained and optimized to continuously improve the accuracy and adaptability of the model.

[0061] 6. In view of the possible large deviations, conduct another model training to calibrate the results: The taste evaluation results of the potato chips samples actually collected are set to , and the target taste evaluation result recorded in the data set is , then we get the deviation value , ; At the same time, the potato variety information corresponding to the production of this batch of potato chips is collected (in is the one-hot encoded vector of potato species, is the starch content, is the thickness of the potato slices) as well as the actual frying temperature and frying time data.

[0062] Set an acceptable deviation threshold , set to 0.5 (this is based on the taste rating using a 0-10 point system).

[0063] if , then the judgment deviation is large and model optimization is needed; if , it is considered that the current model performs well in this batch of production, and production continues with regular monitoring.

[0064] The newly collected potato chip sample data with deviations and their corresponding feature information are organized into a new data set; the new data set is merged with the original data set to obtain an expanded data set.

[0065] Retrain the model using the augmented dataset.

[0066] Through the above-mentioned repetitive process, feedback data from actual production is continuously incorporated into model training, which can gradually improve the accuracy and adaptability of the model, achieve more precise control of the potato chip frying process, and produce potato products that meet the target crispness and taste requirements, while improving production efficiency and product quality stability.

Claims

1. A method for automatically controlling the frying temperature and time of potato chips, characterized in that: The following steps are involved: Step 1: The starch content and slice thickness of various potato varieties are measured in the laboratory, and discrete sampling is performed within the commonly used oil temperature range and frying time range; the taste of the results is evaluated; the data is recorded using a one-hot encoding method and collected into a data set; Step 2, dividing the collected data into a training set, a validation set, and a test set; normalizing the starch content and potato slice thickness in the training set, the validation set, and the test set; Step 3, constructing a multiple linear regression model based on the collected data to form a frying temperature prediction model and a frying time prediction model respectively; Step 4: Train the model, calculate the gradient of the regression coefficient in the model, and use the Adam optimization algorithm to update the regression coefficient of the model according to the calculated gradient; Step 5: Calculate the mean square error, mean absolute error, and coefficient of determination, and use the test set data to evaluate the trained model; Step 6: Apply the trained model evaluated in step 5 to the production equipment. Step 7: Regularly collect potato chip samples actually produced from the production line for taste evaluation. If there is any deviation, retrain and optimize the model.

2. The method for automatically controlling the frying temperature and time of potato chips according to claim 1, characterized in that: In step 2, the collected data is divided into a training set, a validation set, and a test set in a ratio of 70%:15%:15%; The starch content and potato slice thickness in the training set, validation set, and test set were normalized so that their values ​​ranged from 0 to 1. The formula is as follows: in, is the normalized value, is the original value, and are the minimum and maximum values ​​of the feature in the dataset, respectively.

3. The method for automatically controlling the frying temperature and time of potato chips according to claim 1, characterized in that: The step 3 comprises: Step 3-1, set the input feature vector to ,in is the one-hot encoded vector of potato species, is the starch content, is the thickness of potato slices; For frying temperature prediction, the model is expressed as: in is the frying temperature, is the regression coefficient to be estimated, is the error term; For frying time prediction, the model is: in It's frying time. is the regression coefficient to be estimated, is the error term; Step 3-2, randomly initialize the regression coefficients and , The given random number range is -0.1 to 0.1, and the type is a uniformly distributed random number; Step 3-3, using mean square error as the loss function, we have: Frying temperature prediction model, the loss function calculation formula is: in is the number of samples in the training set, It is The actual frying temperature of the samples, is the frying temperature predicted by the model; Frying time prediction model, the loss function calculation formula is: in yes The actual frying time of the sample, is the frying time predicted by the model.

4. The method for automatically controlling the frying temperature and time of potato chips according to claim 1, characterized in that: In step 4, the number of training rounds is set to 1000 rounds, which specifically includes: Step 4-1: For each round of training, traverse each sample in the training set ,in It is The feature vector of the samples, is the corresponding actual frying temperature, is the corresponding actual frying time; Step 4-2, Substitute into the frying temperature prediction model and calculate the predicted temperature , according to the loss function Calculate the loss of temperature prediction and calculate the gradient of the temperature prediction model with respect to the regression coefficient; Step 4-3, Substitute into the frying time prediction model and calculate the predicted time , according to the loss function Calculate the loss of time prediction and calculate the gradient of the time prediction model with respect to the regression coefficients; Step 4-4, using the Adam optimization algorithm to update the regression coefficients of the frying temperature and frying time prediction model according to the calculated gradient; Step 4-5: After every 100 rounds of training, calculate the loss value of the model on the validation set and record it; the loss value includes and .

5. The method for automatically controlling the frying temperature and time of potato chips according to claim 4, characterized in that: The step 4-4 comprises: Step 4-4-1, set the learning rate , hyperparameters ; Step 4-4-2, calculate the first-order moment estimate of the gradient ; Second moment estimate ; Among them, the corner mark is the current training round number; is the current gradient; Step 4-4-3, Modify the first-order moment estimate ; Second moment estimate ; Step 4-4-4, update parameters ,parameter To insert the regression coefficient and , .

6. The method for automatically controlling the frying temperature and time of potato chips according to claim 1, characterized in that: The step 5 comprises: Step 5-1, for the frying temperature prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ； ； ;in is the number of samples in the test set, is the actual frying temperature of the i-th test sample, is the frying temperature predicted by the model, is the average of the actual frying temperatures in the test set; For the frying time prediction model, calculate the mean square error , mean absolute error and the coefficient of determination , the calculation formula is as follows: ； ； ； in is the number of samples in the test set, It is The actual frying time of the test samples, is the frying time predicted by the model, is the average of the actual frying time in the test set; Step 5-2, evaluating the model performance according to the calculation in step 5-1; if and Close to 0, and When it is close to 1, it means that the prediction accuracy of the model is high.

7. The method for automatically controlling the frying temperature and time of potato chips according to claim 1, characterized in that: The step 6 comprises: Step 6-1, in the potato chip production line, install a potato starch content detection device; when the potatoes enter the production line entrance after being washed and peeled, the starch content of the potatoes is automatically detected, and at the same time, the potato type information and slice thickness information are recorded; The slice thickness is automatically recorded by the slicer; Step 6-2, the collected potato type, starch content and slice thickness information are combined into a feature vector, and substituted into the frying temperature and frying time prediction model trained in step 5 to obtain the predicted frying temperature and frying time; Step 6-3, transmitting the predicted frying temperature and time parameters to the control system of the fryer, automatically adjusting the temperature setting value and frying time setting value of the fryer, and realizing automatic control of the potato chip frying process.

8. The method for automatically controlling the frying temperature and time of potato chips according to claim 1 or 2 or 3 or 4 or 5 or 6 or 7, characterized in that: The step 7 comprises: Step 7-1, calculate the deviation value, and set the taste evaluation result of the potato chip sample actually collected to , and the target taste evaluation result recorded in the data set is , then we get the deviation value , ; Step 7-2, set the threshold of the deviation value to ±0.5, and if it exceeds the threshold, proceed to step 7-3 to record the data; Step 7-3, recording the potato type information, actual frying temperature, and actual frying time in step 7-1 as information data, the record refers to a new data set; merging the data set in step 1 of the new data set into an expanded data set; Step 7-4, retrain the model using the merged dataset and repeat steps 2 to 6.

Images