CN-122022551-A - Quality consistency-based potato processing technology parameter optimizing method and system

CN122022551ACN 122022551 ACN122022551 ACN 122022551ACN-122022551-A

Abstract

The invention discloses a potato processing technology parameter optimizing method and system based on quality consistency, and belongs to the technical field of data processing. The method comprises the steps of constructing a digital twin model of a mapping potato processing production line, wherein the digital twin model receives raw material characteristic data, environment state data and process parameter data of each process as input, outputs prediction energy consumption data, prediction yield data and a prediction mean value and a prediction standard deviation of each product quality index, acquires current raw material characteristic data and current environment state data, when an optimization triggering condition is met, takes the current raw material characteristic data and the current environment state data as fixed input, adopts a hierarchical optimization algorithm, invokes the digital twin model as a predictor, solves an optimization problem, and outputs an optimal process parameter set for minimizing an objective function. The invention can dynamically adjust the technological parameters of each working procedure of the potato processing production line, and is suitable for the condition of raw material fluctuation or environmental state change.

Inventors

WANG JIANFEI
WANG NAN
LI JIA
Fan Leixin
YANG LIJUN
MA JUNJIE
SUN KAI

Assignees

雪川六盘山食品(宁夏)有限公司

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (10)

1. A potato processing technology parameter optimizing method based on quality consistency is characterized by comprising the following steps: Constructing a digital twin model of a mapped potato processing production line, wherein the digital twin model receives raw material characteristic data, environment state data and process parameter data of each process as inputs, and outputs predicted energy consumption data, predicted yield data, and predicted mean values and predicted standard deviations of quality indexes of each product; Acquiring current raw material characteristic data and current environment state data; when the optimization triggering condition is met, the current raw material characteristic data and the current environment state data are used as fixed inputs, a hierarchical optimization algorithm is adopted, the digital twin model is called as a predictor, an optimization problem is solved, and an optimal process parameter set for minimizing an objective function is output; The optimized triggering condition is at least one of starting production of a new batch of raw materials, change of raw material characteristics in the production process, change of environmental conditions in the production process and reaching a preset period; The optimization problem comprises decision variables, objective functions and constraint conditions; The decision variable is an adjustable technological parameter set; The objective function is to minimize the overall cost of single batch expectations ; Wherein E (X|R, env) is the predicted total energy consumption under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env; Y (X|R, env) is the predicted qualified product yield under the current technological parameter vector X, the current raw material characteristic vector R and the current environment state vector Env, sigma I (X|R, env) is the predicted standard deviation of the ith product quality index, i=1, 2, & gt, n is the total number of the product quality indexes, n is an integer, c_energy is the real-time energy comprehensive unit price obtained from an energy management system, c_product is the factory average price of the product obtained from an ERP system, c_sigma I is the fluctuation punishment coefficient of the ith product quality index, pi is more than or equal to 1, c_punishment is the safety violation punishment, I { and } is an indication function, and 1 is taken when any product quality index exceeds the safety limit, otherwise 0 is taken; The constraint conditions comprise quality constraint, wherein the quality constraint is that for the ith product quality index, mu i (X|R, env) +k-sigma i (X|R, env) is less than or equal to USL_i, mu_i (X|R, env) -k-sigma i (X|R, env) is more than or equal to LSL_i, k is a process capability coefficient, mu i (X|R, env) is a prediction mean value of the ith product quality index under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env, and USL_i and LSL_i are respectively the upper limit and the lower limit of a standard value of the ith product quality index.
2. The method of claim 1, wherein the feedstock property data comprises moisture content, dry matter content, reducing sugar content, starch content, protein content, size, and specific gravity; The process parameter data comprise cleaning time, cleaning water temperature, water flow speed, steam pressure, steam time, peeling knife rotating speed, cutting thickness, cutting speed, blanching temperature, blanching time, color fixative concentration, frying temperature, frying time, oil quantity, freezing temperature and freezing time; The environmental status data includes temperature, humidity, and carbon dioxide concentration; the product quality indexes comprise oil content, moisture content, color and crisp degree; the predicted energy consumption data comprise total energy consumption, energy consumption of each unit and energy consumption of unit products; the predicted yield data includes yield per unit time and total yield.
3. The method according to claim 1 or 2, wherein the digital twin model comprises a quality prediction model, an energy consumption prediction model and a yield prediction model, each of which is a hybrid model comprising a mechanism part and a data driving part; The data driving part of the quality prediction model is integrated by a fully connected neural network, a traditional kernel method and a gradient lifting tree; The traditional kernel method is support vector regression, gaussian process regression, support vector machine, kernel ridge regression or kernel principal component analysis; the gradient lifting tree is limit gradient lifting XGBoost, lightweight gradient lifting machine LightGBM or category gradient lifting CatBoost; And the data driving parts of the energy consumption prediction model and the yield prediction model adopt gradient lifting trees.
4. A method according to claim 3, wherein the mechanism part of the mass prediction model describes the moisture diffusion of the drying process using the phillips law, the maillard reaction using the arrhenius equation, and the oil content is calculated using the thermal equilibrium equation; the mechanism part of the energy consumption prediction model adopts an energy balance equation and heat and mass transfer to calculate energy consumption; The mechanism part of the yield prediction model adopts an equipment beat and bottleneck analysis model and a raw material-yield relation model.
5. The method according to claim 1, 2 or 4, wherein the calculation formula of the fluctuation penalty coefficient c_σi of the ith product quality indicator is: c_σi=(Si+β·S 0 )/(σ̄i+ε); Wherein Si is the total amount of compensation of relevant clients of the ith product quality index of the last N months, sigma ̄ i is the historical rolling standard deviation mean value of the ith product quality index, beta is a smoothing coefficient, and epsilon takes the value range of 1 multiplied by 10- 8 ～1×10⁻ 4 .
6. The method of claim 1, 2 or 4, wherein the hierarchical optimization algorithm is a three-layer optimization algorithm, wherein a first layer adopts a constraint propagation algorithm to conduct feasibility screening, removes candidate solutions violating hard constraints and outputs a feasible domain, a second layer adopts a Bayesian optimization algorithm to conduct global exploration on the candidate solutions in the feasible domain, evaluates multiple candidate solutions in parallel and outputs multiple feasible points, and a third layer adopts a sequence quadratic programming algorithm to conduct local search and output the optimal solution by taking the optimal feasible point in the feasible points as an initial value.
7. The method of claim 1,2 or 4, further comprising: and obtaining an actual measurement value of the product quality index fed back by the online detector in real time, comparing the actual measurement value with a target value, and finely adjusting the technological parameters.
8. A potato processing technology parameter optimizing system based on quality consistency is characterized by comprising: The model construction module is used for constructing a digital twin model of the mapped potato processing production line, and the digital twin model receives raw material characteristic data, environment state data and process parameter data of each procedure as inputs and outputs predicted energy consumption data, predicted yield data, and predicted mean values and predicted standard deviations of quality indexes of each product; the data acquisition module is used for acquiring the characteristic data of the current raw materials and the current environmental state data; the parameter optimization module is used for taking the current raw material characteristic data and the current environment state data as fixed inputs when the optimization triggering condition is met, adopting a hierarchical optimization algorithm, calling the digital twin model as a predictor, solving an optimization problem and outputting an optimal process parameter set for minimizing an objective function; The optimized triggering condition is at least one of starting production of a new batch of raw materials, change of raw material characteristics in the production process, change of environmental conditions in the production process and reaching a preset period; The optimization problem comprises decision variables, objective functions and constraint conditions; The decision variable is an adjustable technological parameter set; The objective function is to minimize the overall cost of single batch expectations ; Wherein E (X|R, env) is the predicted total energy consumption under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env; Y (X|R, env) is the predicted qualified product yield under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env, sigma I (X|R, env) is the predicted standard deviation of the ith product quality index, i=1, 2, & gt, n is the total number of the product quality indexes, n is an integer, c_energy is the real-time energy comprehensive unit price obtained from an energy management system, c_product is the factory average price of the product obtained from an ERP system, c_sigma I is the fluctuation penalty coefficient of the ith product quality index, pi is more than or equal to 1 (key indexes such as food safety correlation, pi takes a large value such as 2.0, secondary indexes such as color, pi takes 1.2-1.5), c_penalty is the safety violation penalty (takes a large value such as 10. 6 ), I { is an indication function, and any product quality index takes 1 when exceeding the safety limit, otherwise takes 0; The constraint conditions comprise quality constraint, wherein the quality constraint is that for the ith product quality index, mu i (X|R, env) +k-sigma i (X|R, env) < USL_i and mu_i (X|R, env) -k-sigma i (X|R, env) > LSL_i, k is a process capability coefficient (such as k=3), mu i (X|R, env) is a prediction mean value of the ith product quality index under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env, and USL_i and LSL_i are respectively the upper limit and the lower limit of a standard value of the ith product quality index.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements a quality consistency based potato processing parameter optimization method as claimed in any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the quality consistency based potato processing parameter optimizing method of any one of claims 1 to 7.

Description

Quality consistency-based potato processing technology parameter optimizing method and system Technical Field Embodiments of the present invention relate to the field of data processing technology. More particularly, the invention relates to a potato processing technology parameter optimizing method and system based on quality consistency. Background Potato processing refers to the process of converting fresh potatoes, either physically, chemically or biologically, into various forms of products, starting from the potato. Generally comprises the process steps of cleaning, peeling, cutting, blanching, frying, freezing and the like. Because of natural fluctuation of potato raw material characteristics (such as starch content, moisture, reducing sugar content and the like), if process parameters are fixed, unstable product quality (such as uneven color of potato chips, excessive acrylamide content and the like) is easily caused, and therefore, the process parameters of potato processing procedures need to be optimized in real time according to the change of potato raw material characteristics. Currently, statistical methods, such as analysis of variance (ANOVA, analysis of Variance), are generally used to optimize potato processing parameters, and by comparing the inter-group variation with the intra-group variation in experimental data, it is determined whether different process parameters or combinations of process parameters have significant impact on a single target (e.g., product quality, yield, energy consumption, etc.). The direct impact of the process parameters on the single target is then evaluated in combination with the professional evaluator score. The method is dependent on experience, simple experiment design and single-target optimization, is easy to fall into a local optimal solution, and is difficult to comprehensively capture complex nonlinear relations between process parameters and targets. In addition, the optimization method generally adopts a static optimization scheme, and is difficult to adapt to the condition of fluctuation of raw materials or change of environmental conditions. Disclosure of Invention In order to solve at least one technical problem as mentioned above, the invention provides a potato processing technology parameter optimizing method and system based on quality consistency. The technical problems to be solved by the invention are realized by the following technical scheme: the first aspect of the invention provides a potato processing technology parameter optimizing method based on quality consistency, which comprises the following steps: Constructing a digital twin model of a mapped potato processing production line, wherein the digital twin model receives raw material characteristic data, environment state data and process parameter data of each process as inputs, and outputs predicted energy consumption data, predicted yield data, and predicted mean values and predicted standard deviations of quality indexes of each product; Acquiring current raw material characteristic data and current environment state data; when the optimization triggering condition is met, the current raw material characteristic data and the current environment state data are used as fixed inputs, a hierarchical optimization algorithm is adopted, the digital twin model is called as a predictor, an optimization problem is solved, and an optimal process parameter set for minimizing an objective function is output; The optimized triggering condition is at least one of starting production of a new batch of raw materials, change of raw material characteristics in the production process, change of environmental conditions in the production process and reaching a preset period; The optimization problem comprises decision variables, objective functions and constraint conditions; The decision variable is an adjustable technological parameter set; The objective function is to minimize the overall cost of single batch expectations ; Wherein E (X|R, env) is the predicted total energy consumption under the current process parameter vector X, the current raw material characteristic vector R and the current environment state vector Env; Y (X|R, env) is the predicted qualified product yield under the current technological parameter vector X, the current raw material characteristic vector R and the current environment state vector Env, sigma I (X|R, env) is the predicted standard deviation of the ith product quality index, i=1, 2, & gt, n is the total number of the product quality indexes, n is an integer, c_energy is the real-time energy comprehensive unit price obtained from an energy management system, c_product is the factory average price of the product obtained from an ERP system, c_sigma I is the fluctuation punishment coefficient of the ith product quality index, pi is more than or equal to 1, c_punishment is the safety violation punishment, I { and } is an indication function, and 1 is taken when any