Multi-objective process parameter optimization method for tea production line based on improved particle swarm

By improving the particle swarm algorithm and multi-objective optimization method, a moisture content and chemical composition prediction model of the tea production line was established, and the single problem of process parameter optimization in tea processing was solved, and the tea quality improvement and process parameters were improved are achieved. The diversity and convergence of process parameters were achieved. , meets the multi-unit collaborative optimization needs of tea production.

CN118627851BActive Publication Date: 2024-11-12ZHEJIANG GENGXIANG ORGANIC TEA DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411084044.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-12
Estimated Expiration
2044-08-08

AI Technical Summary

Technical Problem

The existing tea processing technology lacks digitalization, informatization and automation, which makes it difficult to guarantee the quality of tea. Optimizing a certain process parameter alone cannot guarantee the optimal quality of finished tea.

Method used

Using the improved particle swarm algorithm, a multi-objective process parameter optimization method is used to establish a moisture content and chemical composition prediction model of each unit of the tea production line. Combined with image information, a collaborative optimization model for the initial production process parameters of tea is constructed, and an adaptive update strategy and angle are used. The divided multi-objective particle swarm optimization algorithm solves the optimal process parameter combination.

Benefits of technology

The collaborative optimization of multi-unit process parameters of the tea production line has been achieved, the quality of finished tea has been improved, the diversity and convergence of process parameters has been ensured, and the actual production needs have been met.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118627851B_ABST
    Figure CN118627851B_ABST
Patent Text Reader

Abstract

The invention discloses a multi-objective process parameter optimization method for a tea production line based on an improved particle swarm, comprising the following steps: collecting data such as process parameters of primary tea production, establishing a moisture content prediction model and a chemical composition prediction model for each unit of the production line after preprocessing; establishing a collaborative optimization model for the process parameters of primary tea production based on the above model, taking the process parameters of each unit as decision variables, and the moisture content of each unit after processing and the chemical composition of raw tea as production targets; improving a particle speed and position update method and a leader particle selection method for the particle swarm in view of the disadvantages of the particle swarm being easy to be selected into a local optimum; taking the process parameters of actual application as constraint conditions, and the moisture content of each unit and a plurality of chemical compositions as targets, applying an improved particle swarm algorithm to solve an optimization model; and outputting an optimal moisture content and chemical composition optimal solution set and a process parameter solution set scheme after multiple iterations of the improved particle swarm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of primary processing of tea, and in particular relates to a multi-objective process parameter optimization method for a tea production line based on an improved particle group. Background Art

[0002] Tea processing requires a certain amount of experience in making tea, and requires workers to be able to adjust the process parameters of each processing unit in real time according to the softness, color and other information of the tea leaves to produce high-quality tea leaves. Currently, most tea processing still relies on the traditional method of "looking, touching and smelling" to judge the quality of tea leaves such as moisture content and appearance during processing, which requires a high level of experience, or uses traditional PID control to simply control process parameters such as temperature. There are problems such as low levels of digitization and informatization, and lack of automation, precision and intelligence.

[0003] At present, the optimization of the tea processing process is mostly focused on optimizing and improving the process parameters, control methods and equipment structure parameters of a certain unit. It does not take into account that tea processing is composed of multiple units that influence each other, and the quality of the finished tea is determined by multiple process units. If only a certain unit is optimized, or multiple units are optimized separately, it cannot be guaranteed that the most ideal tea quality can be obtained in the end.

[0004] For example, reference 1 (Xiang Lihui, Song Zhenshuo, Zhang Yinggen, et al. Effect of tea leaf spreading process on white tea quality during withering process [J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39(07): 266-274.) and reference 2 (Wang Jinjin, Yuan Haibo, Hua Jinjie, et al. Effect of full-fire process parameters on hot air drying characteristics and quality of Gongfu black tea [J]. Transactions of the Chinese Society of Agricultural Engineering, 2020, 36(10): 287-296.) respectively optimized the process parameters of the tea leaf spreading and drying units in the tea production line; reference 3 (Li Bing, Li Weining, Bai Xuanbing, et al. Parameter optimization and experimental study of tea rolling machine based on EDEM [J]. Tea Science, 2020, 40(03): 375-385) optimized the rolling speed, rib height and rolling plate inclination of the rolling equipment. The above-mentioned optimized process parameters or equipment structure have a certain guiding role in actual production, but tea production is a complex production process, and optimizing the process parameters of each unit separately cannot guarantee the quality of the finished tea. Summary of the invention

[0005] In order to solve the shortcomings of the prior art and achieve the purpose of obtaining the optimal combination of multi-unit process parameters to improve the quality of the finished tea produced, the present invention adopts the following technical solutions:

[0006] The multi-objective process parameter optimization method of tea production line based on improved particle swarm includes the following steps:

[0007] Step 1: Obtain production data sets of multiple units of the tea production line, including process parameters of each unit, moisture content of tea leaves before and after processing, and chemical composition of raw tea. After preprocessing, establish a moisture content prediction model for each unit and a chemical composition prediction model for the entire production line;

[0008] The production dataset also includes fresh tea leaf image information, which includes the color and texture characteristics of fresh tea leaves. After the tea leaves are processed, the chemical composition of the raw tea is obtained using ultra-high performance liquid chromatography technology;

[0009] Step 2: Based on the moisture content prediction model and chemical composition prediction model of each unit, the relationship between process parameters and moisture content and chemical composition is obtained. Taking the process parameters of each unit as decision variables, the moisture content and chemical composition of raw tea after processing of each unit as production targets, a collaborative optimization model of tea primary production process parameters is established;

[0010] Step 3: With the process parameters as the particle position and the change of the process parameters as the particle speed, an improved particle swarm algorithm is constructed to perform non-dominated sorting on all particles, select a certain number of particles, and update the speed and position with different adaptive adjustment strategies. The strategy changes with the number of iterations. As the number of iterations increases, the number of particles used for random generation gradually decreases. After a certain number of iterations, all particles are used to find the optimal solution set, and particles are no longer randomly generated. After that, all particles update their positions and speeds according to the global optimal particles. The number of non-dominated solutions in each region in the external archive represents the density of the corresponding region. According to the particle density in the region, a guide particle is selected to guide the update of other particles.

[0011] Step 4: A multi-objective particle swarm optimization algorithm based on adaptive update strategy and angle partitioning is applied to the model to solve the collaborative optimization of tea primary production process parameters. The target moisture content and target chemical composition are taken as the targets, and the process parameters selected in actual production are taken as constraints. The difference between the moisture content and chemical composition and the target is minimized.

[0012] Step 5: After the particle swarm reaches the maximum number of iterations, the moisture content of each unit and the chemical composition of raw tea, as well as the process parameters of each unit, are output as the latest non-dominated solution set.

[0013] Furthermore, the collaborative optimization model of tea primary production process parameters in step 2 obtains the moisture content of tea leaves after the current processing unit through the moisture content prediction model based on the process parameters of the current processing unit, the moisture content of fresh tea leaves before processing in the current processing unit, and the basic state information of tea leaves; and obtains the total moisture content deviation according to the moisture content of tea leaves after each processing unit and the target moisture content;

[0014] Based on the process parameters of each processing unit and the moisture content of fresh tea leaves, the corresponding chemical component content in the tea leaves is obtained through the chemical component prediction model, and the total chemical component deviation is obtained according to the obtained chemical component content and the target content;

[0015] By establishing the relationship between each unit process parameter, total moisture content deviation and total chemical composition deviation, a collaborative optimization model for tea primary production process parameters is constructed:

[0016] .

[0017] Furthermore, the objective function of the moisture content is as follows:

[0018]

[0019] Among them, M0 represents the moisture content of fresh tea leaves, M1, M2, M3, and M4 represent the moisture content when the fixing process parameters are X1, the rolling process parameters are X2, the primary baking process parameters are X3, and the secondary baking process parameters are X4, respectively. t1 、M t2 、M t3 、M t4 Represent the target moisture content of the four processing units, M d Indicates the total moisture content deviation target, X p It represents the basic status information of the tea, α1, α2, α3, and α4 respectively represent the weight coefficients of the moisture content deviation of each unit, and f1, f2, f3, and f4 respectively represent the moisture content prediction models of each unit established based on actual production data.

[0020] Furthermore, the objective function of the chemical composition is as follows:

[0021]

[0022] Among them, C GA , C CAF , C ECG represents the chemical composition content of gallic acid GA, caffeine CAF, and epicatechin gallate ECG in tea under each unit decision variable X of tea processing, n represents the number of decision variables, f5 represents the chemical composition prediction model of multiple units based on actual production data, C TGA , C TCAF , C TECG Indicates the target content of three chemical components, C d It represents the total deviation of chemical components, β1, β2, and β3 represent the weight coefficients of these three components, and the weights are determined by the importance of each chemical component.

[0023] Furthermore, the processing unit includes fixing, rolling, primary and secondary drying, and its process parameters include a set of decision variables related to processing temperature and processing speed; the fixing unit decision variable X1={t 11 ,s 11 ,s 12 ,s 13}, t 11 Indicates the fixing temperature, s 11 Indicates the speed of the killing drum, s 12 Indicates the feeding speed, s 13 represents the dehumidification speed; the decision variable of the drying unit X2={t 21 ,s 21 ,s 22}, t 21 Indicates hot air temperature, s 21 Indicates the roller speed, s 22 represents the feeding speed; the initial drying unit decision variable X3={t 31 ,s 31 ,s 32}, t 31 Indicates hot air temperature, s 31 Indicates the conveying speed 1, s 32 represents the conveying speed 2; the decision variable of the re-drying unit X4={t 41 ,s 41}, t 41 Indicates the hot air temperature of the re-drying, s 41 Indicates the conveying speed of re-drying.

[0024] Furthermore, the constraints of the process parameters in step 4 are as follows:

[0025] 260≤t 11 ≤300

[0026] 20.5≤s 11 ≤21

[0027] 8≤s 12 ≤15

[0028] 60≤s 13 ≤100

[0029] 20≤t 21 ≤100

[0030] 22≤s 21 ≤27

[0031] 14≤s 22 ≤27

[0032] 100≤t 31 ≤130

[0033] 1200≤s 31 ≤1500

[0034] 2.8≤s 32 ≤4.2

[0035] 9.5≤t 41 ≤130

[0036] 900≤s 41 ≤1300.

[0037] Furthermore, in step 3, the speed and position update formula of the particle swarm algorithm is as follows:

[0038]

[0039] Among them, t represents the current iteration number, and the position of the i-th particle is expressed as , corresponding to process parameters X1, X2, X3, X4, the speed is , corresponding to the changes in process parameters in production, represents the optimal position of each particle in the historical search, represents the optimal position of the entire population, that is, the optimal process parameters, ω represents the inertia weight, c1 and c2 represent the personal cognitive weight and social weight respectively, and r1 and r2 represent two random numbers in the interval [0,1].

[0040] Furthermore, in order to speed up the search ability of particles with poor quality, by increasing the random numbers r1 and r2, the randomness of the particles is increased, and the exploration of the solution space is intensified. As the number of iterations increases, the increased randomness becomes 0, which is the same as the randomness of other particles. The calculation formula for the level threshold gp and the value rp of the increased randomness of the particles is as follows:

[0041] .

[0042] Furthermore, the guided particle selection method in step 3 is to introduce the number of divisions of the dynamically adjusted angle region. As the number of iterations increases, the number of divisions of the solution space is gradually increased, and the rate of increase of the number of divisions is also increasing. When the number of iterations is close to the end, the number of regional divisions reaches the upper limit to ensure that the solution space is fully covered. i represents the number of iterations, and max represents the maximum number of divisions. The number of regional divisions for each iteration is calculated as follows:

[0043] .

[0044] Furthermore, step 4 comprises the following steps:

[0045] Step 4.1: Set the algorithm parameters of ARA-MOPSO, including the number of initialized populations, the maximum number of iterations, the level threshold, and the particle randomness increase formula parameters, and initialize the entire population, input the moisture content target and chemical composition target value after tea processing, and the state of fresh tea leaves;

[0046] Step 4.2: Calculate multiple moisture content and chemical composition target values ​​in each population to obtain the optimal Pareto solution set;

[0047] Step 4.3: According to the current number of iterations, determine the number of regions in the solution space, and according to the particle density in each region, randomly select a particle from the region with the lowest density as the optimal particle;

[0048] Step 4.4: Perform non-dominated sorting on the particles in the entire population, and select particles with high non-dominated levels according to the non-dominated level of each particle, and use the strategy in step 3 to update the particle speed and position, and update the Pareto solution set of moisture content and chemical composition;

[0049] Step 4.5: Determine whether the Pareto solution set has reached capacity. If it has, delete the particles in the high-density area and determine whether the maximum number of iterations has been reached. If not, continue to execute 4.3 and 4.4. Otherwise, output the Pareto optimal solution set.

[0050] The advantages and beneficial effects of the present invention are:

[0051] The present invention collects data such as multi-unit process parameters, moisture content, image information and chemical composition of a tea production line, establishes the relationship between each unit process parameter and the moisture content and between multiple unit process parameters and the chemical composition, and on this basis, establishes a collaborative optimization model of process parameters for primary tea production; at the same time, based on the improved particle swarm, a multi-objective particle swarm optimization algorithm based on adaptive adjustment strategy and angle division is proposed, different methods are selected to update particle speed and position according to the non-dominated level of the particle, and the solution space is divided into multiple regions according to the angle based on the number of iterations, which are used to select guiding particles and update the solution set, so that the particle swarm not only ensures the diversity of solutions during the iteration process, but also ensures the convergence and uniformity of solutions; the collaborative optimization model of process parameters for primary tea production is solved by the multi-objective particle swarm optimization algorithm based on adaptive adjustment strategy and angle division, and the actual production is guided theoretically. Through process optimization, a set of optimal solution sets are obtained, so that tea production organizations can select more suitable production goals and process parameters according to actual production needs, and the requirements of producing higher quality tea can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 It is a flow chart of a multi-objective process parameter optimization method for a tea production line based on an improved particle swarm in an embodiment of the present invention.

[0054] Figure 2a Schematic diagram of the spatial region result of angle partitioning solution in an embodiment of the present invention (divided into 2 (i=6) regions).

[0055] Figure 2b Schematic diagram of the spatial region result of angle partitioning solution in an embodiment of the present invention (divided into 12 (i=25) regions).

[0056] Figure 2c 1 is a schematic diagram of the results of angle partitioning the solution space region in an embodiment of the present invention (divided into 41 (i=50) regions).

[0057] Figure 2d 1 is a schematic diagram of the results of angle partitioning the solution space region in an embodiment of the present invention (divided into 91 (i=80) regions).

[0058] Figure 3 It is a flow chart of the optimization method of the improved particle swarm algorithm in an embodiment of the present invention.

[0059] Figure 4 It is a schematic diagram of the optimization result for a tea production line in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0061] like Figure 1As shown in the figure, a multi-objective process parameter optimization method for a tea production line based on an improved particle swarm is proposed. This method uses a machine learning model to establish the relationship between production objectives and process parameters. By improving the particle swarm update strategy of the multi-objective particle swarm algorithm and the selection of guided particles (ARA-MOPSO), a multi-objective particle swarm optimization algorithm based on adaptive adjustment strategy and angle division is proposed. The process parameters of each unit of the tea production line are used as constraints to coordinately optimize multiple unit process parameters to obtain the optimal combination of multi-unit process parameters, thereby improving the quality of the finished tea and solving the technical problem that the existing tea primary processing process relies too much on personal experience and cannot produce the best quality tea. The specific steps include the following:

[0062] Step 1: Obtain the production data sets of multiple units of the tea production line, and establish a moisture content prediction model for each unit and a chemical composition prediction model for the entire production line after preprocessing.

[0063] The data set in the tea production line includes the process parameters of each unit, image information of fresh tea leaves, moisture content of each unit before and after processing, and chemical composition of raw tea. The processing units for collecting data include withering, rolling, primary baking, and re-baking, which are mainly composed of the feed rate and processing temperature of each unit. The image information includes 9 color features and 6 texture features. The color features include RGB, LAB, and HSV. The texture features are extracted through the gray level co-occurrence matrix, which are mean, standard deviation, smoothness, third-order moment, consistency, and entropy.

[0064] In the embodiment of the present invention, before the fresh tea leaves are processed, 20g of fresh tea leaves are uniformly selected and placed under a Hikvision industrial camera to collect images; when the tea leaves start to be processed, the tea production line control software is started to collect the production process parameters of multiple units, and the process parameter regulation of each batch of tea leaves is determined by the workers according to the processing experience and the state of the tea leaves; after the fresh tea leaves and the tea leaves are processed by each unit, 3g of tea leaves are taken to measure the moisture content, and the moisture content measuring device is an MA-150C rapid moisture meter, and the device temperature is set to quickly heat to 140°C within 2min, and when the moisture loss within 30s is less than 0.005g, the work is stopped and the moisture content is locked.

[0065] After the tea leaves are processed, ultra-high performance liquid chromatography is used to obtain the chemical components of the raw tea, including eight chemical components: gallic acid, catechin, epicatechin, epigallocatechin, epicatechin gallate, epigallocatechin gallate, theanine, and caffeine. In the embodiment of the present invention, 0.2 (±0.0005) g of tea leaves are weighed to prepare a sample, which is placed in an ultra-high performance liquid chromatography to measure the corresponding chemical component area, and the content of the chemical components in the tea leaves is calculated based on the standard curve. Gallic acid, caffeine, and epicatechin gallate are used to construct a chemical component prediction model. Based on the above data set, support vector machines, multivariate linear regression, and partial least squares regression are used to establish the most suitable relationship model between moisture content and process parameters for each unit, and random forests are used to establish prediction models for multiple chemical components. After data set preprocessing and model hyperparameter optimization, according to R 2 By comparing the performance of each model with RMSE, it was found that the applicable model for the fixing, primary baking and re-baking units is the support vector machine, and the applicable model for the tumble drying unit is the partial least squares regression.

[0066] Step 2: Based on the moisture content prediction model and chemical composition prediction model of each unit, the relationship between process parameters and moisture content and chemical composition is obtained. Taking the process parameters of each unit as decision variables, the moisture content of each unit after processing and the chemical composition of raw tea as production targets, a collaborative optimization model for the primary production process parameters of tea is established.

[0067] The established mathematical model for collaborative optimization of tea primary processing parameters is shown in the following formula:

[0068] (1)

[0069] (2)

[0070] (3)

[0071] Formula (1) represents the moisture content objective function, where M0 represents the moisture content of fresh tea leaves, and M1, M2, M3, and M4 represent the moisture content when the fixing process parameter is X1, the tumble drying process parameter is X2, the primary drying process parameter is X3, and the re-drying process parameter is X4, respectively. t1 、M t2 、M t3 、M t4 represents the target moisture content of the four processing units, M d is the total moisture content deviation target. prepresents the basic status information of the tea leaves. α1, α2, α3, and α4 are weight coefficients of the moisture content deviation of each unit of tea leaves, and their values ​​are 3, 1, 2, and 5, respectively. f1, f2, f3, and f4 represent the moisture content prediction models of multiple units based on actual production data. For the decision variables of the four units, the decision variable of the killing unit X1={t 11 ,s 11 ,s 12 ,s 13}, t 11 is the fixing temperature, s 11 is the speed of the fixing drum, s 12 Indicates the feeding speed, s 13 is the dehumidification speed; the decision variable of the drying unit X2={t 21 ,s 21 ,s 22}, t 21 Indicates hot air temperature, s 21 is the roller speed, s 22 is the feeding speed; the initial drying unit decision variable X3={t 31 ,s 31 ,s 32}, t 31 Indicates hot air temperature, s 31 Indicates the conveying speed 1, s 32 represents the conveying speed 2; the decision variable of the re-drying unit X4={t 41 ,s 41}, t 41 Indicates the hot air temperature of re-drying, s 41 Indicates the conveying speed of re-drying.

[0072] Formula (2) represents the chemical composition objective function, where C GA , C CAF , C ECG The chemical composition content of GA, CAF and ECG in tea leaves is C TGA , C TCAF , C TECG Indicates the target content of three chemical components, C d is the total deviation of chemical composition. β1, β2, and β3 represent the weight coefficients of these three components. The weights are determined by the importance of each chemical component and are 10, 1, and 4, respectively.

[0073] Formula (3) represents a multi-objective optimization model based on the objective functions of formula (1) and formula (2), with moisture content and chemical composition as the targets.

[0074] Step 3: Improve the particle swarm algorithm. Perform non-dominated sorting on all particles, select a certain number of particles based on the non-dominated level, and use different strategies to update the speed and position. The strategy will also change with the number of iterations. Select the guide particle based on the particle density in each angle area, and maintain and update the external archive.

[0075] The speed and position of the particle swarm algorithm are updated according to formula (4):

[0076] (4)

[0077] Where the current iteration number is t, and the position of the i-th particle is expressed as , represents the process parameters X1, X2, X3, X4 in step 2, and the speed is , represents the change of process parameters in production, and the optimal position of each particle obtained by historical search , represents the optimal position of the entire population, that is, the optimal process parameters, where ω is the inertia weight, c1 and c2 are the personal cognitive weight and social weight respectively, and r1 and r2 are random numbers in the interval [0,1].

[0078] After the particles are initialized, an adaptive strategy is introduced to retain a part of the particles for continued random generation without affecting the convergence speed of the particles. Particles with high non-dominated levels are selected from the entire population for regeneration, because the moisture content deviation of these particles is M d and chemical composition deviation C d The other particles update their speed and position according to formula (4). As the number of iterations increases, the number of particles used for random generation gradually decreases. After a certain number of iterations, all particles are used to find the optimal solution set, and no more particles are randomly generated. After that, all particles update their positions and speeds according to the global optimal particle. In order to speed up the search ability of particles with poor quality, the parameters r1 and r2 in formula (4) are increased to increase the randomness of the particles and increase the exploration of the solution space. As the number of iterations increases, the increased randomness becomes 0, which is the same as the randomness of other particles. The calculation formula for the level threshold gp and the value rp of the increased particle randomness is shown in formula (5):

[0079] (5)

[0080] Guided particle selection method: Introduce the number of divisions of the dynamic adjustment angle region. As the number of iterations increases, the number of solution space divisions is gradually increased, and the rate of increase in the number of divisions is also increasing. When the number of iterations is close to the end, the number of regional divisions reaches the upper limit to ensure that the solution space is fully covered. With i representing the number of iterations and max representing the maximum number of divisions, the number of regional divisions for each iteration is calculated as shown in formula (6):

[0081] (6)

[0082] The number of non-dominated solutions in each region in the external archive represents the density of the corresponding region. After the region is divided, a non-dominated solution is randomly selected from the low-density region of the archive as the optimal particle to guide the update of other particles. The division result is as follows: Figure 2a to Figure 2d shown.

[0083] To verify the effectiveness of the improved algorithm, ZDT1, ZDT2, ZDT3, and ZDT6 of the ZDT test function are used as optimization problems, the classic MOPSO, NSGA-II, and MOEA / D algorithms are used as optimization algorithms, and the IGD and HV indicators are used as algorithm performance evaluation indicators. Table 1 Results of IGD and HV performance indicators optimized by four algorithms on the ZDT test function problem. The smaller the IGD value and the larger the HV value, the better the algorithm performance. The results show that ARA-MOPSO regenerates solutions with lower non-dominated levels at the beginning, increases the randomness of solutions with lower non-dominated levels later, and selects the optimal particles by angle partitioning, ensuring the convergence and diversity of the Pareto optimal solution. The proposed method has better results for solving multi-objective optimization problems, with more uniform distribution of solutions and wider coverage.

[0084] Table 1 IGD and HV index results of ARA-MOPSO and other algorithms on ZDT test function

[0085]

[0086] Step 4: A multi-objective particle swarm optimization algorithm based on adaptive adjustment strategy and angle division is applied to the model to solve the collaborative optimization of tea primary production process parameters. The target moisture content and target chemical composition are taken as the targets, and the process parameters selected in actual production are taken as constraints. The difference between the moisture content and chemical composition and the target is minimized.

[0087] In the embodiment of the present invention, the process parameters are used as constraints, and the deviation between the moisture content of tea leaves after withering, rolling, primary baking and re-baking and the multiple chemical components of raw tea and the defined target moisture content and chemical components is minimized. The ARA-MOPSO algorithm is used to solve the above model, and the population of ARA-MOPSO is set to 100 and the number of iterations is set to 50. Medium-quality tea leaves are selected for optimization experiment comparison, and the actual process parameters are determined by experienced workers. The moisture contents after withering, rolling, primary baking and re-baking are 60.94%, 51.91%, 12.98% and 3.68% respectively, and the chemical components GA, CAF and ECG are 1.17 mg / g, 38.34 mg / g and 11.05 mg / g respectively. The target moisture contents of withering, rolling, primary baking and re-baking set by the multi-objective optimization model are 59%, 51%, 11% and 3.73% respectively, and the chemical components GA, CAF and ECG targets are 1.3, 40 and 12 g / mg respectively. The constraints of the process parameters are as follows:

[0088] 260≤t 11 ≤300

[0089] 20.5≤s 11 ≤21

[0090] 8≤s 12 ≤15

[0091] 60≤s 13 ≤100

[0092] 20≤t 21 ≤100

[0093] 22≤s 21 ≤27

[0094] 14≤s 22 ≤27

[0095] 100≤t 31 ≤130

[0096] 1200≤s 31 ≤1500

[0097] 2.8≤s 32 ≤4.2

[0098] 9.5≤t 41 ≤130

[0099] 900≤s 41 ≤1300

[0100] The algorithm flow is as follows Figure 3 As shown, the following steps are included:

[0101] Step 4.1: Set the algorithm parameters of ARA-MOPSO, including the number of initialized populations, the maximum number of iterations, the level threshold, and the particle randomness increase formula parameters, and initialize the entire population. Input the target moisture content and chemical composition of tea leaves after processing and the state of fresh tea leaves.

[0102] Step 4.2: Calculate multiple target values ​​(moisture content target and chemical composition target value) in each population and obtain the optimal Pareto solution set.

[0103] Step 4.3: According to the current number of iterations, determine the number of regions in the solution space, and according to the particle density in each region, randomly select a particle from the region with the lowest density as the optimal particle.

[0104] Step 4.4: Perform non-dominated sorting on the particles in the entire population. According to the non-dominated level of each particle, select the particle with a high non-dominated level. Use the strategy in step 3 to update the particle velocity (process parameter change rate) and position (process parameter), and update the Pareto solution set of moisture content and chemical composition.

[0105] Step 4.5: Determine whether the Pareto solution set has reached capacity. If it has, delete the particles in the high-density area and determine whether the maximum number of iterations has been reached. If not, continue to execute 4.3 and 4.4, otherwise output the Pareto optimal solution set.

[0106] Step 5: After the particle swarm reaches the maximum number of iterations, the latest non-dominated solution set (the solution set of each unit's moisture content value and the chemical composition of raw tea) and the process parameter setting values ​​of each unit are output.

[0107] In the embodiment of the present invention, after iterative optimization, the Pareto frontier is obtained as follows: Figure 4 As shown in Table 2, the two targets are moisture content and chemical composition. The moisture content of the withering unit, drum drying, and primary drying unit are shown in Table 2. The moisture content has decreased and is closer to the average value. The moisture content of the re-drying unit is less than 6%, which is more in line with the actual production requirements. From the chemical composition results, it can be concluded that the GA content has increased to a certain extent, the CAF content has remained basically unchanged, and the ECG content has increased significantly. Tables 3 and 4 are the process parameter optimization results of the four units of withering, drum drying, primary drying and re-drying. In the actual production process, different schemes can be selected according to needs.

[0108] Table 2 Multi-objective optimization results of moisture content and chemical composition of each unit

[0109]

[0110] Table 3 Optimization results of process parameters for fixing and rolling drying units

[0111]

[0112] Table 4 Optimization results of process parameters for primary drying and secondary drying units

[0113]

[0114] Through the study of tea production process optimization methods, tea production units can optimize the process parameters of comprehensive optimization indicators based on this method when optimizing the quality of finished tea products, which is of great significance to improving the efficiency of tea production.

[0115] The main idea of ​​the above improvement is: in the process of searching for the optimal solution by the multi-objective particle swarm, in the initialization stage, the positions and velocities of all particles in the population need to be randomly initialized. In the subsequent iteration process, the position and velocity of the particles will be updated according to the selected optimal particles to achieve a better search for potential solutions in the solution space. However, since the initial velocity and position of the particles are random, the extensiveness and uniformity of the initialization have a certain impact on the subsequent exploration of the particles, which may cause the particle swarm to fall into the local optimum and fail to obtain the global optimal solution, affecting the convergence speed. Because the particle initialization is concentrated in a certain local area, ignoring the exploration of other areas in the solution space, it has certain limitations. In the process of particle swarm solution, selecting a part of the particles to regenerate or increase the randomness can achieve full exploration of the solution space and accelerate the convergence speed; the selection of guide particles affects the overall movement direction of the particle swarm. By dividing the solution space according to angles and selecting guide particles according to the particle density in each area, the distribution of the solution is more uniform, and a better non-dominated solution set is obtained.

[0116] Of course, the above description is only a specific embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. All equivalent changes or modifications made according to the structure, characteristics and principles described in the patent application scope of the present invention should be included in the patent application scope of the present invention.

[0117] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A multi-objective process parameter optimization method for a tea production line based on an improved particle swarm, characterized in that The steps include: Step 1: Obtain production data sets of multiple units of the tea production line, including process parameters of each unit, moisture content of tea leaves before and after processing, and chemical composition of raw tea, and establish a moisture content prediction model for each unit and a chemical composition prediction model for the entire production line; Step 2: Based on the moisture content prediction model and chemical composition prediction model of each unit, the relationship between process parameters and moisture content and chemical composition is obtained. Taking the process parameters of each unit as decision variables, the moisture content and chemical composition of raw tea after processing of each unit as production targets, a collaborative optimization model of tea primary production process parameters is established; The collaborative optimization model of tea primary production process parameters is based on the process parameters of the current processing unit, the moisture content of the fresh tea leaves before the current processing unit, and the basic state information of the tea leaves. The moisture content of the tea leaves after the current processing unit is obtained through the moisture content prediction model; the total moisture content deviation is obtained according to the moisture content of the tea leaves after each processing unit and the target moisture content; Based on the process parameters of each processing unit and the moisture content of fresh tea leaves, the corresponding chemical component content in the tea leaves is obtained through the chemical component prediction model, and the total chemical component deviation is obtained according to the obtained chemical component content and the target content; By establishing the relationship between each unit process parameter, total moisture content deviation and total chemical composition deviation, a collaborative optimization model for tea primary production process parameters was constructed. The objective function of the moisture content is as follows: Among them, M0 represents the moisture content of fresh tea leaves, M1, M2, M3, and M4 represent the moisture content when the fixing process parameters are X1, the rolling process parameters are X2, the primary baking process parameters are X3, and the secondary baking process parameters are X4, respectively. t1 、M t2 、M t3 、M t4 Represent the target moisture content of the four processing units, M d Indicates the total moisture content deviation target, X p represents the basic status information of the tea, α1, α2, α3, and α4 represent the weight coefficients of the moisture content deviation of each unit, and f1, f2, f3, and f4 represent the moisture content prediction models of each unit established based on actual production data; The objective function of the chemical composition is as follows: Among them, C GA , C CAF , C ECG represents the chemical composition content of gallic acid GA, caffeine CAF, and epicatechin gallate ECG in tea under each unit decision variable X of tea processing, n represents the number of decision variables, f5 represents the chemical composition prediction model of multiple units based on actual production data, C TGA , C TCAF , C TECG Indicates the target content of three chemical components, C d It represents the total deviation of chemical components, β1, β2, and β3 represent the weight coefficients of these three components, and the weights are determined by the importance of each chemical component; The processing unit includes fixing, rolling, primary and secondary drying, and its process parameters include a group of decision variables related to processing temperature and processing speed; the fixing unit decision variable X1 = {t 11 ,s 11 ,s 12 ,s 13 }, t 11 Indicates the fixing temperature, s 11 Indicates the speed of the killing drum, s 12 Indicates the feeding speed, s 13 represents the dehumidification speed; the decision variable of the drying unit X2 = {t 21 ,s 21 ,s 22 }, t 21 Indicates hot air temperature, s 21 Indicates the roller speed, s 22 Indicates the feeding speed; the initial drying unit decision variable X3 = {t 31 ,s 31 ,s 32 }, t 31 Indicates hot air temperature, s 31 Indicates the conveying speed 1, s 32 represents the conveying speed 2; the decision variable of the re-drying unit X4 = {t 41 ,s 41 }, t 41 Indicates the hot air temperature of the re-drying, s 41 Indicates the conveying speed of re-drying; Step 3: With the process parameters as the particle position and the change of the process parameters as the particle speed, an improved particle swarm algorithm is constructed to perform non-dominated sorting on all particles, select a certain number of particles, and update the speed and position with different adaptive adjustment strategies. The strategy changes with the number of iterations. The number of non-dominated solutions in each region in the external archive represents the density of the corresponding region. According to the particle density in the region, a guide particle is selected to guide the update of other particles. The speed and position update formula of the particle swarm algorithm is as follows: Among them, t represents the current iteration number, and the position of the i-th particle is expressed as Corresponding process parameters, the speed is Corresponding to the changes in process parameters in production, represents the optimal position of each particle in the historical search, represents the optimal position of the entire population, that is, the optimal process parameters, ω represents the inertia weight, c1 and c2 represent the personal cognitive weight and social weight respectively, r1 and r2 represent two random numbers in the interval [0,1] respectively; By increasing the random numbers r1 and r2, the randomness of the particles is increased, and the exploration of the solution space is intensified. As the number of iterations increases, the increased randomness becomes 0, which is the same as the randomness of other particles. The calculation formula for the level threshold gp and the value rp of the increased randomness of the particles is as follows: The guided particle selection method is to introduce the number of divisions of the dynamic adjustment angle area. As the number of iterations increases, the number of solution space divisions is gradually increased, and the rate of increase in the number of divisions is also increasing. When the number of iterations is close to the end, the number of regional divisions reaches the upper limit. i' represents the number of iterations, and max represents the maximum number of divisions. The number of regional divisions for each iteration is calculated as follows: (1-cos(90*i' / max / 0.8))*90+1, Step 4: Based on the adaptive update strategy and multi-objective particle swarm optimization algorithm, the model is applied to solve the collaborative optimization of tea primary production process parameters. The target moisture content and target chemical composition are taken as the goals, and the process parameters selected in actual production are taken as constraints. The difference between the moisture content and chemical composition and the target is minimized. Step 5: After the particle swarm reaches the maximum number of iterations, the moisture content of each unit and the chemical composition of raw tea, as well as the process parameters of each unit, are output as the latest non-dominated solution set.

2. The multi-objective process parameter optimization method for a tea production line based on an improved particle swarm according to claim 1, characterized in that: The constraints of the process parameters in step 4 are as follows: 260≤t 11 ≤300, 5≤s 11 ≤21, 8≤s 12 ≤15, 60≤s 13 ≤100, 20≤t 21 ≤100, 22≤s 21 ≤27, 14≤s 22 ≤27, 100≤t 31 ≤130, 1200≤s 31 ≤1500, 8≤s 32 ≤4.2, 5≤t 41 ≤130, 900≤s 41 ≤1300。 3. The multi-objective process parameter optimization method for a tea production line based on an improved particle swarm according to claim 1, characterized in that: The step 4 comprises the following steps: Step 4.1: Set the algorithm parameters, input the target moisture content and chemical composition of tea leaves after processing, and the state of fresh tea leaves; Step 4.2: Calculate multiple moisture content and chemical composition target values ​​in each population to obtain the optimal Pareto solution set; Step 4.3: According to the current number of iterations, determine the number of regions in the solution space, and according to the particle density in each region, randomly select a particle from the region with the lowest density as the optimal particle; Step 4.4: Perform non-dominated sorting on the particles in the entire population, and select particles with high non-dominated levels according to the non-dominated level of each particle, and use the strategy in step 3 to update the particle speed and position, and update the Pareto solution set of moisture content and chemical composition; Step 4.5: Determine whether the Pareto solution set has reached capacity. If it has, delete the particles in the high-density area and determine whether the maximum number of iterations has been reached. If not, continue to execute 4.3 and 4.

4. Otherwise, output the Pareto optimal solution set.