A method for optimizing sizing process parameters based on particle swarm optimization

Abstract

The invention relates to the technical field of data processing, and in particular to a method for optimizing glue process parameters based on a particle swarm algorithm, comprising: using a particle swarm algorithm to process data of various process parameters collected during a glue application process, determining a particle update preference according to a change in glue application intensity before and after particle update, and a directional consistency between a particle position after update and a global optimal position, determining a particle position preference according to a distance between a particle position before and after update and a global optimal position, determining a particle speed adjustment factor in combination with the particle update preference and the particle position preference, using the speed adjustment factor to adaptively adjust a particle next update speed, so as to dynamically update the particle position, thereby optimizing the glue process parameters and improving the efficiency of searching for optimal process parameters and the optimization effect.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for optimizing sizing process parameters based on a particle swarm algorithm. Background Art

[0002] In the footwear manufacturing industry, the gluing process is one of the key production links, and its quality directly affects the performance of the final product. The traditional gluing process often relies on manual experience and trial and error, which leads to a series of problems, including low production efficiency, serious material waste, and unstable product quality. Therefore, it is particularly urgent to achieve intelligent optimization of gluing process parameters. Specifically, the gluing process parameters include but are not limited to the amount of glue per unit area, coating speed, drying time, material moisture content, material temperature, ambient wind speed, etc.

[0003] In recent years, with the development of intelligent manufacturing technology, particle swarm algorithm has become an effective method for optimizing various process parameters in industrial production due to its easy implementation, efficient global search capability and strong robustness. Particle swarm algorithm simulates group behavior in nature and establishes a mathematical model to find the optimal combination of process parameters, which can flexibly respond to different production environments and conditions. For example, the existing Chinese patent application document with publication number CN117688744A provides a method, device, equipment, and storage medium for optimizing process parameters of injection molding equipment, including: using historical parameter data of injection molding equipment to train production time consumption model, production energy consumption model, and dimensional error model; generating according to historical process parameter combinations The process parameter combinations are grouped and input into the dimension error model. If the product dimension error values ​​corresponding to each process parameter combination meet the error requirements, they are input into the production time consumption model to select the time-optimized process parameter combination, or they are input into the production energy consumption model to select the energy-optimized process parameter combination; the selected process parameter combination is optimized using the particle swarm optimization algorithm. This patent application document can be applied to industrial Internet production, where the pre-trained model selects the time-optimized or energy-optimized process parameter combination, and the process parameter combination obtained by the particle swarm algorithm is optimized to make the production time or energy consumption lower than the manually set process parameter combination.

[0004] Particle swarm algorithm is an optimization algorithm that finds the optimal solution by simulating the foraging behavior of bird flocks. In this algorithm, each particle represents a possible solution. Each particle has a speed to determine the direction and distance of its next move. Particle swarm algorithm updates the particle position by updating the particle speed. The particle speed update formula is: , in the formula, and Respectively represent The speed of each particle before and after the update, is the learning factor of the historical optimal position, Controls the degree to which the particle moves toward its historical optimal position. is the learning factor of the global optimal position, Controls the degree to which particles move toward the global optimal position. is a random number, For the The current position of a particle, For the The historical optimal position of a particle, is the global optimal position of all particles, Usually a pre-set experience value, and The randomness of particle updates is increased to help particles escape from the historical optimal position and move closer to the global optimal position.

[0005] The particle speed update of the above-mentioned particle swarm algorithm is to update the position of each particle by presetting a fixed learning factor and combining it with a random number. The setting of random numbers can make each particle position update full of randomness, so that the particle can more easily break away from its historical optimal position and move to the global optimal position. However, this random position update often requires a sufficiently large data sample. In the process of optimizing the gluing process parameters, it takes a long period to verify the effect of each particle position update, because each particle position update will get a data combination composed of various process parameters. The various process parameters contained in this data combination need to be tested in the actual production environment to verify whether these parameters have optimized the gluing process. Therefore, this fixed learning factor plus random value makes it difficult for the learning factor to adapt to the changes in various situations that occur during the particle iteration process, thereby increasing the time cost of searching for the optimal parameters, resulting in low search efficiency. This method of updating the position of each particle often consumes a large amount of work, resulting in low efficiency in searching for the optimal position of the particle, affecting the process effect.

[0006] Therefore, a more flexible and adaptive method is urgently needed to optimize the sizing process parameters in order to improve the search efficiency and optimization effect of the optimal process parameters. Summary of the invention

[0007] In order to solve the problem that when optimizing the gluing process parameters by using a particle swarm algorithm, the particle swarm algorithm controls the speed update of particles by combining a fixed learning factor with a random number, which increases the time cost of searching for the optimal parameters and affects the search efficiency and optimization effect of the optimal parameters, the present invention proposes a method for optimizing the gluing process parameters based on a particle swarm algorithm, comprising:

[0008] The particle swarm algorithm is used to process the data of various process parameters collected during the sizing process. The particle swarm algorithm generates multiple particles based on the data of various process parameters, and updates the positions of the particles multiple times through iterations. After each update, the following operations are performed:

[0009] Determine the change in gluing strength according to the difference in gluing strength before and after the particle is updated, determine the directional consistency between the updated position of the particle and the global optimal position according to the angle between the current global optimal position and the historical optimal position of the particle relative to the updated position of the particle, and take the product of the change in gluing strength and the directional consistency as the update preference degree of the particle;

[0010] Determining the particle's position preference according to the distance between the particle's position before and after the update and the global optimal position; the position preference is used to reflect the closeness between the particle's updated position and the global optimal position;

[0011] The particle speed adjustment factor is determined in combination with the particle update preference degree and the particle position preference degree. The speed adjustment factor is used to adaptively adjust the speed of the next particle update to update the particle position, thereby optimizing the sizing process parameters.

[0012] This technical solution processes various process parameter data of the sizing process through a particle swarm algorithm, and can transform a complex multi-dimensional parameter optimization problem into a particle swarm optimization problem. Particles are generated based on these data, and each particle represents a combination of a set of process parameters. Through iterative updates, the optimal process parameter combination is gradually searched. After each update, the direction consistency of the updated position of the particle and the global optimal position is reflected by the angle between the particle's historical optimal position and the global optimal position relative to the particle's updated position. The direction consistency reflects whether the direction of the particle's updated position is closer to the direction of the global optimal position. The particle's update optimization degree is evaluated according to the direction consistency and the change in the sizing intensity before and after the particle update, which can reflect the optimization effect of the particle after each update. Further, the particle's position optimization degree is evaluated by the distance between the particle's position before and after the update and the global optimal position, which is used to reflect the proximity of the particle's updated position to the global optimal position, which can better guide the particle's position to move closer to the global optimal position. The particle speed adjustment factor is determined by comprehensively considering the update optimization degree and the position optimization degree. The adjustment factor is used to dynamically adjust the particle's speed, which can avoid premature convergence or blind exploration, and can more accurately adjust the particle's speed in a targeted manner. The particle's position is dynamically updated, which improves the efficiency of searching for the optimal process parameters and the optimization effect.

[0013] Preferably, the calculation formula for the particle position preference is:

[0014]

[0015] In the formula, For the After the update The position preference of each particle, and Respectively The particle in The distance between the position before and after the update and the global optimal position, is the natural exponential function, is the total number of particles.

[0016] This technical solution quantifies the preference of particle positions by comparing the average distance of particles from the global optimal position with the individual distance. The closer the particles are to the global optimal position, the higher the preference, thereby effectively guiding the positions of particles closer to the global optimal position.

[0017] Preferably, the variation of the sizing intensity , For the After the update The change in sizing intensity of each particle, and Respectively The particle in The glue strength before and after the update, For all particles in The mean of the absolute values ​​of the differences in sizing strength before and after the update.

[0018] Preferably, the directional consistency , It is The particle in The updated position is consistent with the direction of the global optimal position. It is The particle in The historical optimal position and the global optimal position after the update are relative to the The particle in The angle of the updated position, is the cosine function.

[0019] Preferably, the method of determining the particle speed adjustment factor by combining the particle update preference and the particle position preference is: taking the inverse of the product of the particle update preference and the particle position preference as the particle speed adjustment factor.

[0020] Preferably, using the speed adjustment factor to adaptively adjust the speed of the next particle update includes:

[0021] The particle speed update formula is adjusted using the particle speed adjustment factor to obtain the particle speed after the next update. The adjusted particle speed update formula is:

[0022]

[0023] in, and Respectively Particle No. sequence The speed after the update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, For the The particle in The best historical position after the update. For the The global optimal position after the update is For the The particle in The updated position, For the After the update The adjustment factor for each particle is Greater than 1.

[0024] This technical solution dynamically adjusts the learning factor of the global optimal position by introducing an adjustment factor. This solution can adaptively change the particle speed during the particle update process, so that the particle moves to the global optimal position more efficiently. The application of the adjustment factor makes the particle speed update more targeted. The particles with more increased gluing intensity and better directional consistency are more inclined to the global optimal position, thereby accelerating the convergence speed, reducing the ineffective exploration of particles in non-optimal directions, and improving the search efficiency and optimization effect.

[0025] Preferably, the gluing strength of the particles before and after the updating is the gluing strength obtained in actual testing based on the values ​​of various process parameters corresponding to the positions of the particles before and after the updating.

[0026] Preferably, the method for determining the angle is: obtaining a vector from the updated position of each particle to the historical optimal position of the particle, and a vector from the updated position of the particle to the global optimal position, and taking the angle between the two vectors as the angle between the global optimal position and the historical optimal position of the particle relative to the updated position of the particle.

[0027] This technical solution can intuitively evaluate whether the updated position of the particle is closer to the global optimal position by calculating the vector angle between the particle's updated position and its historical optimal position and the global optimal position, thereby better guiding the next movement direction of the particle. The smaller the angle, the better the directional consistency between the particle's updated position and the global optimal position, and the higher the degree of preference. This method can effectively screen out particles that are closer to the global optimal position.

[0028] Preferably, the method for determining the adjustment factor of a particle according to the particle update preference and the particle position preference is: taking the inverse of the product of the particle update preference and the particle position preference as the particle adjustment factor.

[0029] Preferably, the method of adaptively adjusting the speed of the next particle update by using the adjustment factor further includes: when the particle is updated for the first time, the particle speed update formula used is:

[0030]

[0031] in, For the The initial velocity of a particle, For the The velocity of a particle after the first update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, and All are preset values. For the The initial position of a particle, For the The historical optimal position of a particle, and same, is the global optimal position, Is a random number.

[0032] The present invention has the following effects:

[0033] The present invention determines the update preference degree of particles by combining the difference in gluing intensity and the consistency of direction, and can more accurately evaluate the optimization effect of particles. The distance between the particles and the global optimal position before and after the update is used to determine the position preference degree, which can effectively reduce the influence of particles far away from the optimal solution and improve the search efficiency and optimization accuracy. The adjustment factor is determined by comprehensively considering the update preference degree and the position preference degree, and the speed of the particles is adaptively adjusted to dynamically update the position of the particles, accelerate the convergence process, improve the search efficiency of the optimal process parameters in the gluing process, and improve the parameter optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0035] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0037] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0038] Reference Figure 1 The present invention provides a method for optimizing sizing process parameters based on a particle swarm algorithm, comprising steps S1 to S4:

[0039] Step S1: collecting data of multiple process parameters in the gluing process and preprocessing the data of multiple process parameters.

[0040] In the process of producing shoes and boots, glue is required. Gluing can enhance the bonding strength between shoe parts and ensure that shoes and boots will not fall off or crack easily during use. In order to obtain the most suitable production conditions during the gluing process, this step collects data of multiple process parameters during the gluing process, including but not limited to the following items: glue amount per unit area, coating speed, drying time, material moisture content, material temperature, ambient wind speed, etc. These data will affect the strength of the gluing process. By analyzing these data, better gluing process parameters can be obtained to optimize the process flow.

[0041] Record the upper and lower limits of each process parameter data. The upper and lower limits of the data of the process parameters are recorded as and , the total data items of the collected process parameters are recorded as The data of each process parameter is preprocessed to unify the order of magnitude of the data. Specifically, the value of the data of each process parameter is scaled to between.

[0042] In the process of optimizing the gluing process parameters, the particle swarm algorithm performs multiple iterations based on the particle velocity update formula, and eventually obtains a global optimal position shared by all particles. The global optimal position refers to a set of parameter values ​​that are most conducive to optimizing the gluing process. After each iteration, the historical optimal position of each particle represents the best position found by each particle in the search process, that is, the best set of parameter values ​​encountered by the particle in its search history. For example, if a particle obtains a better gluing effect after a certain iteration, its historical optimal position will be updated. After each iteration, the global optimal position represents the best set of parameter values ​​found by all particles in their search history. If a particle finds the best set of parameter values ​​in the entire particle swarm, then this is the global optimal position.

[0043] Next, this step first initializes the particle swarm algorithm for the collected Each process parameter has a range of values. In order to uniformly select the position of the initial particles, the initial particles are uniformly selected within the range of each process parameter. values, It is an experience value, which is set here as ,but N Total process parameters There are a total of 100 combinations of values, each of which corresponds to the position of an initial particle. The position of the initial particles is ensured to be evenly distributed within the range of the values ​​of various process parameters, so as to better cover the entire search space and improve the optimization effect. The learning factor of the historical optimal position of a particle , the learning factor of the global optimal position .

[0044] For example, the data of three process parameters are collected, namely: the amount of glue applied per unit area, the value range is 0 to 100, and 4 values ​​are evenly selected: 0, 25, 50, 75; the coating speed, the value range is 0 to 10, and 4 values ​​are evenly selected: 0, 2.5, 5, 7.5; the drying time, the value range is 0 to 50, and 4 values ​​are evenly selected: 0, 12.5, 25, 37.5. The data of these three process parameters are obtained in total The position of one of the initial particles is , the position of the other initial particle is , and so on, a total of 64 initial particle positions are obtained.

[0045] S2: Determine the update preference of the particle according to the change in the gluing strength before and after the particle is updated, and the directional consistency between the updated position of the particle and the global optimal position.

[0046] The particle swarm algorithm is used to process the data of various process parameters collected during the sizing process. The particle position is updated through multiple iterations of the particle swarm algorithm. Each time the particle is updated, the position of each particle before the update and the values ​​of the process parameters at that position are obtained, as well as the position of each particle after the update and the values ​​of the process parameters at that position.

[0047] For example, for After the update For each particle:

[0048] Get the The position of the particle before the position update, that is, Before the iteration The position of a particle is , No. The updated position of the particle, i.e. After the iteration The position of a particle is , No. The global optimal position after iteration is , No. After the iteration, The historical optimal position of a particle is , It is The number of iterations corresponding to the historical optimal position of a particle, Less than or equal to ;

[0049] Since each location includes Process parameters, then Particles at position No. The value of the process parameter data is , No. Particles at position No. The value of the process parameter data is , the global optimal position No. The value of the process parameter data is .

[0050] This step calculates the update preference of each particle mainly considering the following two points:

[0051] The first is the change in the gluing strength of the particle before and after the update. The more the gluing strength increases, the better the optimization effect of the particle position. Specifically, after each update, the gluing strength of the values ​​of the process parameters of each particle after the update in the actual test and the gluing strength of the values ​​of the process parameters of each particle before the update in the actual test are obtained. The change in the gluing strength of the particle before and after the update is determined based on the difference in the gluing strength before and after the particle is updated.

[0052] For example, for After the update For each particle, The values ​​of various process parameters of the position of each particle before updating in the actual test are , No. The values ​​of various process parameters of the updated position of each particle in the actual test are .

[0053] For example, if Greater than , then it means Update to The optimization effect of the position of each particle is getting better. Less than , then it means The update makes the The optimization effect of the position of the particle is getting worse. If the two are equal, it means that the The update did not affect The optimization effect of the position of each particle.

[0054] Therefore, after each update, the change in sizing strength for each particle is calculated:

[0055]

[0056] In the formula, Indicates After the update The change in sizing intensity of each particle, and Respectively The particle in The glue strength before and after the update, For all particles in The mean of the absolute values ​​of the differences in sizing strength before and after the update.

[0057] In this formula, It reflects the change of the sizing strength brought by the particles before and after the iteration. The larger the value, the stronger the After the iteration The more the gluing strength of a particle increases, the better the position of the particle after the update, and the higher the update preference of the particle. The gluing strength of the particle before the iteration is the gluing strength of the values ​​of the various process parameters of the position of the particle before the iteration in the actual test, and the gluing strength of the particle after the iteration is the gluing strength of the values ​​of the various process parameters of the position of the particle after the update in the actual test.

[0058] It is further explained that after each update, for any particle, if the gluing intensity of the values ​​of various process parameters at the position of the particle after the update in the actual test is higher than the gluing intensity of the values ​​of various process parameters at the position of the particle before the iteration in the actual test, it means that the optimization degree of the particle in this iteration is higher; if the gluing intensity of the values ​​of various process parameters at the position of the particle after the update in the actual test is lower than the gluing intensity of the values ​​of various process parameters at the position of the particle before the iteration in the actual test, it means that the optimization effect of the particle in this iteration is worse.

[0059] The second is the directional consistency between the updated position of the particle and the global optimal position. The directional consistency is determined by the angle between the current global optimal position and the particle's historical optimal position relative to the updated position of the particle. Specifically, the first vector is formed by the updated position of the particle to the particle's historical optimal position, and the second vector is formed by the updated position of the particle to the global optimal position. The directional consistency between the updated position of the particle and the global optimal position is evaluated by the angle between the first vector and the second vector. Because each position is represented by The angle is calculated by first obtaining the vector from the updated position of each particle to the historical optimal position of the particle and the vector from the updated position of the particle to the global optimal position, and taking the angle between these two vectors as the angle between the historical optimal position and the global optimal position relative to the updated position of the particle.

[0060] Therefore, after each update, the directional consistency between the updated position of each particle and the global optimal position is calculated:

[0061]

[0062] In the formula, It is The particle in The updated position is consistent with the direction of the global optimal position. It is The particle in The historical optimal position and the global optimal position after the update are relative to the The particle in The angle of the updated position, is the cosine function.

[0063] In this formula, It reflects the directional consistency between the updated position of the particle and the global optimal position. It is After the iteration, The historical optimal position and global optimal position of the particle are relative to the The cosine value of the angle between the updated positions of the particles. The smaller the angle, the larger the cosine value, indicating that the direction of the particle's historical optimal position is more consistent with the direction of the global optimal position, and the higher the update preference of the particle. When it is 0 degrees, is 1, then is 1, when When it is 180 degrees, for ,at this time is 0, that is, the smaller the angle is, the better the directional consistency between the updated position of the particle and the global optimal position is, and the higher the particle update preference is; the larger the angle is, the worse the directional consistency between the updated position of the particle and the global optimal position is, and the lower the update preference is.

[0064] Finally, the update preference of each particle is calculated based on the change in the particle's gluing intensity before and after the iteration and the directional consistency between the particle's updated position and the global optimal position:

[0065]

[0066] in, For the After the iteration The update optimization degree of each particle, Indicates After the update The change in sizing intensity of each particle, It is The particle in The updated position is consistent with the direction of the global optimal position.

[0067] This formula combines the directional consistency between the updated position of the particle and the global optimal position and the change in gluing intensity brought by the particle before and after iteration to obtain the particle's update preference degree. If the directional consistency between the updated position of the particle and the global optimal position is better, and the gluing intensity of the particle after iteration is higher, it means that the optimization effect of the particle after iteration is better, and the particle's update preference degree is higher.

[0068] S3: Determine the particle's position preference for the next update based on the distance between the particle's position before and after the update and the global optimal position.

[0069] In step S2, the directional consistency between the updated position of the particle and the global optimal position and the change in the gluing strength brought by the particle before and after the iteration have been combined to obtain the particle's update optimization degree. However, this optimization degree can only reflect the optimization effect of the particle after the current iteration. Since the change in the gluing strength is not monotonically increasing, there may be fluctuations, resulting in the gluing strength of the particle after the current iteration. Although it has increased, the gluing strength of the particle after the next iteration may decrease. Even if the consistency between the particle's position update direction and the direction of the global optimal position is good, due to the volatility of the gluing strength, it is still impossible to guarantee that the particle can efficiently tend to the global optimal position.

[0070] Therefore, in order to make the particles tend to the global optimal position more effectively after each iteration, it is necessary to comprehensively consider the change in the distance between the particle and the global optimal position before and after the iteration. Therefore, for any particle, it is hoped that it can tend to the global optimal position after each iteration. Therefore, the position preference of each particle is obtained according to the change in the distance between each particle and the global optimal position before and after the iteration, which is used to reflect the closeness of the particle's position to the global optimal position.

[0071] Specifically, after each iteration, the global optimal position is obtained, and the distance between the position of each particle before iteration and the global optimal position, as well as the distance between the updated position of each particle and the global optimal position are calculated. If the difference between these two distances is large, it means that after this iteration, the updated position of the particle gradually deviates from the global optimal position, and the preferred degree of the updated position of the particle is lower. Conversely, if the difference between the two distances is small, it means that after this iteration, the updated position of the particle is closer to the global optimal position, and the preferred degree of the updated position of the particle is higher.

[0072] For After the update For each particle, first calculate Before the update, The position of the particle and The distance to the global optimal position after the update :

[0073]

[0074] Calculate the After the update, The position of the particle and The distance to the global optimal position after the update :

[0075]

[0076] Next, calculate the The position preference of each particle:

[0077]

[0078] In the formula, Indicates After the update The position preference of each particle, Indicates Updated before The position of each particle is related to ( The distance from the global optimal position after the update, Indicates After the update, The distance between the updated position of each particle and the global optimal position, Represents an exponential function with the natural constant e as base.

[0079] In this formula, The molecule of part After the update, the average distance from the updated position of all particles to the global optimal position, the denominator is The ratio of the distance between the updated position of the particle and the global optimal position. The larger the ratio, the better. After the update, The updated position of the particle is close to the global optimal position, then The position of each particle is more preferred.

[0080] In this formula, Indicates Before and after the update, The greater the difference between the position of the particle and the global optimal position, the greater the difference. The particle passes through The next update is further away from the global optimal position, so it is necessary to reduce the particle's position preference and increase the adjustment factor. The degree of the gap is adjusted in the form of a negative exponential function. The larger the degree of the gap, the lower the position preference.

[0081] The formula calculates the particle position preference It is used to evaluate the degree to which particles are close to the global optimal position during the iteration process. The higher the position optimization degree, the closer the particle is to the global optimal position and the better the optimization effect is, so as to better guide the particle to move to the global optimal position. The higher the position optimization degree, the lower the learning factor of the global optimal position can be appropriately reduced, the large movement of particles can be reduced, and the search accuracy can be improved. The lower the optimization degree, the higher the learning factor of the global optimal position can be appropriately increased to enhance the exploration ability of particles.

[0082] S4: Determine a speed adjustment factor of the particle based on the particle update preference and the particle position preference, and use the speed adjustment factor to adaptively adjust the speed of the next particle update to update the particle position, thereby optimizing the sizing process parameters.

[0083] Specifically, the formula for obtaining the particle adjustment factor according to the particle update preference and position preference is:

[0084]

[0085] In the formula, It is After the update The update optimization degree of each particle, It is After the update The position preference of each particle, It is After the update The speed adjustment factor of each particle is used to adjust the particle speed update formula.

[0086] In this formula, when the update preference and position preference of a particle are both high, it means that the position of the particle is closer to the global optimal position. Under the premise that the learning factor of the historical optimal position of the particle remains unchanged, it is necessary to reduce the learning factor of the global optimal position to prevent the large movement of particles, reduce the large-scale movement of particles, prevent excessive adjustment, and improve the search accuracy; on the contrary, if the update preference and position preference of the particle are lower, it means that the position of the particle is far from the global optimal position, and it is necessary to appropriately increase the learning factor of the global optimal position to enhance the particle's exploration ability so that the particle can move to the global optimal position faster.

[0087] This strategy aims to dynamically adjust the learning factor of the global optimal position through the performance of particles in the iterative process, thereby accelerating the process of finding the global optimal position and improving the efficiency of optimizing the sizing process parameters.

[0088] Furthermore, the speed adjustment factor is used to adaptively adjust the speed of the next particle update to achieve adaptive updating of the particle speed, as follows:

[0089] The particle speed update formula is adjusted using the particle speed adjustment factor to obtain the particle speed after the next update. The adjusted particle speed update formula is:

[0090]

[0091] In the formula, and Respectively Particle No. sequence The speed after the update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, For the The particle in The best historical position after the update. For the The global optimal position after the update is For the The particle in The updated position, For the After the update The adjustment factor for each particle is Greater than 1.

[0092] The particle update speed formula before adjustment is:

[0093]

[0094] In the formula, and Respectively represent The speed of each particle before and after the update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, is a random number, For the The position of a particle, For the The historical optimal position of a particle, at this time, and same, is the global optimal position, that is, the position of the particle with the largest glue strength. At this time, , Usually it is a pre-set empirical value, which has been set in step S2 .

[0095] When the particle swarm algorithm updates the particle position for the first iteration, the particle speed is updated based on the particle update speed formula before adjustment, and the initial performance of the particle in the update process is obtained as the basis for subsequent analysis. Then, in the subsequent iterative process of updating the particle position, the particle speed is dynamically updated through the adjusted particle update speed formula, thereby realizing the update of the particle position.

[0096] It can be observed that the adjusted particle velocity update formula discards the random number in the original formula. , and after each iteration, adjust the factor according to the velocity of each particle Adaptively adjust the learning factor of the global optimal position ,because The degree of movement of particles to the global optimal position is controlled. In the iterative process, the adjustment factor of the particles is calculated by the position change characteristics of the particles. Dynamic adjustment, when When is larger, particles tend to move toward the global optimal position; when When it is smaller, the movement speed of the particles will slow down, reducing the possibility of over-exploration, so that the position of the particles can converge to the global optimal position faster and approach the global optimal position faster.

[0097] This adaptive learning factor approach abandons the random numbers and fixed learning factors in the original algorithm, making the particle speed update more targeted and able to reach the global optimal position faster and more accurately. In other words, it no longer relies on random numbers to adjust the fixed learning factor to determine the direction of particle movement. Instead, it dynamically adjusts the learning factor based on the particle's performance during the iteration process (specifically, the gluing effect corresponding to the particle's updated position after each iteration), allowing the particle to find the global optimal position (the best set of parameter values) more quickly.

[0098] Subsequent gluing is carried out according to the gluing process parameters at the final global optimal position. The adjusted particle swarm algorithm can converge to the optimal gluing process parameters more quickly and accurately, which improves the efficiency of the particle swarm algorithm in finding the optimal gluing process parameters and ensures the optimization effect of the gluing process parameters.

[0099] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0100] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for optimizing sizing process parameters based on particle swarm algorithm, characterized in that: include: The particle swarm algorithm is used to process the data of various process parameters collected during the sizing process. The particle swarm algorithm generates multiple particles based on the data of various process parameters, and updates the positions of the particles multiple times through iterations. After each update, the following operations are performed: Determine the change in gluing strength according to the difference in gluing strength before and after the particle is updated, determine the directional consistency between the updated position of the particle and the global optimal position according to the angle between the current global optimal position and the historical optimal position of the particle relative to the updated position of the particle, and take the product of the change in gluing strength and the directional consistency as the update preference degree of the particle; Determining the particle's position preference according to the distance between the particle's position before and after the update and the global optimal position; the position preference is used to reflect the closeness between the particle's updated position and the global optimal position; The method of determining the particle speed adjustment factor in combination with the particle update preference and the particle position preference is as follows: taking the inverse of the product of the particle update preference and the position preference as the particle speed adjustment factor, and using the speed adjustment factor to adaptively adjust the particle speed for the next update includes: using the particle speed adjustment factor to adjust the particle speed update formula to obtain the particle speed after the next update, and the adjusted particle speed update formula is: in, and Respectively Particle No. sequence The speed after the update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, For the The particle in The best historical position after the update. For the The global optimal position after the update is For the The particle in The updated position, For the After the update The adjustment factor for each particle is Greater than 1 to update the position of the particles and optimize the sizing process parameters.

2. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 1, characterized in that: The calculation formula for the particle position preference is: ； In the formula, For the After the update The position preference of each particle, and Respectively The particle in The distance between the position before and after the update and the global optimal position, is the natural exponential function, is the total number of particles.

3. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 1, characterized in that: The variation of the sizing intensity , For the After the update The change in sizing intensity of each particle, and Respectively The particle in The glue strength before and after the update, For all particles in The mean of the absolute values ​​of the differences in sizing strength before and after the update.

4. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 1, characterized in that: The directional consistency , It is The particle in The updated position is consistent with the direction of the global optimal position. It is The particle in The historical optimal position and the global optimal position after the update are relative to the The particle in The angle of the updated position, is the cosine function.

5. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 3, characterized in that: The gluing strength of the particle before and after the update is the gluing strength obtained in actual testing based on the values ​​of various process parameters corresponding to the positions of the particle before and after the update.

6. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 3, characterized in that: The method for determining the angle is: obtaining a vector from the updated position of each particle to the historical optimal position of the particle, and a vector from the updated position of the particle to the global optimal position, and taking the angle between the two vectors as the angle between the global optimal position and the historical optimal position of the particle relative to the updated position of the particle.

7. The method for optimizing sizing process parameters based on particle swarm algorithm according to claim 1, characterized in that: The method of adaptively adjusting the speed of the next particle update by using the adjustment factor also includes: when the particle is updated for the first time, the particle speed update formula used is: ； in, For the The initial velocity of a particle, For the The velocity of a particle after the first update, It is The learning factor of the historical optimal position of a particle, is the learning factor of the global optimal position, and All are preset values. For the The initial position of a particle, For the The historical optimal position of a particle, and same, is the global optimal position, Is a random number.

Images