Stainless steel single-core tube production method based on improved population algorithm

Abstract

The invention discloses a stainless steel single-core tube production method based on an improved population algorithm, which comprises the following steps: S1, collecting process parameters in a stainless steel single-core tube production process, and constructing a process parameter data set; s2, constructing a multi-objective decision network model, and generating a comprehensive performance score through conflict reconciliation between process objectives; s3, optimizing process parameters by adopting a particle swarm optimization algorithm; s4, performing target performance prediction on each particle position combination by using the comprehensive performance score, and iteratively updating the particle positions until an optimal process parameter combination is obtained; s5, the optimal process parameter combination is applied to each section of actual production of the stainless steel single-core pipe; and S6, collecting actual quality data of the stainless steel single-core tube generated under optimization control, calculating errors and feeding back the errors to the target weight channel. According to the method, the multi-objective decision network model, the particle swarm optimization algorithm and the industrial quality data modeling technology are combined, and stainless steel single-core pipe production based on the improved swarm algorithm is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of stainless steel single-core tube production, and particularly to a production method of stainless steel single-core tube based on an improved swarm algorithm. Background Art

[0002] In the current stainless steel pipe manufacturing industry, especially in the manufacturing process of stainless steel single-core tubes, the setting and adjustment of process parameters greatly affect the finished product quality, manufacturing energy efficiency and process stability of products. The traditional production of stainless steel single-core tubes mainly relies on the experience of operators to control and adjust key process parameters such as drawing speed, lubrication pressure, annealing temperature, coolant temperature and forming tension. Although some advanced manufacturing enterprises have introduced basic digital control equipment and quality inspection systems, the setting of process parameters still shows the characteristics of static setting, manual experience dominance and lagging response. This control method is difficult to cope with complex working conditions such as fluctuations in raw materials of different batches, changes in production rhythm, and fluctuations in equipment status, and is extremely likely to cause uneven wall thickness, surface quality fluctuations, excessive energy consumption or fracture risk during the drawing process, thus directly affecting the product consistency, energy utilization rate and equipment operation stability of the production line.

[0003] At present, research on the optimization of process parameters in the manufacturing process has gradually introduced machine learning and intelligent optimization algorithms, such as artificial neural networks, genetic algorithms, particle swarm optimization, etc. Among them, the particle swarm optimization algorithm (PSO) has been applied to the optimization problems of metal forming processes due to its characteristics of few parameters, fast convergence and easy implementation. However, the conventional particle swarm optimization algorithm still has certain limitations when facing industrial problems with multiple objectives, strong nonlinearity and high-dimensional coupling. Specifically, the particle swarm optimization algorithm is prone to falling into local optima and it is difficult to ensure the global optimization effect; in addition, its fitness function is often designed singly and it is difficult to comprehensively represent the trade-off relationship between multiple process quality objectives; furthermore, traditional algorithms usually lack a real-time feedback mechanism and cannot dynamically adjust the optimization direction according to the quality inspection results in the actual production process, resulting in the lack of adaptive ability of the process control system.

[0004] On the other hand, although some studies have attempted to combine deep learning methods with optimization algorithms, such as using neural networks to predict process quality and assist in optimizing parameters, in the typical multi-process series manufacturing scenario of stainless steel single-core tubes, due to its strong process coupling, obvious mutual influence of quality indicators, and strict control boundary requirements, higher requirements are put forward for the response speed, prediction accuracy, and closed-loop control ability of the optimization system. Most traditional data-driven models stay at the level of single-objective fitting and lack a modeling and reconciliation mechanism for conflicts between multiple quality objectives. At the same time, existing studies often ignore the actual control constraints of process parameter adjustment, such as setting ranges, adjustment granularity, and the correlation between parameters, resulting in the lack of feasibility of the output optimized parameter combinations for deployment and execution in actual equipment.

[0005] In addition, the utilization of quality inspection results in existing technologies mostly focuses on the post-process analysis stage, lacking a real-time feedback mechanism for inspection data and failing to effectively use the inspection error results to correct the target structure and evaluation mechanism inside the optimization model. This "one-way" optimization process shows poor adaptability and robustness when facing the operation of complex industrial systems and is difficult to maintain the optimal control state in a dynamically changing production environment.

[0006] Therefore, in the face of the actual problems of "multiple objectives, multiple variables, strong constraints, and need for closed-loop" in the manufacturing process of stainless steel single-core tubes, a new parameter configuration method combining process cognition, data-driven, and intelligent optimization is urgently needed to achieve intelligent modeling, collaborative optimization, and online adaptive update capabilities of key control parameters in the whole-process technology. This method should be able to extract feature vectors that can express product performance from real-time collected process data, establish the correlation structure between multi-objective quality performances, introduce a dynamic scoring mechanism to evaluate the comprehensive performance of parameter combinations, use advanced swarm intelligence algorithms to globally optimize the parameter space, and based on the feedback results, real-time correct the scoring structure after production execution to achieve a complete closed-loop from process modeling, parameter optimization to equipment control and feedback adjustment.

[0007] Therefore, how to provide a production method for stainless steel single-core tubes based on an improved swarm algorithm is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0008] One purpose of the present invention is to propose a method for producing stainless steel single-core tubes based on an improved swarm algorithm. The present invention combines advanced technologies such as multi-objective decision-making network models, particle swarm optimization algorithms, and industrial quality data modeling. By constructing a scoring structure with adjustable target weights, introducing a dynamic weight adjustment mechanism based on process feedback, and utilizing a swarm intelligence algorithm to perform global optimization in a high-dimensional parameter space, the present invention achieves optimization of the production process of stainless steel single-core tubes under complex constraints. The system uses actual process parameters and multi-dimensional quality indicators as the modeling basis to form a closed-loop optimization control path that integrates scoring drive and iterative feedback. It can adapt to raw material fluctuations, equipment operation differences, and changes in product quality requirements, and dynamically outputs a process parameter combination with execution feasibility and optimal performance, significantly improving the consistency, yield rate, and intelligent level of the production process of stainless steel single-core tube products.

[0009] A method for producing a stainless steel single-core tube based on an improved swarm algorithm according to an embodiment of the present invention includes the following steps:

[0010] S1. Collect process parameters during the production of stainless steel single-core tubes and construct a process parameter data set;

[0011] S2. Based on the process parameter data set, a multi-objective decision network model is constructed, with process parameters as input and quality data as output. An adjustable target weight channel is provided to generate a comprehensive performance score by reconciling conflicts between process targets.

[0012] S3, using particle swarm optimization algorithm to optimize process parameters, expressing particle position as process parameter and particle velocity as process adjustment;

[0013] S4. During the particle swarm iteration process, the target performance of each particle position combination is predicted using the comprehensive performance score, the comprehensive fitness value is calculated, the particle position is updated according to the comprehensive fitness value, the particle speed is adjusted according to the rate of change of the comprehensive fitness value, and the particle position is iteratively updated until the optimal process parameter combination is obtained;

[0014] S5. Apply the optimal process parameter combination to each stage of the actual production of stainless steel single-core tubes;

[0015] S6. Collect the actual quality data of the stainless steel single-core tube generated under the optimization control, compare it with the target performance prediction value, calculate the error and feed it back to the target weight channel, and correct the weight configuration between each target.

[0016] Optionally, the quality data specifically include wall thickness fluctuation rate, surface roughness, unit energy consumption and drawing fracture rate.

[0017] Optionally, the process parameters specifically include drawing speed, lubrication pressure, annealing temperature, coolant temperature and forming tension.

[0018] Optionally, the S2 specifically includes:

[0019] S21. Construct a multi-objective decision network model with a multi-objective output structure. The multi-objective decision network model adopts a multi-layer perceptron structure and is composed of an input layer, multiple hidden layers, and an output layer. The input layer has 5 nodes, corresponding to the input process parameters, and the output layer contains 4 nodes, corresponding to the output quality data.

[0020] S22. Set up a target weight fusion channel at the output layer to build a process target reconciliation mechanism, assigning a corresponding weight coefficient to each output prediction value. The target weight fusion channel supports independent weighting adjustment of each output target during the inference process to generate a comprehensive performance score.

[0021] S23. Calculate the comprehensive performance score and conduct a weighted evaluation of the performance of the process parameter combination on each quality data:

[0022]

[0023] Among them, Ψ comprehensive performance score, δ w Predicted wall thickness fluctuation, R s Predicted surface roughness, E u Predicted unit energy consumption, λ b Predicted pull-out fracture rate, The historical mean of wall thickness fluctuation, Historical surface roughness average, Historical average of unit energy consumption, The historical average of the pull-out fracture rate, ω1, ω2, ω3, ω4 adjustable weight coefficients, very small positive numbers, μ conflict reconciliation penalty coefficient, ω j The weight of the j-th quality data, q j Normalized deviation value of the jth quality data, ω i The weight of the i-th quality data, q i Normalized deviation value of the i-th quality data;

[0024] S24. Dynamically adjust the weight coefficient of each target according to process requirements and correct the relative weight relationship between targets.

[0025] Optionally, the S3 specifically includes:

[0026] S31. Construct a particle swarm search space suitable for optimizing process parameters in the production of stainless steel single-core tubes, and define the current position of each particle as a set of process parameter combinations;

[0027] S32. Set upper and lower bounds for each process parameter dimension. The bounds are determined based on the equipment control capability, historical operating data, and safe process limits. The particle swarm search space consists of a range of process parameter combinations. Each position represents a set of process parameter combinations that are controllable and executable in single-core tube production.

[0028] S33. Set the velocity vector of each particle, which is expressed as a variable process adjustment amount of the corresponding process parameter in the current iteration, and each dimensional velocity component corresponds to the dynamic change amount of the process parameter.

[0029] Optionally, the S4 specifically includes:

[0030] S41. During the particle swarm optimization iteration process, the target performance of each particle position combination is predicted using the comprehensive performance score, and the comprehensive fitness value of each particle position combination is calculated:

[0031]

[0032] Among them, i (t) is the comprehensive fitness score of the i-th particle in the t-th iteration, Ψ is the comprehensive performance score, μ is the conflict reconciliation penalty coefficient, ω j is the weight of the j-th quality data, q j is the normalized deviation value of the j-th quality data, ω i is the weight of the i-th quality data, q i is the normalized deviation value of the i-th quality data, ω k is the original target weight, To normalize the relative error between the predicted value and the reference value and amplify it through an exponential function, j is the index of the target weight, k is the target number index, is the historical mean of the k-th quality indicator, and ε is a very small positive number;

[0033] S43, based on the comprehensive fitness value Ψ of the current particle i (t) Compare with the historical optimal fitness and the global optimal fitness. If the comprehensive fitness value of the current particle is less than the historical individual optimal fitness, then update the individual optimal position to the current particle position; if the comprehensive fitness value of the current particle is less than the global optimal fitness, then update the global optimal position to the current particle position;

[0034] S44. Calculate the rate of change of the comprehensive fitness value ΔΨ i =|Ψ i (t)-Ψ i (t-1)|, the drawing speed, lubrication pressure, annealing temperature, coolant temperature and forming tension in the particle velocity vector are dynamically adjusted according to the change rate, θ is the set fitness change rate threshold, when ΔΨi When it is <θ, the shrinking velocity component is used to enhance the local search ability; when ΔΨ i ≥θ, the magnifying velocity component is used to improve the search jumping ability;

[0035] S45. Iteratively update the particle positions until the optimal process parameter combination is obtained.

[0036] Optionally, the S45 specifically includes:

[0037] S451. According to the updated velocity vector and the current particle position, perform a position update operation. Based on the process parameter combination represented by the current particle, numerically adjust the five process parameters of drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension in sequence. The adjustment amount of each process variable is provided by the corresponding component in the velocity vector to obtain the next-round particle position;

[0038] S452. Repeatedly execute the target performance prediction, comprehensive fitness calculation, position update, and velocity adjustment operations until the iteration round reaches the maximum iteration number T max = 200 times or within 10 consecutive rounds, the global optimal fitness score value of the particle swarm satisfies the change convergence condition Ψ g (t)-Ψ g (t - 1)| < ∈, ∈ = 10 -4 , and output the optimal process parameter combination corresponding to the global particle position with the optimal comprehensive score. Optionally, the S6 specifically includes:

[0039] Optionally, the S5 specifically includes:

[0040] S51. Map the optimal process parameter combination, including drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension, to the five key process sections of the stainless steel single-core tube production line, namely the drawing section, lubrication section, annealing section, cooling section, and tension control section respectively; by establishing parameter interface connections with the control modules of each section, accurately load the corresponding parameter values into each execution system as the standard process setting values for the current production batch, and at the same time complete the setting value writing and status synchronization, so that each section is in the loaded and pending execution state before the task starts. The loading process supports real-time parameter verification and automatic backup;

[0041] S52. After receiving the production task execution instruction, automatically call the set optimal process parameter combination, and use it as the main control benchmark to perform real-time process scheduling and segmented control on the entire production line; each control section executes the closed-loop control strategy according to the standard set value, and dynamically adjusts the execution behavior by collecting the deviation between the current operating state and the target setting, so as to realize real-time parameter tracking and automatic adjustment; the entire production execution process is carried out strictly in accordance with the optimal process parameter combination, and from the input of section-level parameters, the execution of logic control to the feedback correction of sensors, all are under continuous and consistent target control.

[0042] Optionally, the S6 specifically includes:

[0043] S61. During the production process of the stainless steel single-core pipe, continuously collect the actual operation data of each section controlled by the optimal process parameter combination, and combine the index data output by the quality detection module to construct a quality feedback data set corresponding to the current batch. The quality data includes wall thickness fluctuation rate, surface roughness, unit energy consumption, and drawing fracture rate;

[0044] S62. Compare the measured quality indicators with the target performance values predicted by the multi-objective decision network model one by one to construct a quality error vector:

[0045]

[0046] where, Δy k is the absolute error value of the quality data, is the measured value, is the model predicted value, k ∈ 1, 2, 3, 4, corresponding to four types of quality data respectively;

[0047] S63. Input the quality error vector into the adjustable target weight channel to perform feedback correction on the weight configuration of the four quality objectives in the scoring function:

[0048]

[0049] where, ω k is the original target weight, the corrected k-th target weight, β is the error sensitivity adjustment coefficient, Δy k is the absolute error value of the quality data, j is the index of the target weight;

[0050] S64. Input the updated target weight configuration back into the multi-objective scoring function as the basis for calculating the fitness in the next round of particle swarm optimization iteration.

[0051] The beneficial effects of the present invention are:

[0052] The present invention combines a multi-objective decision-making network model, a particle swarm optimization algorithm, and industrial quality data modeling technology, successfully overcoming the limitations of process parameter setting in the existing production of stainless steel single-core pipes. Traditional production methods mostly rely on manual experience to set process parameters and lack sufficient adaptability and closed-loop feedback mechanisms, resulting in large fluctuations in product quality, low production efficiency, and difficulty in coping with unstable factors caused by raw material fluctuations and equipment status changes. In contrast, the present invention can precisely balance the conflicts between various quality objectives by dynamically optimizing the process parameter combination, improving the consistency and quality stability of products.

[0053] By introducing an optimization method based on an improved swarm algorithm, the present invention can achieve global optimization in a complex and multi-dimensional process parameter space, avoiding the deficiency that traditional optimization algorithms are prone to falling into local optimal solutions. In the actual production process, the system adjusts the target weights according to real-time feedback, automatically correcting and optimizing the process parameter combination, thus ensuring the continuous optimization of the production process and high-precision quality control. This adaptive optimization mechanism not only improves the yield and product quality of stainless steel single-core pipes but also significantly reduces energy consumption and production costs, increasing production efficiency.

[0054] In addition, the intelligent control system of the present invention makes the process parameter optimization process more automated and precise, reducing manual intervention and improving the intelligent level of the production line. Through precise process parameter control and multi-objective optimization, the present invention provides an efficient, stable, and well-adaptive solution for the production of stainless steel single-core pipes, meeting the requirements of modern manufacturing for high-quality and high-efficiency production. Description of the Drawings

[0055] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention and do not constitute a limitation to the present invention. In the drawings:

[0056] Figure 1 is a flowchart of a method for producing stainless steel single-core pipes based on an improved swarm algorithm proposed by the present invention;

[0057] Figure 2 is a schematic diagram of a method for producing stainless steel single-core pipes based on an improved swarm algorithm proposed by the present invention;

[0058] Figure 3 is a data flow diagram of a method for producing stainless steel single-core pipes based on an improved swarm algorithm proposed by the present invention. Detailed Embodiments

[0059] Now, the present invention will be further described in detail with reference to the drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0060] Reference Figures 1-3 , a production method of stainless steel single-core pipe based on an improved swarm algorithm, comprising the following steps:

[0061] S1. Collect the process parameters in the production process of the stainless steel single-core pipe and construct a process parameter data set;

[0062] S2. Based on the process parameter data set, construct a multi-objective decision network model, with the process parameters as the input, the quality data as the output, and there is an adjustable target weight channel, and generate a comprehensive performance score through the reconciliation of conflicts between process objectives;

[0063] S3. Use the particle swarm optimization algorithm to optimize the process parameters, represent the particle position as the process parameters, and the particle velocity as the process adjustment amount;

[0064] S4. During the particle swarm iteration process, use the comprehensive performance score to predict the target performance of each group of particle position combinations, calculate the comprehensive fitness value, update the particle position according to the comprehensive fitness value, adjust the particle velocity magnitude according to the change rate of the comprehensive fitness value, and iteratively update the particle position until the optimal process parameter combination is obtained;

[0065] S5. Apply the optimal process parameter combination to each section of the actual production of the stainless steel single-core pipe;

[0066] S6. Collect the actual quality data of the stainless steel single-core pipe generated under optimized control, compare it with the target performance prediction value, calculate the error and feedback it to the target weight channel to correct the weight configuration between each target.

[0067] The present invention combines a multi-objective decision network model with a particle swarm optimization algorithm, and improves the production efficiency and product quality by optimizing the process parameter configuration in the production process of the stainless steel single-core pipe. Based on the process parameter collection and the multi-objective decision model, the system can dynamically adjust the weights of each target, accurately generate a comprehensive performance score, so as to achieve the global optimization of the process parameters. Through the particle swarm optimization algorithm, the relationship between the adjustment amount of the process parameters and the particle position is effectively explored, avoiding the local optimum problem in the traditional method, and ensuring that the optimal process combination is always found during the optimization process. In actual production, the system can collect production data in real time, compare it with the predicted value, calculate the error and feedback it to the target weight channel, and automatically correct the weight configuration between the targets, further improving the stability and adaptability of the production process. This closed-loop feedback mechanism enables the system to continuously optimize under different production batches and raw material conditions, significantly improving the finished product consistency and production automation level of the stainless steel single-core pipe.

[0068] In this embodiment, the quality data specifically includes wall thickness volatility, surface roughness, unit energy consumption, and drawing fracture rate.

[0069] The present invention precisely evaluates various quality indicators in the production process by collecting quality data including wall thickness volatility, surface roughness, unit energy consumption, and drawing fracture rate. By constructing a multi-dimensional quality data space, it ensures that each quality objective is reasonably optimized, and dynamically adjusts the weights of various quality indicators during the optimization process, enhancing the consistency and stability of stainless steel single-core tube products. This method can provide real-time feedback of production data, enhance the adaptive ability of the system, and ensure continuous optimization of production efficiency and product quality.

[0070] In this embodiment, the process parameters specifically include drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension.

[0071] The present invention optimizes the production process of stainless steel single-core tubes by precisely controlling key process parameters such as drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension. Through comprehensive optimization of these process parameters, it ensures that the parameter adjustment in each link can maximize product quality and production efficiency. The system can dynamically adjust various process parameters during actual production, reducing the deviation caused by manual setting, thereby achieving more stable and precise production control and improving the overall quality and production efficiency of stainless steel single-core tubes.

[0072] In this embodiment, S2 specifically includes:

[0073] S21. Construct a multi-objective decision-making network model with a multi-objective output structure. The multi-objective decision-making network model adopts a multi-layer perceptron structure, which consists of an input layer, multiple hidden layers, and an output layer. The number of nodes in the input layer is 5, corresponding to the input process parameters, and the output layer contains 4 nodes, corresponding to the output quality data.

[0074] S22. Set a target weight fusion channel in the output layer, construct a process objective reconciliation mechanism, assign corresponding weight coefficients to each output prediction value. The target weight fusion channel supports independent weighted adjustment of each output target during the inference process to generate a comprehensive performance score.

[0075] S23. Calculate the comprehensive performance score and conduct a weighted evaluation of the performance of the process parameter combination on each quality data:

[0076]

[0077] where, Ψ is the comprehensive performance score, δ w predicted wall thickness volatility, R s predicted surface roughness, E u predicted unit energy consumption, λ b predicted drawing fracture rate, historical average wall thickness volatility, historical average surface roughness, Historical average value of unit energy consumption Historical average value of drawing fracture rate, adjustable weight coefficients ω1, ω2, ω3, ω4, extremely small positive number, μ conflict reconciliation penalty coefficient, ω j Weight of the jth quality data, q j Normalized deviation value of the jth quality data, ω i Weight of the ith quality data, q i Normalized deviation value of the ith quality data;

[0078] S24. Dynamically adjust the weight coefficients of each target according to process requirements to correct the relative weight relationship between targets.

[0079] Through constructing a multi-objective decision network model and combining the multi-layer perceptron structure with the target weight fusion channel, the present invention realizes the dynamic optimization of process parameters in the production process of stainless steel single-core pipes. By independently weighting and adjusting each quality index to generate a comprehensive performance score, it ensures a reasonable balance of each quality target in the optimization process. At the same time, the system can dynamically adjust the target weight coefficients according to process requirements, optimize the relative importance between each target, and improve the adaptive ability of the production process and the consistency of product quality.

[0080] In this embodiment, the S3 specifically includes:

[0081] S31. Construct a particle swarm search space applicable to the optimization of stainless steel single-core pipe production process parameters, and define the current position of each particle as a set of process parameter combinations;

[0082] S32. Set the upper and lower boundary ranges for each process parameter dimension respectively. The boundary values are determined based on equipment control capabilities, historical operation data, and safety process limits. The particle swarm search space is composed of the process parameter combination range, and each position represents a set of adjustable and executable process parameter combinations in single-core pipe production;

[0083] S33. Set the velocity vector of each particle, which is expressed as the process adjustment amount that the corresponding process parameter can change in the current iteration. Each dimensional velocity component corresponds to the dynamic change amount of the process parameter.

[0084] By constructing a particle swarm search space, the present invention precisely defines the process parameter combinations of each particle, and sets parameter boundaries based on equipment control capabilities, historical data, and safety limits to ensure that the adjustment range of process parameters meets the actual production requirements. The velocity vector of each particle represents the change in the process adjustment amount, enabling the optimization process to perform dynamic adjustment in each process parameter dimension. Through this method, the process parameters can be precisely controlled during the optimization process, improving the flexibility and adaptability of the production process, and ensuring the quality and efficiency of stainless steel single-core pipe production.

[0085] In this embodiment, S4 specifically includes:

[0086] S41. During the particle swarm optimization iteration process, use the comprehensive performance score to perform target performance prediction on each group of particle position combinations, and calculate the comprehensive fitness value of each group of particle position combinations:

[0087]

[0088] Among them, Ψ i (t) is the comprehensive fitness score value of the i-th particle in the t-th iteration, Ψ is the comprehensive performance score, μ is the conflict reconciliation penalty coefficient, ω j is the weight of the j-th quality data, q j is the normalized deviation value of the j-th quality data, ω i is the weight of the i-th quality data, q i is the normalized deviation value of the i-th quality data, ω k is the original target weight, is to normalize the relative error between the predicted value and the reference value and amplify it through the exponential function, j is the index of the target weight, k is the target number index, is the historical mean value of the k-th quality index, ε is a very small positive number;

[0089] S43. Compare the comprehensive fitness value Ψ i (t) of the current particle with the historical best fitness and the global best fitness. If the comprehensive fitness value of the current particle is less than the historical individual best fitness, update the individual best position to the current particle position; if the comprehensive fitness value of the current particle is less than the global best fitness, update the global best position to the current particle position;

[0090] S44. Calculate the change rate ΔΨ of the comprehensive fitness value i =|Ψ i (t)-Ψ i (t - 1)|, and dynamically adjust the drawing speed, lubrication pressure, annealing temperature, coolant temperature and forming tension in the particle velocity vector according to the change rate. θ is the set fitness change rate threshold. When ΔΨ i <θ, reduce the velocity component to enhance the local search ability; when ΔΨ i ≥θ, amplify the velocity component to improve the search jump ability;

[0091] S45. Iteratively update the particle positions until the optimal process parameter combination is obtained.

[0092] In this embodiment, S45 specifically includes:

[0093] S451. Perform a position update operation based on the updated velocity vector and the current particle position. Based on the process parameter combination represented by the current particle, numerically adjust the five process parameters of drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension in sequence. The adjustment amount of each process variable is provided by the corresponding component in the velocity vector to obtain the next-round particle position.

[0094] S452. Repeat the operations of target performance prediction, comprehensive fitness calculation, position update, and velocity adjustment until the iteration round reaches the maximum iteration number T max = 200 times or within 10 consecutive rounds, the global optimal fitness score value of the particle swarm satisfies the change convergence condition Ψ g (t) - Ψ g (t - 1)| < ∈, ∈ = 10 -4 , and output the optimal process parameter combination corresponding to the global particle position with the optimal comprehensive score.

[0095] In the present invention, through the particle swarm optimization algorithm combined with the comprehensive performance score, the fitness value of each group of particle position combinations is accurately evaluated, and the target weight and quality index deviation are dynamically adjusted. By introducing a conflict reconciliation penalty coefficient and normalized error processing, the global search and local exploration capabilities of the process parameters are effectively optimized to ensure that the process parameter combination always remains optimal during the optimization process. By comparing the current particle fitness with the historical optimal fitness, the system can automatically update the optimal position and adjust the particle velocity, improving the jumping ability or local accuracy of the search, and finally determining the optimal process parameter combination. This method greatly improves the self-adaptability, flexibility, and optimization efficiency of the production process, and effectively ensures the production quality and consistency of the stainless steel single-core tube.

[0096] In this embodiment, the S5 specifically includes:

[0097] S51. Map the optimal process parameter combination, including drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension, to the five key process sections of the stainless steel single-core tube production line, namely the drawing section, lubrication section, annealing section, cooling section, and tension control section respectively; by establishing parameter interface connections with the control modules of each section, accurately load the corresponding parameter values into each execution system as the standard process setting values for the current production batch, and at the same time complete the writing of the setting values and status synchronization, so that each section is in the loaded and waiting-to-execute state before the task starts. The loading process supports real-time parameter verification and automatic backup.

[0098] S52. After receiving the production task execution instruction, automatically call the set optimal process parameter combination, and use it as the main control reference to perform real-time process scheduling and section control on the entire production line; each control section executes the closed-loop control strategy according to the standard set value, and dynamically adjusts the execution behavior by collecting the deviation between the current operating state and the target setting, so as to achieve real-time parameter tracking and automatic adjustment; the entire production execution process is carried out strictly in accordance with the optimal process parameter combination, and from the input of section-level parameters, the execution of logic control to the feedback correction of sensors, all are under continuous and consistent target control.

[0099] In this embodiment, the S6 specifically includes:

[0100] S61. During the production process of the stainless steel single-core pipe, continuously collect the actual operation data of each section controlled by the optimal process parameter combination, and combine the index data output by the quality detection module to construct a quality feedback data set corresponding to the current batch. The quality data includes wall thickness volatility, surface roughness, unit energy consumption, and drawing fracture rate.

[0101] S62. Compare the measured quality indicators with the target performance values predicted by the multi-objective decision-making network model one by one to construct a quality error vector:

[0102]

[0103] Among them, Δy k is the absolute error value of the quality data, is the measured value, is the model predicted value, k ∈ 1, 2, 3, 4, corresponding to four types of quality data respectively;

[0104] S63. Input the quality error vector into the adjustable target weight channel to perform feedback correction on the weight configuration of the four quality targets in the scoring function:

[0105]

[0106] Among them, ω k is the original target weight, the corrected k-th target weight, β is the error sensitivity adjustment coefficient, Δy k is the absolute error value of the quality data, and j is the index of the target weight;

[0107] S64. Input the updated target weight configuration back into the multi-objective scoring function as the fitness calculation basis for the next round of particle swarm optimization iteration.

[0108] The present invention ensures the precise matching of the control modules and process parameters in each section by accurately loading the optimal combination of process parameters into five key process sections of the stainless steel single-core tube production line. Before the task starts, the system automatically synchronizes the set values, conducts real-time verification, and performs automatic backup, thereby ensuring that each section is in the loaded and pending execution state at startup. Once the production task instruction is received, the system automatically calls the optimal parameter combination and uses it as the main control reference for the production line for real-time scheduling and control. Each section executes a closed-loop control strategy based on the set value and dynamically adjusts the execution behavior through the deviation amount to ensure that the entire production process strictly follows the optimal combination of process parameters, improving the adaptive capacity, accuracy, and stability of the production line, and significantly enhancing the production efficiency and product quality consistency.

[0109] Example 1:

[0110] To verify the feasibility of the present invention in implementation, the present invention is applied to a certain stainless steel single-core tube production enterprise, XYZ Stainless Steel Factory located in Wuxi City, Jiangsu Province. This factory mainly produces high-precision stainless steel single-core tubes, which are widely used in industries such as aerospace, automotive manufacturing, and petrochemical industry. During the daily production process, the enterprise faces the following problems: due to the quality fluctuations of raw materials in different batches, different equipment states, and the instability of the process parameters set manually, problems such as uneven wall thickness, unqualified surface roughness, and excessive energy consumption often occur during the production process. These problems not only affect the product quality but also result in higher production costs and longer production cycles.

[0111] To solve these problems, the enterprise decides to apply the production process optimization method based on the improved swarm algorithm in the present invention. By collecting the process parameters on the production line in real time, including drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension, etc., data modeling is carried out through a multi-objective decision-making network model, and the process parameters are optimized through the particle swarm optimization algorithm, and finally the optimal combination of process parameters suitable for production is obtained.

[0112] In the actual application process, the multi-objective decision-making network model of the present invention analyzes the historical data and real-time production data of each process parameter, calculates the parameter combination suitable for the current production conditions, and solves the conflicts between different quality objectives in the production process by dynamically adjusting the weight coefficients. Within two months after the system deployment, the production efficiency and product quality have been significantly improved. Specifically, the production efficiency has increased, the product quality has been enhanced, and the energy consumption has been reduced. The increase in production efficiency means that through optimizing the process parameters, the production cycle has been shortened from the original 24 hours per production line to 19 hours. The downtime of the production line has been reduced by 15%, and the operating efficiency of the equipment has been greatly improved; the improvement in product quality means that according to the quality inspection results, the wall thickness fluctuation rate of the produced stainless steel single-core pipe has been reduced from the original 2.5% to 1.2%, the surface roughness has been reduced from the original Ra = 0.6 μm to Ra = 0.3 μm, the unit energy consumption has been reduced from the original 5.6 kWh per meter to 4.2 kWh, and the drawing fracture rate has been reduced from the original 1.8% to 0.9%; the reduction in energy consumption means that the optimized production process has reduced the energy consumption by about 25%, effectively reducing the production cost and improving the overall efficiency of the enterprise.

[0113] Table 1 Comparison table of the optimization effect of the production method of stainless steel single-core pipe based on the improved population algorithm

[0114]

[0115] Table 1 shows that the process optimization method applied in the present invention has brought significant improvements. In the optimized production process, the drawing speed has increased by 20%, and the lubrication pressure and annealing temperature have been reasonably adjusted, resulting in a decrease in the coolant temperature and an appropriate reduction in the forming tension, thereby reducing the quality fluctuation of the product. Especially in key quality indicators such as the wall thickness fluctuation rate, surface roughness, unit energy consumption, and drawing fracture rate, significant improvements have been shown after optimization. In addition, the shortening of the production cycle and the reduction of energy consumption have directly saved costs for the enterprise and improved the operating efficiency of the production line.

[0116] Through real-time data feedback and dynamic optimization, the intelligent control system of the present invention enables the production process to be continuously optimized, not only meeting the high-quality requirements of stainless steel single-core pipes, but also improving the flexibility and adaptability of the production line, and ultimately enhancing the comprehensive benefits of the enterprise. This embodiment demonstrates the strong adaptability and excellent optimization effect of the present invention in the actual production environment, verifying its feasibility and innovation.

[0117] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A production method of stainless steel single-core tube based on an improved swarm algorithm, characterized in that, It includes the following steps: S1. Collect the process parameters in the production process of the stainless steel single-core tube and construct a process parameter dataset; S2. Based on the process parameter dataset, construct a multi-objective decision network model. The input is the process parameter, the output is the quality data, and there is an adjustable target weight channel. Generate a comprehensive performance score through conflict reconciliation between process objectives; S3. Use the particle swarm optimization algorithm to optimize the process parameters. Represent the particle position as the process parameter and the particle velocity as the process adjustment amount; S4. During the particle swarm iteration process, use the comprehensive performance score to predict the target performance of each group of particle position combinations, calculate the comprehensive fitness value, update the particle position according to the comprehensive fitness value, adjust the particle velocity magnitude according to the change rate of the comprehensive fitness value, and iteratively update the particle position until the optimal process parameter combination is obtained; S5. Apply the optimal process parameter combination to each section of the actual production of the stainless steel single-core tube; S6. Collect the actual quality data of the stainless steel single-core tube generated under optimized control, compare it with the target performance prediction value, calculate the error and feedback it to the target weight channel to correct the weight configuration between each target.

2. The production method of a stainless steel single-core tube based on an improved swarm algorithm according to claim 1, wherein The process parameters specifically include drawing speed, lubricating pressure, annealing temperature, coolant temperature, and forming tension.

3. A production method of stainless steel single-core pipe based on an improved swarm algorithm according to claim 1, characterized in that, The quality data specifically includes wall thickness volatility, surface roughness, unit energy consumption, and drawing fracture rate.

4. A production method of stainless steel single-core tube based on an improved population algorithm according to claim 1, characterized in that, The specific content of S2 includes: S21. Construct a multi-objective decision network model with a multi-objective output structure. The multi-objective decision network model adopts a multi-layer perceptron structure, which consists of an input layer, multiple hidden layers, and an output layer. The number of input layer nodes is 5, corresponding to the input process parameters. The output layer contains 4 nodes, corresponding to the output quality data; S22. Set a target weight fusion channel in the output layer, construct a process objective reconciliation mechanism, assign corresponding weight coefficients to each output prediction value, and the target weight fusion channel supports independent weighted adjustment of each output target during the inference process to generate a comprehensive performance score; S23. Calculate the comprehensive performance score and conduct a weighted evaluation of the performance of the process parameter combination on each quality data; Among them, Ψ is the comprehensive performance score, δ w Predicted wall thickness volatility, R s Predicted surface roughness, E u Predicted unit energy consumption, λ b Predicted drawing fracture rate, Mean value of historical wall thickness volatility, Mean value of historical surface roughness, Historical mean value of unit energy consumption, Historical mean value of drawing fracture rate, ω1, ω2, ω3, ω4 are adjustable weight coefficients, very small positive numbers, μ is the conflict reconciliation penalty coefficient, ω j Weight of the jth quality data, q j Normalized deviation value of the jth quality data, ω i Weight of the ith quality data, q i Normalized deviation value of the ith quality data; S24. Dynamically adjust each target weight coefficient according to the process requirements and correct the relative weight relationship between the targets.

5. A production method of stainless steel single-core tube based on an improved swarm algorithm according to claim 1, characterized in that The specific content of S3 includes: S31. Construct a particle swarm search space suitable for optimizing the production process parameters of the stainless steel single-core tube, and define the current position of each particle as a set of process parameter combinations; S32. Set the upper and lower boundary ranges for each process parameter dimension respectively. The boundary values are determined based on the equipment control ability, historical operation data, and safety process limit. The particle swarm search space is composed of the process parameter combination range, and each position represents a set of adjustable and executable process parameter combinations in the production of the single-core tube; S33. Set the velocity vector of each particle, which is expressed as the variable process adjustment amount of the corresponding process parameter in the current iteration, and each dimensional velocity component corresponds to the dynamic change amount of the process parameter.

6. A production method of a stainless steel single-core tube based on an improved swarm algorithm according to claim 1, characterized in that, The specific content of S4 includes: S41. During the particle swarm optimization iteration process, use the comprehensive performance score to predict the target performance of each group of particle position combinations and calculate the comprehensive fitness value of each group of particle position combinations: Among them, i (t) is the comprehensive fitness score of the i-th particle in the t-th iteration, Ψ is the comprehensive performance score, μ is the conflict reconciliation penalty coefficient, ω j is the weight of the j-th quality data, q j is the normalized deviation value of the j-th quality data, ω i is the weight of the i-th quality data, q i is the normalized deviation value of the i-th quality data, ω k is the original target weight, To normalize the relative error between the predicted value and the reference value and amplify it through an exponential function, j is the index of the target weight, k is the target number index, is the historical mean of the k-th quality indicator, and ε is a very small positive number; S43. According to the comprehensive fitness value Ψ i (t) of the current particle, compare it with the historical optimal fitness and the global optimal fitness. If the comprehensive fitness value of the current particle is less than the historical individual optimal fitness, update the individual optimal position to the current particle position; if the comprehensive fitness value of the current particle is less than the global optimal fitness, update the global optimal position to the current particle position; S44. Calculate the change rate ΔΨ of the comprehensive fitness value i = |Ψ i (t) - Ψ i (t - 1)|. Dynamically adjust the drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension in the particle velocity vector according to the change rate. θ is the set threshold of the fitness change rate. When ΔΨ i < θ, reduce the velocity component to enhance the local search ability; when ΔΨ i ≥ θ, increase the velocity component to improve the search jumping ability; S45. Iteratively update the particle positions until the optimal process parameter combination is obtained.

7. A production method of stainless steel single-core pipe based on an improved swarm algorithm according to claim 7, characterized in that, The specific steps of S45 include: S451. Based on the updated velocity vector and the current particle position, perform a position update operation. Based on the process parameter combination represented by the current particle, numerically adjust the five process parameters of drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension in sequence. The adjustment amount of each process variable is provided by the corresponding component in the velocity vector to obtain the next-round particle position. S452. Repeat the operations of target performance prediction, comprehensive fitness calculation, position update, and velocity adjustment until the number of iterations reaches the maximum number of iterations T max = 200 times or within 10 consecutive rounds, the global optimal fitness score value of the particle swarm satisfies the change convergence condition Ψ g (t) - Ψ g (t - 1)| < ∈, ∈ = 10 -4 , and output the optimal process parameter combination corresponding to the global particle position with the optimal comprehensive score.

8. A production method of stainless steel single-core tube based on an improved population algorithm according to claim 1, characterized in that, The specific steps of S5 include: S51. Map the optimal process parameter combination, including drawing speed, lubrication pressure, annealing temperature, coolant temperature, and forming tension, respectively, to the five key process sections of the stainless steel single-core pipe production line, namely the drawing section, lubrication section, annealing section, cooling section, and tension control section; by establishing parameter interface connections with the control modules of each section, accurately load the corresponding parameter values into each execution system as the standard process setting values for the current production batch, and at the same time complete the writing of the setting values and status synchronization, so that each section is in the loaded and pending execution state before the task starts. The loading process supports real-time parameter verification and automatic backup. S52. After receiving the production task execution instruction, automatically call the set optimal process parameter combination and use it as the main control benchmark to perform real-time process scheduling and sectional control on the entire production line; each control section executes a closed-loop control strategy based on the standard setting value, and dynamically adjusts the execution behavior by collecting the deviation between the current operating state and the target setting to achieve real-time parameter tracking and automatic adjustment; the entire production execution process is carried out strictly in accordance with the optimal process parameter combination, and from the input of section-level parameters, execution logic control to sensor feedback correction, it is under continuous and consistent target control.

9. A production method of stainless steel single-core pipe based on an improved swarm algorithm according to claim 1, characterized in that The specific steps of S6 include: S61. During the production process of the stainless steel single-core pipe, continuously collect the actual operation data of each section controlled by the optimal process parameter combination, and combine the index data output by the quality inspection module to construct the quality feedback data set corresponding to the current batch. The quality data includes wall thickness fluctuation rate, surface roughness, unit energy consumption, and drawing fracture rate. S62. Compare the measured quality indicators with the target performance values predicted by the multi-objective decision-making network model one by one to construct a quality error vector: Among them, Δy k is the absolute error value of the quality data, is the measured value, is the model predicted value, k ∈ {1, 2, 3, 4}, corresponding to four types of quality data respectively; S63. Input the quality error vector into the adjustable target weight channel to feedback and correct the weight configuration of the four quality objectives in the scoring function: Among them, ω k is the original target weight, the corrected target weight of the k-th item, β is the error sensitivity adjustment coefficient, and Δy k is the absolute error value of the quality data, and j is the index of the target weight; S64. Input the updated target weight configuration back into the multi-objective scoring function as the basis for calculating the fitness in the next round of particle swarm optimization iteration.

Images