Method for online prediction of wall thickness and dynamic optimization of saw cutting amount after sizing of hot-rolled seamless steel pipe

Abstract

The invention provides a method for online prediction of wall thickness and dynamic optimization of saw cutting amount after sizing of a hot-rolled seamless steel pipe, and relates to the technical field of mechanical automation control. The method comprises the following steps: acquiring online thickness measurement data of a pierced billet, production process parameters of a steel pipe, equipment parameters of a sizing mill and actually measured wall thickness data of a plurality of finished pipes; based on the actually measured wall thickness data, a fitting function coefficient of an actually measured finished pipe head and tail wall thickness change curve is obtained; derivation features are calculated; establishing a coefficient regression model of the wall thickness prediction function, and forming a wall thickness prediction model and a saw cutting amount prediction strategy which can be used for unknown incoming materials; and executing a mother pipe cutting optimization strategy in combination with different sizing modes, and outputting a corresponding sawing scheme and a material utilization rate evaluation result. The method solves the problems that the wall thickness of the steel pipe fluctuates non-linearly in the length direction after the sizing process, and IMS thickness measurement data cannot directly reflect the wall thickness distribution of the finished pipe.

Description

Technical Field

[0001] This invention relates to the field of mechanical automation control technology, and in particular to a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes. Background Technology

[0002] In the current seamless steel pipe production process, although an IMS thickness measurement system is installed behind the continuous rolling mill to detect the wall thickness of the rough pipe, the wall thickness of the rough pipe changes after the sizing process. Therefore, the IMS wall thickness data cannot directly reflect the final wall thickness distribution of the finished steel pipe, thus limiting its direct application in the head and tail sawing process.

[0003] In actual end-to-end sawing operations, steel pipes after cooling need to be end-to-end cut using a pipe saw to meet the final finished pipe length requirements and wall thickness compliance standards. Currently, the common practice is for operators at the cooling bed to manually inspect the wall thickness of each group of pipes, determine the acceptable wall thickness range, and empirically set the sawing amount accordingly. This information is then manually entered into the HMI system to execute the sawing task. This current manual inspection method has the following drawbacks:

[0004] (1) It is difficult to detect the wall thickness of the finished pipes on the cooling bed. The number of samples is limited and cannot represent the overall wall thickness of the whole batch of steel pipes.

[0005] (2) Each steel pipe being inspected is usually measured only once at a single axial section position, making it impossible to obtain information on the wall thickness variation along the length.

[0006] (3) Reliance on operator experience and inconsistent measurement standards can easily lead to judgment bias.

[0007] The current production situation further exacerbates the above problems: the number of steel pipes in each batch on the cooling bed can reach dozens. If operators have to judge the acceptable range of wall thickness and set the sawing amount for each pipe, the workload is extremely large. At the same time, the sawing operation at the beginning and end is fast-paced. A batch needs to be sawed within an hour. It is difficult for manual workers to complete the optimization of each pipe in a short time. In order to balance the production rhythm, a conservative uniform sawing amount is often adopted, which leads to the common phenomenon of over-cutting and significantly affects the yield.

[0008] In summary, the current wall thickness determination method, which relies primarily on manual sampling on the cooling bed, suffers from serious deficiencies in data representativeness and accuracy. It also struggles to effectively identify out-of-tolerance areas at the beginning and end of the steel pipe, leading to conservatively set sawing amounts, wasted effective pipe length, and a significant impact on product yield. Therefore, to establish a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes, the following key issues urgently need to be addressed:

[0009] (1) Since IMS wall thickness data cannot directly reflect the final wall thickness distribution of the finished pipe, it is necessary to establish a prediction model from IMS wall thickness data to the wall thickness of the finished pipe in order to predict the wall thickness distribution of the finished pipe.

[0010] (2) Based on the prediction, it is necessary to identify the abnormal wall thickness area along the axial direction of the steel pipe, assist in judging the range of deviation at the beginning and end, and provide a quantitative basis for judging the sawing amount;

[0011] (3) Under the premise of ensuring that the finished steel pipes meet the requirements of order length and wall thickness compliance, it is necessary to set the optimal sawing amount through algorithm optimization to minimize invalid cutting and improve the yield of each steel pipe.

[0012] Based on the above characteristics, this invention will rely on IMS detection data and combine it with the characteristics of the sizing process to establish a wall thickness prediction model. Based on the prediction, targeted sawing optimization algorithms will be designed for both tail-aligned and non-tail-aligned sizing methods. Summary of the Invention

[0013] To address the technical problems of nonlinear fluctuations in the wall thickness of steel pipes along the length direction after sizing and the inability of IMS thickness measurement data to directly reflect the wall thickness distribution of the finished pipe, and to address the issue of the high workload and difficulty in matching manual sawing amount setting with the rolling rhythm, this invention provides a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes. The technical solution is as follows:

[0014] A method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes, the method comprising:

[0015] S1. Obtain online thickness measurement data of raw pipe, steel pipe production process parameters, sizing machine equipment parameters, and actual measured wall thickness data of finished pipe;

[0016] S2. Based on the measured wall thickness data of the finished pipe, fit the wall thickness variation function along the length direction to obtain the fitting function coefficients of the measured wall thickness variation curve at the head and tail of the finished pipe.

[0017] S3. Calculate derived features. Using the fitted function coefficients obtained in step S2 as the target variables, feature importance analysis is performed using random forest, L1 regularized regression with cross-validation (LassoCV), and elastic net regularized regression with cross-validation (ElasticNetCV) feature selection algorithms. Key features are selected through intersection priority and weighted fusion strategies.

[0018] S4. Based on the key features selected in step S3, establish coefficient regression models for calculating the wall thickness prediction function, forming a wall thickness prediction model and sawing volume prediction strategy that can be used for unknown incoming materials.

[0019] S5. Determine the sawing amount based on the wall thickness prediction model and sawing amount prediction strategy, and execute the mother tube cutting optimization strategy in combination with different fixed length methods to output the corresponding sawing scheme and material utilization evaluation results.

[0020] The online thickness measurement data of the raw pipe in S1 is the average of at least 6 channel wall thickness values ​​obtained at sampling intervals s within a sampling section length D from pipe head to pipe tail and from pipe tail to pipe head. and Where D ranges from 500 to 1000 mm, and s ranges from 5 to 10 mm.

[0021] Steel pipe production process parameters include raw pipe specifications and wall thickness (t). nominal outer diameter d of the rough pipe nominal Finished pipe specifications and wall thickness (t) spec and the outer diameter d of the finished pipe spec ;

[0022] Sizing mill equipment parameters include the rotational speed (rpm) of each frame of the sizing mill. u Inlet velocity v entry Export speed v exit and final rolling temperature T rolling The sizing machine has 14 frames, u = 1, 2, 3, ..., 14;

[0023] The measured wall thickness data of the finished pipe is obtained by averaging at least three channels of wall thickness values ​​collected at sampling intervals s' within a sampling section length D, which is the same as the online thickness measurement data of the raw pipe, from the pipe head to the pipe tail and from the pipe tail to the pipe head. The position vector of the sampling point along the length direction of the i-th finished pipe head is then taken. Collect the corresponding wall thickness measurement vector The sampling point position vector along the length direction at the tail of the i-th finished pipe Collect the corresponding wall thickness measurement vector Where s'≥50mm, y i,j For the i-th finished pipe at position x j The measured wall thickness at the location was used to fill in missing values ​​using linear interpolation to form a complete measurement sequence.

[0024] In step S2, for the measured wall thickness data of each finished pipe, the measurement sequences of the head and tail regions are fitted with a quadratic polynomial function, as shown in the following formula:

[0025] T post (x)=a0+a1x+a2x 2

[0026] Among them, T post(x) represents the wall thickness of the steel pipe at a distance x from the pipe end, in mm; a0, a1, and a2 are the coefficients of the fitting function.

[0027] For head measurement data and tail measurement data Fit the corresponding parameter vectors using the least squares method respectively. and To minimize the sum of squared errors, we obtain the fitting coefficients, which serve as parameters for the wall thickness variation model. The formula is as follows:

[0028]

[0029] in, and To determine the position of the i-th finished pipe at position x j The measured wall thickness at the head and tail sections.

[0030] The calculation process for derived features in S3 is as follows:

[0031] S311. Calculate the head and tail derived features:

[0032] The head-derived features include the average thickness T of the front section of the head tube. head Head of the rough pipe wall thickness slope k head Maximum wall thickness T of the front section of the head pipe max,head The sudden change value of wall thickness ΔT in the middle section of the head tube. mid,head Head section wall thickness and curvature k tail,head and the abrupt change in wall thickness ΔT between the head and tail sections of the tube tail,head The following formula is used for calculation:

[0033]

[0034] in, x is the distance from the head k = k·smm online thickness measurement of the rough pipe, in mm; and x within 6 sampling points from the tube head k and The mean value, in mm; s is the sampling interval. D is the sampling segment length;

[0035] The tail-end derived features include the average thickness T of the front section of the tail-end tube. tail The slope of the rough pipe wall at the tail end. tail Maximum wall thickness T at the front section of the tail section of the rough pipe max,tail The abrupt change in wall thickness ΔT in the middle section of the tail section of the tube mid,tail Tail section wall thickness curvature k tail,tail and the abrupt change in wall thickness ΔT at the tail end of the tubetail,tail The calculation is performed as follows:

[0036]

[0037] In the formula, x is the distance from the tail k = k·smm online thickness measurement of the rough pipe, in mm; and x within 6 sampling points from the end of the tube k and The mean value, in mm; s is the sampling interval.

[0038] D is the sampling segment length;

[0039] S312. Derived features calculated based on sizing machine equipment parameters:

[0040] The derived characteristics of the sizing mill equipment parameters include the final rolling thermodynamic temperature T. k Plastic deformation capacity T of hot-rolled materials e The rolling speed gradient Δv and the standard deviation of the rolling mill motor speed σ rpm The calculation is performed as follows:

[0041] T k =T rolling +273.15

[0042]

[0043] Δv=v exit -v entry

[0044]

[0045] In the formula, Q is the activation energy constant, J / mol; R is the gas constant, J / (mol·K); T rolling v represents the final rolling temperature, in °C. exit and v entry These are the outlet velocity and inlet velocity, respectively, in m / s and rpm. u Let u be the rotational speed of each frame, where u = 1, 2, ..., 14; The average rotational speed of the frame;

[0046] S313. Calculation of derived features based on steel pipe production process parameters:

[0047] The derived characteristics of the steel pipe production process parameters include the thickness reduction rate and the rolling load difference ΔA, which are calculated in the following way:

[0048]

[0049] In the formula, t nominal The wall thickness of the rough pipe is specified in mm; d nominal The outer diameter of the rough pipe is in mm; t spec The wall thickness of the finished pipe is in mm; d spec The outer diameter of the finished pipe is in mm.

[0050] The specific process of feature importance analysis in S3 is as follows:

[0051] Using online thickness measurement data of raw pipes, steel pipe production process parameters, sizing machine equipment parameters, and calculated derived features as input variables, and with the wall thickness variation function coefficients a0, a1, and a2 fitted in S2 as target variables, three regression tasks were constructed, and feature importance analysis was performed using the following feature selection methods:

[0052] S321. Model based on Random Forest regression model, input derived features, calculate the importance score of each derived feature through mean squared error division standard, and normalize to the [0,1] interval;

[0053] S322. Use L1 regularized regression with cross-validation (LassoCV) to automatically adjust the regularization strength, screen out features with zero coefficients, and consider the features corresponding to non-zero coefficients as important features.

[0054] S323. Use ElasticNetCV with cross-validation to perform sparsity selection of features by fusing L1 and L2 regularization, evaluate the contribution of each feature in the linear combination, and obtain a list of ranked features by importance.

[0055] The specific process for selecting key features in S3 is as follows:

[0056] S331. Based on intersection priority: Features with a weight greater than 30% in all models are retained as stable and robust features and are retained as the first-level feature set.

[0057] S332. Weighted Fusion: For the remaining features after selection in S331, set fusion weights and calculate the feature fusion score S. f The following formula is used for calculation:

[0058] S f =0.4·I RF +0.3·I Lasso +0.3·I ENet

[0059] In the formula, I RF I Lasso I ENetThese are the normalized feature importance scores for three models: Random Forest, L1 Regularized Regression with Cross-Validation, and Elastic Net Regularized Regression with Cross-Validation.

[0060] The model is based on a random forest regression model. The importance score of each feature is calculated using the mean squared error criterion and normalized to the [0,1] interval. The normalized score is used as the random forest feature importance score I. RF ;

[0061] L1 regularized regression with cross-validation is used to automatically adjust the regularization strength, filtering out features with zero coefficients and normalizing the absolute values ​​of their non-zero coefficients to the [0,1] interval. These normalized values ​​are then used as the Lasso feature importance score I. Lasso ;

[0062] Elastic mesh regularized regression with cross-validation, incorporating L1 and L2 regularization, is used to sparsify feature selection, evaluate the contribution of each feature in the linear combination, and normalize the absolute values ​​of its regression coefficients to the [0,1] interval. These normalized values ​​are then used as the elastic mesh feature importance score I. Enet ;

[0063] According to the fusion score S f The remaining features are sorted, and the top 30% of features by importance score are selected as the secondary feature set. Finally, a comprehensive key feature subset consisting of primary and secondary features is output for modeling.

[0064] S4 includes:

[0065] S41. Based on the key feature variables obtained from S3, construct prediction functions for the wall thickness of the head and tail of the finished pipe along the length direction, with the prediction functions for the head and tail sections having the same form.

[0066] S42. Before the raw tube enters the sizing machine unit, the industrial site collects online thickness measurement data and relevant production process parameters of the raw tube in real time. After the sizing process is completed, the operating parameters of the sizing machine are obtained synchronously. Based on this, the wall thickness prediction function coefficient is calculated and substituted into the prediction function. The continuous wall thickness distribution trend at both ends of the finished tube is simulated within the length range of the head and tail of the finished tube from 0 to the sampling section length Dmm, and the prediction and reconstruction of the wall thickness curve of the finished tube is completed.

[0067] S43, Based on the target finished pipe specification thickness t spec The relationship with the prediction function, respectively for the head prediction function and tail prediction function Make a judgment and define the recommended sawing amount x. cut The recommended sawing amount is determined by finding the first point in the thickness prediction function where the thickness is less than or equal to the specification requirement.

[0068] x cut =min{x∈[0,L]|T(x)≤t spec}

[0069] In the formula, T(x) represents the head prediction function. and tail prediction function

[0070] The head prediction function in S41 With coefficient The calculation formula is as follows:

[0071]

[0072] In the formula, x is the predicted position, in mm; T represents the predicted wall thickness of the finished pipe at location x, in mm. head Δv is the average thickness of the front section of the rough tube, in mm; Δv is the rolling speed gradient of the steel pipe, in m / s; T k The final rolling thermodynamic temperature, K; t spec Here is the wall thickness of the finished pipe, in mm; Δ is a constant correction term; k head T represents the slope of the rough pipe wall thickness at the head, in mm / mm; max,head ΔT represents the maximum wall thickness of the front section of the rough pipe, in mm; mid,head The value of the abrupt change in wall thickness in the middle section of the head tube is shown in mm; k tail,head The curvature of the wall thickness of the head section and tail section of the rough pipe, in mm. -1 ;ΔT tail,head , where is the abrupt change in wall thickness of the head section and tail section, in mm; Q is the activation energy constant, in J / mol; R is the gas constant, in J / (mol·K); All regression coefficients are obtained through multiple linear regression analysis based on measured head data. Here, o takes the values ​​0, 1, 2, or 3 to distinguish the coefficients corresponding to different features.

[0073] Tail prediction function With coefficient The calculation formula is as follows:

[0074]

[0075] In the formula, T represents the predicted wall thickness of the finished pipe at location x, in mm. tail Δv is the average thickness of the front section of the tail section of the rough tube, in mm; Δv is the rolling speed gradient of the steel pipe, in m / s; T k The final rolling thermodynamic temperature, K; t spec Here is the wall thickness of the finished pipe, in mm; Δ is a constant correction term; k tail T represents the linear slope of the wall thickness at the tail of the rough pipe, in mm / mm;max,tail ΔT represents the maximum wall thickness of the front section of the tail section of the rough tube, in mm; mid,tail The value of the abrupt change in wall thickness in the middle section of the tail section of the tube, in mm; k tail,tail The curvature of the wall thickness of the tail section of the rough tube, in mm. -1 ;ΔT tail,tail , where is the abrupt change in wall thickness at the tail end of the rough tube, in mm; Q is the activation energy constant, in J / mol; R is the gas constant, in J / (mol·K); All regression coefficients are obtained through multiple linear regression analysis based on measured tail data, where o takes the values ​​0, 1, 2 or 3 to distinguish the coefficients corresponding to different features.

[0076] In S5, based on the head prediction function and tail prediction function The determined recommended sawing amount and Combined with the process settings, the standard length range of the sub-tube [L] min ,L max The length L of the main tube in the slatted saw was detected in real time. i We construct sawing optimization strategies applicable to tail-aligned and non-tail-aligned scenarios using the short-scale ratio threshold ε, and output the optimal sawing scheme and material utilization evaluation results.

[0077] The calculation process for the optimal sawing scheme and material utilization evaluation results for both tail-aligned and non-tail-aligned scenarios is as follows:

[0078] S51, Tail Alignment Scenario:

[0079] Based on the predicted sawing amount of all the mother tubes to be cut, a uniform head removal length S is determined. H The calculation formula is as follows:

[0080]

[0081] In the formula, The recommended cutting amount is predicted for the head of the i-th mother tube; N is the number of mother tubes in this group of gang saws;

[0082] The number of segments (n) to be cut from the head of the target sub-tube to a fixed length (l) is: i.align -1, the length of the remaining tail segment is r i,align Furthermore, the tail-aligned cutting length S is calculated uniformly based on the tail-predicted recommended sawing amount. T The calculation formulas for the three are as follows:

[0083]

[0084] r i,align =R i -(n i,align-1)·l

[0085]

[0086] In the formula, l satisfies L min ≤l≤L max ; Predict and recommend the sawing amount for the tail of the i-th mother tube;

[0087] Based on the above cutting method, an integer optimization model is constructed for the fixed length l: to maximize the material utilization rate U(l) of the tail-aligned fixed length method under the constraints that the short length rate does not exceed the short length rate threshold ε, the fixed length range is constrained, the length of the remaining tail section is non-negative, and the tail-cut section does not form a short ruler tube. The integer optimization model is constructed as follows:

[0088]

[0089] stL min ≤l≤L max

[0090]

[0091] In the formula, N represents the number of main tubes in the gang saw; L i R is the original length of the i-th mother tube, in mm; i Let be the effective length of the i-th mother tube after sawing, in mm; l be the target length, a decision variable; n be the number of sub-tube segments cut from each mother tube; r i,align S is the length of the tail section of the i-th mother pipe, in mm; T To standardize the tail resection amount, mm; M short M represents the total number of short ruler tubes; total The total number of all sub-tubes;

[0092] The optimized fixed length Cut length from head and tail Together with the recommended parameters for the group of gang saw cutting under the tail-aligned fixed-length sawing method, the corresponding sawing scheme, short length rate and material utilization rate evaluation results were obtained;

[0093] S52, Non-tail-aligned scenarios:

[0094] Based on the predicted sawing amount of all the mother tubes to be cut, a uniform head removal length S is determined. H The calculation formula is as follows:

[0095]

[0096] For each mother pipe P i The remaining length after removing the excessively thick sections at the head and tail is denoted as R. i The calculation formula is as follows:

[0097] R i =L i -S H

[0098] In the non-tail-aligned method, each mother tube is cut sequentially from the head according to a certain target length l until the remaining segment is less than l, and the last segment is the tail segment r. i,nonalign Whether it is retained as a sub-tube depends on whether it meets the minimum sub-tube length requirement L. min The number of standard length sections in each mother tube is denoted as n. i,nonalign The calculation formula is as follows:

[0099]

[0100] r i,nonalign =R i -n i,nonalign ·l

[0101] To maximize material utilization in non-tail-aligned fixed-length cutting Simultaneously, by controlling the proportion of short ruler tubes to not exceed the short ruler ratio threshold ε and constraining the fixed ruler length range, the following integer optimization model is constructed:

[0102]

[0103] stL min ≤l≤L max

[0104]

[0105] In the formula, n i,nonalign r represents the number of daughter tubes completely cut from the i-th mother tube; i δ represents the remaining length of the main pipe, in mm. i M is an indicator variable for whether to retain the tail segment sub-tube; short M represents the number of short ruler tubes. total The total number of all sub-tubes; ε is the short-scale rate threshold;

[0106] The optimized fixed length with head resection length Together with the recommended parameters for this group of gang saw cutting under the non-tail-aligned fixed-length sawing method, the corresponding sawing scheme, short length rate and material utilization rate evaluation results are obtained. Based on the specific length of the remaining section at the tail of each mother tube, it is determined whether to retain it as an effective daughter tube. Finally, the sawing scheme, short length rate and material utilization rate evaluation results for this batch are obtained.

[0107] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0108] The above solution addresses the issues of nonlinear fluctuations in steel pipe wall thickness along its length after the sizing process and the inability of IMS thickness measurement data to directly reflect the wall thickness distribution of the finished pipe. Through multi-model feature importance analysis and key feature screening, the accuracy of wall thickness prediction is improved. A dynamic optimization model for sawing is established based on the prediction results, enabling intelligent optimization of the main pipe cutting method, effectively improving material utilization and reducing the short-length rate. It also solves the difficulty of relying on manual measurement for main pipe sawing, contributing to improved production automation and intelligence. Attached Figure Description

[0109] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0110] Figure 1 This is a flowchart of a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes provided in an embodiment of the present invention. Detailed Implementation

[0111] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0112] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0113] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0114] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0115] This invention provides a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes. For example... Figure 1 The flowchart shown is a method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes. This method may include the following steps:

[0116] S1. Obtain online thickness measurement data of raw pipe, steel pipe production process parameters, sizing machine equipment parameters, and actual measured wall thickness data of finished pipe;

[0117] S2. Based on the measured wall thickness data of the finished pipe, fit the wall thickness variation function along the length direction to obtain the fitting function coefficients of the measured wall thickness variation curve at the head and tail of the finished pipe.

[0118] S3. Calculate derived features. Using the fitted function coefficients obtained in step S2 as target variables, use random forest, L1 regularized regression with cross-validation and elastic net regularized regression with cross-validation feature selection algorithms to perform feature importance analysis. Select key features through intersection priority and weighted fusion strategies.

[0119] S4. Based on the key features selected in step S3, establish coefficient regression models for calculating the wall thickness prediction function, forming a wall thickness prediction model and sawing volume prediction strategy that can be used for unknown incoming materials.

[0120] S5. Determine the sawing amount based on the wall thickness prediction model and sawing amount prediction strategy, and execute the mother tube cutting optimization strategy in combination with different fixed length methods to output the corresponding sawing scheme and material utilization evaluation results.

[0121] The following description, in conjunction with specific embodiments, illustrates this point.

[0122] This embodiment uses field data from a steel pipe factory before and after sizing, including online thickness measurement data T of the rough pipe with a sampling section length of 600mm and a sampling interval of 10mm from pipe head to pipe tail and from pipe tail to pipe head. pre (x k (x) k =10k, k=0,1,...,60), rough pipe specifications and wall thickness t nominal outer diameter d of the rough pipe nominal Finished pipe specifications and wall thickness (t) spec Finished pipe specifications, outer diameter d spec The sizing mill has 14 stands, and the rotational speed of each stand is 1-10 rpm. 14 Inlet velocity v entry Export speed v exit Final rolling temperature T rolling Measured wall thickness data of several finished pipes y i,j .

[0123] S1: Obtain online thickness measurement data of raw pipes, steel pipe production process parameters, sizing machine equipment parameters, and measured wall thickness data of several finished pipes, as shown in Tables 1, 2, 3, 4, 5, and 6:

[0124] Table 1 Thickness data of the head pipe along the length direction

[0125]

[0126]

[0127] Table 2 Thickness data of the tail section of the unfinished pipe along its length

[0128] batch Lot No. 0 / mm 10 / mm … 590 / mm 600 / mm 241219A21 242B12743241219A210063 15.42 15.38 … 13.48 13.36 241220A07 242B12774241220A070040 16.39 16.32 … 14.55 14.46 241220A07 242B12774241220A070050 16.87 16.79 ... 15.18 15.09 241220A07 242B12774241220A070051 16.91 16.86 … 15.12 15.05 241220A07 242B12774241220A070054 16.88 16.80 … 14.93 14.84 241220A07 242B12774241220A070055 16.58 16.52 … 14.58 14.50 250418A19 252B03819250418A190057 16.54 16.49 … 14.40 14.32 250418A19 252B03819250418A190059 16.51 16.44 … 13.93 13.93 250418A19 252B03819250418A190060 16.55 16.47 … 13.81 13.72 250418A19 252B03819250418A190065 16.51 16.43 … 14.79 14.72

[0129] Table 3 Steel Pipe Manufacturing Process Parameters

[0130] batch Lot No. <![CDATA[t nominal / mm]]> <![CDATA[d nominal / mm]]> <![CDATA[t spec / mm]]> <![CDATA[d spec / mm]]> 241219A21 242B12743241219A210063 16.00 224.80 16.00 203.00 241220A07 242B12774241220A070040 18.00 224.80 18.00 194.00 241220A07 242B12774241220A070050 18.00 224.80 18.00 194.00 241220A07 242B12774241220A070051 18.00 224.80 18.00 194.00 241220A07 242B12774241220A070054 18.00 224.80 18.00 194.00 241220A07 242B12774241220A070055 18.00 224.80 18.00 194.00 250418A19 252B03819250418A190057 18.00 224.80 18.00 180.00 250418A19 252B03819250418A190059 18.00 224.80 18.00 180.00 250418A19 252B03819250418A190060 18.00 224.80 18.00 180.00 250418A19 252B03819250418A190065 18.00 224.80 18.00 180.00

[0131] Table 4 Sizing Machine Equipment Parameters

[0132]

[0133]

[0134] Table 5 Measured wall thickness data of the finished head tube

[0135] batch Lot No. 0 / mm 50 / mm … 550 / mm 600 / mm 241219A21 242B12743241219A210063 17.10 17.42 … 15.98 15.97 241220A07 242B12774241220A070040 18.16 18.23 … 17.94 17.92 241220A07 242B12774241220A070050 18.64 18.49 ... 17.89 17.89 241220A07 242B12774241220A070051 18.29 18.26 … 17.89 17.88 241220A07 242B12774241220A070054 18.30 18.66 … 17.98 17.98 241220A07 242B12774241220A070055 17.58 18.34 … 17.85 17.80 250418A19 252B03819250418A190057 18.03 19.20 … 17.85 17.87 250418A19 252B03819250418A190059 18.55 19.37 … 17.94 17.93 250418A19 252B03819250418A190060 18.78 18.27 … 17.82 17.72 250418A19 252B03819250418A190065 18.96 18.91 … 17.93 17.91

[0136] Table 6 Measured wall thickness data of the finished pipe at the tail end

[0137] batch Lot No. 0 / mm 50 / mm … 550 / mm 600 / mm 241219A21 242B12743241219A210063 15.92 16.03 … 16.85 16.95 241220A07 242B12774241220A070040 17.88 17.90 … 18.01 18.06 241220A07 242B12774241220A070050 17.78 17.84 ... 18.36 18.44 241220A07 242B12774241220A070051 17.69 17.75 … 18.29 18.33 241220A07 242B12774241220A070054 17.75 17.81 … 18.33 18.40 241220A07 242B12774241220A070055 17.68 17.75 … 18.02 18.11 250418A19 252B03819250418A190057 17.68 17.75 … 18.12 18.21 250418A19 252B03819250418A190059 17.70 17.78 … 18.22 18.34 250418A19 252B03819250418A190060 17.58 17.64 … 18.00 18.08 250418A19 252B03819250418A190065 17.65 17.71 … 18.18 18.24

[0138] S2: Based on the measured wall thickness data of several finished pipes, a function for the variation of wall thickness along the length direction is fitted for each finished pipe to obtain the fitting function coefficients of the measured wall thickness variation curves at the head and tail of the finished pipe. As shown in Table 7:

[0139] Table 7. Fitting function coefficients for each finished pipe

[0140]

[0141] S3: Calculate derived features. Using the function coefficients as target variables, feature selection algorithms such as Random Forest, LassoCV with cross-validation, and ElasticNetCV with cross-validation are employed to analyze feature importance. Key features are selected using an intersection-first and weighted fusion strategy. The specific steps are as follows:

[0142] (1) The derived features of the head wall thickness prediction function include the average thickness T of the front section of the head tube. head Head of the rough pipe wall thickness slope k head The maximum wall thickness T of the front section of the head pipe max,head The abrupt change in wall thickness ΔT in the middle section of the head tube mid,head The wall thickness and curvature k of the head section of the unfinished pipe tail,headThe abrupt change in wall thickness ΔT between the head and tail sections of the tube tail,head Final rolling thermodynamic temperature T k Plastic deformation capacity T of hot-rolled materials e Steel pipe rolling speed gradient Δv, rolling mill motor speed standard deviation σ rpm The thickness reduction rate and rolling load difference ΔA were calculated, and the results are shown in Table 8.

[0143] Table 8. Derivative characteristics of each finished pipe head

[0144]

[0145] Table 9. Derivative characteristics of each finished pipe tail section

[0146] batch Lot No. <![CDATA[T tail / mm]]> <![CDATA[k tail / mm]]> <![CDATA[T max,tail / mm]]> ... Δv / (m / s) 241219A21 242B12743241219A210063 15.9124 0.0078 16.5400 ... 0.1000 241220A07 242B12774241220A070040 16.5721 0.0051 17.2105 ... 0.1500 241220A07 242B12774241220A070050 17.1029 0.0049 17.6677 ... 0.1500 241220A07 242B12774241220A070051 17.0895 0.0051 17.6602 ... 0.1500 241220A07 242B12774241220A070054 17.0943 0.0050 17.6725 ... 0.1500 241220A07 242B12774241220A070055 16.9070 0.0064 17.5808 ... 0.1500 250418A19 252B03819250418A190057 16.5711 0.0037 16.9310 ... 0.2500 250418A19 252B03819250418A190059 16.4243 0.0053 16.9501 ... 0.2500 250418A19 252B03819250418A190060 15.8559 0.0059 16.4622 ... 0.2500 250418A19 252B03819250418A190065 16.7778 0.0045 17.2623 ... 0.2500

[0147] (2) All features were analyzed using function coefficients as target variables. Random forest, L1 regularized regression with cross-validation, and elastic network regularized regression with cross-validation were used for feature importance analysis. Key features were selected based on the intersection-first and weighted fusion strategy. The results of feature selection are shown in Tables 10 and 11.

[0148] Table 10 Head Feature Screening Results

[0149]

[0150] Table 11 Results of tail feature screening

[0151]

[0152] S4: Based on the selected features and the measured head and tail data, regression coefficients are obtained through multiple linear regression analysis. The following head wall thickness prediction function is then established to predict the sawing amount:

[0153]

[0154]

[0155] Similarly, a tail wall thickness prediction function is established to predict the sawing amount:

[0156]

[0157] Based on the head and tail wall thickness prediction functions, prediction curves for wall thickness variation in the first 600mm of the head and the first 600mm of the tail are generated respectively. The intersection points of these prediction curves and the finished pipe specification wall thickness are then calculated, and the intersection point determines the predicted sawing amount. The theoretical sawing amount is calculated by intersecting the measured wall thickness data of the finished pipe with the specification wall thickness. For multiple steel pipes within the same batch, a uniform sawing amount is generally used on-site. Therefore, this method first calculates the predicted sawing amount for each steel pipe in the batch, and takes the maximum value as the uniform predicted sawing amount for the batch. The predicted sawing amount, theoretical sawing amount, on-site sawing amount, and uniform predicted sawing amount are compared to obtain the average uniform savings, as shown in Tables 12 and 13.

[0158] Table 12 Predicted Cutting Amount for Each Finished Pipe Head

[0159]

[0160]

[0161] Table 13 Predicted sawing quantity for the tail end of each finished pipe

[0162]

[0163] This example uses selected steel pipes from three batches: 241219A21, 241220A07, and 250418A19. Comparison shows that the predicted values ​​for all samples are significantly better than the actual sawing values, while still being no less than the theoretical sawing values. The formulas for calculating the average head-to-tail savings and improved yield are as follows:

[0164]

[0165] Average uniform savings = Average head uniform savings + Average tail uniform savings

[0166]

[0167] In the formula, N is the batch number.

[0168] Calculations were performed on selected steel pipes from three batches: 241219A21, 241220A07, and 250418A19. The average uniform savings and increased yield are shown in Table 14.

[0169] Table 14 Average Unified Savings and Improved Yield

[0170] batch Average uniform savings / mm Improve yield / % 241219A21 119.09 1.25 241220A07 433.62 4.56 250418A19 371.03 3.91

[0171] S5: Based on the wall thickness prediction model and the sawing amount prediction strategy, the sawing point is determined. Combined with different fixed length methods, the mother tube cutting optimization strategy is executed to obtain the optimized fixed length, predicted head and tail sawing amount, corresponding sawing scheme, optimal fixed length, material utilization rate and short length rate evaluation results. The tail alignment scheme data results and non-tail alignment scheme data results are shown in Table 15 and Table 16.

[0172] Table 15 Data Results of Tail Alignment Scheme

[0173]

[0174] Table 16 Data Results for Non-Tail Alignment Schemes

[0175]

[0176]

[0177] The symbols involved in the above process are explained in Table 17.

[0178] Table 17 Explanation of Formula Symbols

[0179]

[0180]

[0181]

[0182]

[0183] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipes, characterized in that, The method includes: S1. Obtain online thickness measurement data of raw pipe, steel pipe production process parameters, sizing machine equipment parameters, and actual measured wall thickness data of finished pipe; S2. Based on the measured wall thickness data of the finished pipe, fit the wall thickness variation function along the length direction to obtain the fitting function coefficients of the measured wall thickness variation curve at the head and tail of the finished pipe. S3. Calculate derived features. Using the fitted function coefficients obtained in step S2 as target variables, use random forest, L1 regularized regression with cross-validation and elastic net regularized regression with cross-validation feature selection algorithms to perform feature importance analysis. Select key features through intersection priority and weighted fusion strategies. S4. Based on the key features selected in step S3, establish coefficient regression models for calculating the wall thickness prediction function, forming a wall thickness prediction model and sawing volume prediction strategy that can be used for unknown incoming materials. S5. Determine the sawing amount based on the wall thickness prediction model and sawing amount prediction strategy, and execute the mother tube cutting optimization strategy in combination with different fixed length methods to output the corresponding sawing scheme and material utilization evaluation results.

2. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, The online thickness measurement data of the raw pipe in S1 is the average of at least 6 channel wall thickness values ​​obtained at sampling intervals s within a sampling section length D from pipe head to pipe tail and from pipe tail to pipe head. and Where D ranges from 500 to 1000 mm, and s ranges from 5 to 10 mm. ); Steel pipe production process parameters include raw pipe specifications and wall thickness (t). nominal outer diameter d of the rough pipe nominal Finished pipe specifications and wall thickness (t) spec and the outer diameter d of the finished pipe spec ; Sizing mill equipment parameters include the rotational speed (rpm) of each frame of the sizing mill. u Inlet velocity v entry Export speed v exit and final rolling temperature T rolling The sizing machine has 14 frames, u = 1, 2, 3, ..., 14; The measured wall thickness data of the finished pipe is obtained by averaging at least three channels of wall thickness values ​​collected at sampling intervals s' within a sampling section length D, which is the same as the online thickness measurement data of the raw pipe, from the pipe head to the pipe tail and from the pipe tail to the pipe head. The position vector of the sampling point along the length direction of the i-th finished pipe head is then taken. Collect the corresponding wall thickness measurement vector The sampling point position vector along the length direction at the tail of the i-th finished pipe Collect the corresponding wall thickness measurement vector Where s'≥50mm, y i,j For the i-th finished pipe at position x j The measured wall thickness at the location was used to fill in missing values ​​using linear interpolation to form a complete measurement sequence.

3. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, In step S2, for the measured wall thickness data of each finished pipe, the measurement sequences of the head and tail regions are fitted with a quadratic polynomial function, as shown in the following formula: T post (x)=a0+a1x+a2x 2 Among them, T post (x) represents the wall thickness of the steel pipe at a distance x from the pipe end, in mm; a0, a1, and a2 are the coefficients of the fitting function. For head measurement data and tail measurement data Fit the corresponding parameter vectors using the least squares method respectively. and To minimize the sum of squared errors, we obtain the fitting coefficients, which serve as parameters for the wall thickness variation model. The formula is as follows: in, and To determine the position of the i-th finished pipe at position x j The measured wall thickness at the head and tail sections.

4. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, The calculation process for derived features in S3 is as follows: S311. Calculate the head and tail derived features: The head-derived features include the average thickness T of the front section of the head tube. head Head of the rough pipe wall thickness slope k head Maximum wall thickness T of the front section of the head pipe max,head The sudden change value of wall thickness ΔT in the middle section of the head tube. mid,head Head section wall thickness and curvature k tail,head and the abrupt change in wall thickness ΔT between the head and tail sections of the tube tail,head The following formula is used for calculation: in, x is the distance from the head k = k·smm online thickness measurement of the rough pipe, in mm; and x within 6 sampling points from the tube head k and The mean value, in mm; s is the sampling interval. D is the sampling segment length; The tail-end derived features include the average thickness T of the front section of the tail-end tube. tail The slope of the rough pipe wall at the tail end. tail Maximum wall thickness T at the front section of the tail section of the rough pipe max,tail The abrupt change in wall thickness ΔT in the middle section of the tail section of the tube mid,tail Tail section wall thickness curvature k tail,tail and the abrupt change in wall thickness ΔT at the tail end of the tube tail,tail The calculation is performed as follows: In the formula, x is the distance from the tail k = k·smm online thickness measurement of the rough pipe, in mm; and x within 6 sampling points from the end of the tube k and The mean value, in mm; s is the sampling interval. D is the sampling segment length; S312. Derived features calculated based on sizing machine equipment parameters: The derived characteristics of the sizing mill equipment parameters include the final rolling thermodynamic temperature T. k Plastic deformation capacity T of hot-rolled materials e The rolling speed gradient Δv and the standard deviation of the rolling mill motor speed σ rpm The calculation is performed as follows: T k =T rolling +273.15 Δv=v exit -v entry In the formula, Q is the activation energy constant, J / mol; R is the gas constant, J / (mol·K); T rolling v represents the final rolling temperature, in °C. exit and v entry These are the outlet velocity and inlet velocity, respectively, in m / s and rpm. u Let u be the rotational speed of each frame, where u = 1, 2, ..., 14; The average rotational speed of the frame; S313. Calculation of derived features based on steel pipe production process parameters: The derived characteristics of the steel pipe production process parameters include the thickness reduction rate and the rolling load difference ΔA, which are calculated in the following way: In the formula, t nominal The wall thickness of the rough pipe is specified in mm; d nominal The outer diameter of the rough pipe is in mm; t spec The wall thickness of the finished pipe is in mm; d spec The outer diameter of the finished pipe is in mm.

5. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, The specific process of feature importance analysis in S3 is as follows: Using online thickness measurement data of raw pipes, steel pipe production process parameters, sizing machine equipment parameters, and calculated derived features as input variables, and with the wall thickness variation function coefficients a0, a1, and a2 fitted in S2 as target variables, three regression tasks were constructed, and feature importance analysis was performed using the following feature selection methods: S321. Model based on random forest regression model, input derived features, calculate the importance score of each derived feature through mean squared error division standard, and normalize to the [0,1] interval; S322. Use L1 regularized regression with cross-validation to automatically adjust the regularization strength, screen out features with zero coefficients, and consider the features corresponding to non-zero coefficients as important features. S323. Use elastic net regularization with cross-validation to fuse L1 and L2 regularization to perform sparsity selection of features, evaluate the contribution of each feature in the linear combination, and obtain a feature importance ranking list.

6. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, The specific process for selecting key features in S3 is as follows: S331. Based on intersection priority: Features with a weight greater than 30% in all models are retained as stable and robust features and are retained as the first-level feature set. S332. Weighted Fusion: For the remaining features after selection in S331, set fusion weights and calculate the feature fusion score S. f The following formula is used for calculation: S f =0.4·I RF +0.3·I Lasso +0.3·I ENet In the formula, I RF I Lasso I ENet These are the normalized feature importance scores for three models: Random Forest, L1 Regularized Regression with Cross-Validation, and Elastic Net Regularized Regression with Cross-Validation. The model is based on a random forest regression model. The importance score of each feature is calculated using the mean squared error criterion and normalized to the [0,1] interval. The normalized score is used as the random forest feature importance score I. RF ; L1 regularized regression with cross-validation is used to automatically adjust the regularization strength, filtering out features with zero coefficients and normalizing the absolute values ​​of their non-zero coefficients to the [0,1] interval. These normalized values ​​are then used as the Lasso feature importance score I. Lasso ; Elastic mesh regularized regression with cross-validation, incorporating L1 and L2 regularization, is used to sparsify feature selection, evaluate the contribution of each feature in the linear combination, and normalize the absolute values ​​of its regression coefficients to the [0,1] interval. These normalized values ​​are then used as the elastic mesh feature importance score I. Enet ; According to the fusion score S f The remaining features are sorted, and the top 30% of features by importance score are selected as the secondary feature set. Finally, a comprehensive key feature subset consisting of primary and secondary features is output for modeling.

7. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, S4 includes: S41. Based on the key feature variables obtained from S3, construct prediction functions for the wall thickness of the head and tail of the finished pipe along the length direction, with the prediction functions for the head and tail sections having the same form. S42. Before the raw tube enters the sizing machine unit, the industrial site collects online thickness measurement data and relevant production process parameters of the raw tube in real time. After the sizing process is completed, the operating parameters of the sizing machine are obtained synchronously. Based on this, the wall thickness prediction function coefficient is calculated and substituted into the prediction function. The continuous wall thickness distribution trend at both ends of the finished tube is simulated within the length range of the head and tail of the finished tube from 0 to the sampling section length Dmm, and the prediction and reconstruction of the wall thickness curve of the finished tube is completed. S43, Based on the target finished pipe specification thickness t spec The relationship with the prediction function, respectively for the head prediction function and tail prediction function Make a judgment and define the recommended sawing amount x. cut The recommended sawing amount is determined by finding the first point in the thickness prediction function where the thickness is less than or equal to the specification requirement. x cut min{x∈[0,L]|T(x)≤t spec } In the formula, T(x) represents the head prediction function. and tail prediction function 8. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 7, characterized in that, The head prediction function in S41 With coefficient The calculation formula is as follows: In the formula, x is the predicted position, in mm; T represents the predicted wall thickness of the finished pipe at location x, in mm. head Δv is the average thickness of the front section of the rough tube, in mm; Δv is the rolling speed gradient of the steel pipe, in m / s; T k The final rolling thermodynamic temperature, K; t spec Here is the wall thickness of the finished pipe, in mm; Δ is a constant correction term; k head T represents the slope of the rough pipe wall thickness at the head, in mm / mm; max,head ΔT represents the maximum wall thickness of the front section of the rough pipe, in mm; mid,head The value of the abrupt change in wall thickness in the middle section of the head tube is shown in mm; k tail,head The curvature of the wall thickness of the head section and the tail section of the rough pipe, in mm. -1 ;ΔT tail,head , where is the abrupt change in wall thickness of the head section and tail section, in mm; Q is the activation energy constant, in J / mol; R is the gas constant, in J / (mol·K); All regression coefficients are obtained through multiple linear regression analysis based on measured head data. Here, o takes the values ​​0, 1, 2, or 3 to distinguish the coefficients corresponding to different features. Tail prediction function With coefficient The calculation formula is as follows: In the formula, T represents the predicted wall thickness of the finished pipe at location x, in mm. tail Δv is the average thickness of the front section of the tail section of the rough tube, in mm; Δv is the rolling speed gradient of the steel pipe, in m / s; T k The final rolling thermodynamic temperature, K; t spec Here is the wall thickness of the finished pipe, in mm; Δ is a constant correction term; k tail T represents the linear slope of the wall thickness at the tail of the rough pipe, in mm / mm; max,tail ΔT represents the maximum wall thickness of the front section of the tail section of the rough tube, in mm; mid,tail The value of the abrupt change in wall thickness in the middle section of the tail section of the tube, in mm; k tail,tail The curvature of the wall thickness of the tail section of the rough tube, in mm. -1 ;ΔT tail,tail , where is the abrupt change in wall thickness at the tail end of the rough tube, in mm; Q is the activation energy constant, in J / mol; R is the gas constant, in J / (mol·K); All regression coefficients are obtained through multiple linear regression analysis based on measured tail data, where o takes the values ​​0, 1, 2 or 3 to distinguish the coefficients corresponding to different features.

9. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 1, characterized in that, In S5, based on the head prediction function and tail prediction function The determined recommended sawing amount and Combined with the process settings, the standard length range of the sub-tube [L] min ,L max ] Length L of the main tube in the group saw detected in real time i We construct sawing optimization strategies applicable to tail-aligned and non-tail-aligned scenarios using the short-scale ratio threshold ε, and output the optimal sawing scheme and material utilization evaluation results.

10. The method for online prediction of wall thickness and dynamic optimization of sawing amount after sizing of hot-rolled seamless steel pipe according to claim 9, characterized in that, The calculation process for the optimal sawing scheme and material utilization evaluation results for both tail-aligned and non-tail-aligned scenarios is as follows: S51, Tail Alignment Scenario: Based on the predicted sawing amount of all the mother tubes to be cut, a uniform head removal length S is determined. H The calculation formula is as follows: In the formula, The recommended cutting amount is predicted for the head of the i-th mother tube; N is the number of mother tubes in this group of gang saws; The number of segments (n) to be cut from the head of the target sub-tube to a fixed length (l) is: i.align -1, the length of the remaining tail segment is r i,align Furthermore, the tail-aligned cutting length S is calculated uniformly based on the tail-predicted recommended sawing amount. T The calculation formulas for the three are as follows: r i,align =R i -(n i,align -1)·l In the formula, l satisfies L min ≤l≤L max ; Predict and recommend the sawing amount for the tail of the i-th mother tube; Based on the above cutting method, an integer optimization model is constructed for the fixed length l: to maximize the material utilization rate U(l) of the tail-aligned fixed length method under the constraints that the short length rate does not exceed the short length rate threshold ε, the fixed length range is constrained, the length of the remaining tail section is non-negative, and the tail-cut section does not form a short ruler tube. The integer optimization model is constructed as follows: s.t.L min ≤l≤L max In the formula, N represents the number of main tubes in the gang saw; L i R is the original length of the i-th mother tube, in mm; i Let be the effective length of the i-th mother tube after sawing, in mm; l be the target length, a decision variable; n be the number of sub-tube segments cut from each mother tube; r i,align S is the length of the tail section of the i-th mother pipe, in mm; T To standardize the tail resection amount, mm; M short M represents the total number of short ruler tubes; total The total number of all sub-tubes; The optimized fixed length Cut length from head and tail Together with the recommended parameters for the group of gang saw cutting under the tail-aligned fixed-length sawing method, the corresponding sawing scheme, short length rate and material utilization rate evaluation results were obtained; S52, Non-tail-aligned scenarios: Based on the predicted sawing amount of all the mother tubes to be cut, a uniform head removal length S is determined. H The calculation formula is as follows: For each mother pipe P i The remaining length after removing the excessively thick sections at the head and tail is denoted as R. i The calculation formula is as follows: R i =L i -S H In the non-tail-aligned method, each mother tube is cut sequentially from the head according to a certain target length l until the remaining segment is less than l, and the last segment is the tail segment r. i,nonalign Whether it is retained as a sub-tube depends on whether it meets the minimum sub-tube length requirement L. min The number of standard length sections in each mother tube is denoted as n. i,nonalign The calculation formula is as follows: r i,nonalign =R i -n i,nonalign ·l To maximize material utilization in non-tail-aligned fixed-length cutting Simultaneously, by controlling the proportion of short ruler tubes to not exceed the short ruler ratio threshold ε and constraining the fixed ruler length range, the following integer optimization model is constructed: s.t.L min ≤l≤L max In the formula, n i,nonalign r represents the number of daughter tubes completely cut from the i-th mother tube; i δ represents the remaining length of the main pipe, in mm. i M is an indicator variable for whether to retain the tail segment sub-tube; short M represents the number of short ruler tubes. total The total number of all sub-tubes; ε is the short-scale rate threshold; The optimized fixed length with head resection length Together with the recommended parameters for this group of gang saw cutting under the non-tail-aligned fixed-length sawing method, the corresponding sawing scheme, short length rate and material utilization rate evaluation results are obtained. Based on the specific length of the remaining section at the tail of each mother tube, it is determined whether to retain it as an effective daughter tube. Finally, the sawing scheme, short length rate and material utilization rate evaluation results for this batch are obtained.

Images