Alloy cost optimization method for rolled product

Abstract

The invention discloses an alloy cost optimization method for rolled products. The alloy cost optimization method comprises the following steps: 1, determining a steel type of which the alloy cost is to be optimized according to the qualified rate of historical rolled products on a rolling production line; 2, obtaining each mechanical property prediction model and a model standard deviation based on multiple linear regression; 3, converting each element component in the steel grade to be optimized from a continuous number to a discrete number under the constraint of process requirements; 4, performing full permutation and combination on each element component with the established discrete number representation to form N groups of data after full permutation and combination; 5, sequentially carrying out mechanical property calculation based on a mechanical property prediction model on the N groups of data; 6, screening calculation results according to process requirements; and 7, calculating the total cost of the screened M groups of data under each group of data, and forming a recommendation table on the basis of forming M groups of sorts in a cost descending order or an ascending order according to a calculation result. And the method has relatively high reasonability and credibility and wide popularization value.

Description

Technical Field

[0001] The present invention belongs to the technical field of metallurgical production, and particularly relates to a method for optimizing the alloy cost of rolled products. Background Art

[0002] With the improvement of equipment and production strength in the steel industry, the market situation has become increasingly severe. Quality and cost are inevitably the fundamentals of competition in the same industry. This means that improving quality and reducing costs are direct challenges faced by steel enterprises, and continuous optimization of related products is required. And how to establish a practical and reliable solution has become a difficult point to overcome in the industry.

[0003] The invention application with the application number: CN201110036223.X discloses a "dynamic control method for the mechanical properties of hot-rolled strip steel based on a performance prediction model". Before strip steel rolling, it first retrieves chemical compositions, preset rolling process parameters, and specification parameters, then calculates the content of precipitated alloy compounds according to the preset coiling temperature in the preset rolling process parameters, and then brings all the data into the mechanical property prediction model of hot-rolled strip steel to predict the mechanical properties of the rolled strip steel. By adjusting the rolling process parameters, the predicted value of the mechanical properties of the strip steel is made to satisfy the constraint of the mechanical property target value, and then rolling is carried out according to the adjusted rolling process parameters.

[0004] The invention application with the application number: CN201210046441.6 discloses a "furnace energy-saving control method for heating furnace based on strip steel mechanical property prediction model", including controlling the furnace temperature of the heating furnace based on a process control computer, and establishing an industrial prediction model for the mechanical properties of hot-rolled strip steel; before the continuous casting billet enters the heating furnace, first predict the mechanical properties of the rolled strip steel according to the chemical composition and preset process parameters; if the predicted value of the mechanical properties is greater than its target value, consider optimizing the strip steel furnace outlet process temperature; on the premise of ensuring the mechanical properties of the strip steel and not violating other necessary constraints, optimize the control of the strip steel furnace outlet process temperature by reducing the furnace outlet temperature to reduce the gas consumption of the heating furnace, and realize the reduction of the gas consumption per unit product energy consumption of the heating furnace.

[0005] The invention application with the application number: CN201410415017.3 discloses a "finish rolling energy-saving control method based on mechanical property prediction and rolling energy consumption model", including the following steps: collecting the chemical composition, rolling process parameters, and mechanical property target value of the strip steel, checking the input parameters of the mechanical property prediction model, calling the rolling energy consumption calculation process, calculating the total finish rolling energy consumption using the rolling energy consumption model, combining the mechanical property prediction model and the rolling energy consumption model to optimize the finish rolling exit temperature; comparing the total finish rolling energy consumption before and after optimization, if the reduction amplitude of the optimized energy consumption > the energy consumption optimization threshold, modify the process temperature condition in the rolling process parameters, and carry out continuous casting billet rolling. Summary of the Invention

[0006] The purpose of the present invention is to provide a practical and highly reliable method for optimizing the alloy cost of rolled products.

[0007] To achieve the above technical objectives, the present invention provides a method for optimizing the alloy cost of rolled products, and its technical solution is as follows:

[0008] A method for optimizing the alloy cost of rolled products includes the following steps:

[0009] S1: Determine the steel grades with alloy cost to be optimized according to the qualification rate of historical rolled products on the rolling production line;

[0010] S2: Obtain each mechanical property prediction model based on multiple linear regression;

[0011] S3: Under the constraints of process requirements, convert each elemental component in the steel grade to be optimized from a continuous number to a discrete number;

[0012] S4: Perform full permutation and combination on each elemental component represented by discrete numbers to form N groups of data after full permutation and combination;

[0013] S5: Perform mechanical property calculations for the N groups of data in sequence based on the mechanical property prediction model;

[0014] S6: Screen the calculation results of each group of data according to process requirements;

[0015] S7: Calculate the total cost for each of the M groups of data after screening, and form a recommendation list based on the M groups of sorted data in descending or ascending order of cost according to the calculation results.

[0016] Furthermore,

[0017] The recommendation list formed based on the M groups of sorted data in descending or ascending order of cost in step S7 can be directly generated according to the M groups of sorted data, or formed by combining the M groups of sorted data with the predicted values of each mechanical property prediction model.

[0018] Furthermore,

[0019] In step S2, the model standard deviation is also obtained based on multiple linear regression;

[0020] Accordingly, the screening of the calculation results in step S6 is completed based on the standard score idea considering the influence of the fluctuation of the specific values of each element on the accuracy of the mechanical property prediction model.

[0021] Furthermore,

[0022] Under the condition of considering the influence of the fluctuation of the specific numerical values of the elements on the accuracy of the mechanical property prediction model, the calculation results are screened based on the standard score idea. This is achieved by first compensating and correcting the model standard deviation obtained by multiple linear regression for each array based on the fluctuation of the specific numerical values of the elements, and then screening the calculation results according to the compensated and corrected model standard deviation in combination with the standard score idea.

[0023] The specific steps are as follows:

[0024] S61: Determine the fluctuation amplitude of each element in each group of data;

[0025] S62: Calculate the influence degree of the fluctuation caused by each element on the accuracy of the mechanical property prediction model at the current value of the element according to the fluctuation amplitude of each element;

[0026] S63: Sum up the influence degrees of the fluctuations caused by each element in each group of data on the accuracy of the mechanical property prediction model to determine the influence degree of the fluctuation of each group of data on the accuracy of the mechanical property prediction model;

[0027] S64: Based on the influence degree of the fluctuation of each group of data on the accuracy of the mechanical property prediction model, perform compensation and correction calculations on the model standard deviation obtained by multiple linear regression;

[0028] S65: Based on the compensated and corrected model standard deviation, screen the calculation results of each group of data according to the standard score idea.

[0029] Furthermore,

[0030] In step S62, the calculation of the influence degree of the fluctuation caused by each element on the accuracy of the mechanical property prediction model at the current value of the element is specifically carried out according to the following formula:

[0031]

[0032] In the above formula,

[0033] Y1: The influence degree of the fluctuation caused by the current element at the current value on the accuracy of the mechanical property prediction model;

[0034] K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current element at the current value;

[0035] A: The fluctuation amplitude of the current element.

[0036] Furthermore,

[0037] The sensitivity coefficient K of the accuracy of the mechanical property prediction model to the current element at the current value is determined according to the following formula:

[0038]

[0039] In the above formula,

[0040] K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current element under the current value;

[0041] Xi_max: The maximum value of the current i-th element;

[0042] P_Y Xi_max : The predicted value of the mechanical property prediction model corresponding to the current i-th element taking the maximum value;

[0043] Xi_aim: The target value of the current i-th element;

[0044] P_Y Xi_aim : The predicted value of the mechanical property prediction model corresponding to the current i-th element taking the target value.

[0045] Furthermore,

[0046] The compensation and correction calculation for the model standard deviation obtained based on multiple linear regression in step S64 is specifically carried out according to the following formula:

[0047]

[0048] STD_r: The compensated and corrected model standard deviation under the current array;

[0049] STD: The model standard deviation obtained based on multiple linear regression;

[0050] Y: The fluctuation influence degree of the current array on the accuracy of the mechanical property prediction model.

[0051] Furthermore,

[0052] Step S65 is specifically as follows:

[0053] S651: Determine the compliance rate of each mechanical property prediction model based on the standard score idea;

[0054] S652: Sort the compliance rates of each mechanical property prediction model under each group of data, select the smallest value of the compliance rate, and compare this value with the set threshold; if this value is greater than or equal to the set threshold, then perform screening, otherwise reject.

[0055] Furthermore,

[0056] Step S651 is specifically carried out according to the following formula:

[0057] Qr_Y i = min[(P_Y i -Y i_min) / (α×Y i _STD_r),(Y i _max - P_Y i ) / (α×Y i _STD_r)],

[0058] In the above formula,

[0059] Qr_Y i : The compliance rate of the current i-th mechanical property prediction model;

[0060] P_Y i : The predicted value of the current i-th mechanical property prediction model;

[0061] Y i _min: The set minimum value of the current i-th mechanical property;

[0062] Y i _max: The set maximum value of the current i-th mechanical property;

[0063] Y i _STD_r: Model standard deviation;

[0064] α: Multiple of the standard deviation.

[0065] Furthermore,

[0066] The multiple α of the standard deviation is taken to be greater than or equal to 3.

[0067] Furthermore,

[0068] Steps S1 to S7 are all based on L4.

[0069] Furthermore,

[0070] Each mechanical property prediction model in step S2 includes: yield strength prediction model, tensile strength prediction model, fracture elongation prediction model, and impact energy prediction model.

[0071] The alloy cost optimization method for a rolled product of the present invention has high rationality and credibility, and has broad popularization value on the basis of being practical and feasible. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic diagram of the implementation steps of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0073] Next, the alloy cost optimization method for a rolled product of the present invention will be further specifically described according to the accompanying drawings of the specification and the detailed implementation manners.

[0074] To fully understand this technical solution, the following elaborates on the technical setup ideas, principles, and processes of this technical solution:

[0075] To provide a practical, convenient, and highly reliable cost optimization solution, this technical solution first selects the objects to be optimized based on the qualification rates of historical rolled products; then for the steel grades to be optimized, according to process requirements, the components are screened and eliminated based on the constraints of mechanical properties; the costs of each value under the constraint of mechanical properties are calculated for the remaining components after elimination, and then based on the calculation results, a cost optimization recommendation table is formed. See Figure 1 , and the specific steps are as follows:

[0076] S1: Determine the steel grades with alloy costs to be optimized based on the qualification rates of historical rolled products on the rolling production line;

[0077] S2: Obtain each mechanical property prediction model and model standard deviation based on multiple linear regression;

[0078] S3: Under the constraint of process requirements, convert each elemental component in the steel grade to be optimized from a continuous number to a discrete number; the conversion of each elemental component here can be carried out according to the set step size for each. It should be noted that: the step size mentioned here and the step size mentioned in the following embodiments both refer to the data interval length for converting continuous numbers to discrete numbers, rather than the iteration step size in the optimization process.

[0079] S4: Perform a full permutation and combination of each elemental component represented by discrete numbers to form N groups of data after the full permutation and combination;

[0080] S5: Perform mechanical property calculations for the N groups of data in sequence based on the mechanical property prediction model;

[0081] S6: Screen the calculation results of each group of data according to process requirements;

[0082] S7: Calculate the total cost for each of the M groups of data after screening, and form a recommendation table based on the M groups of sorted data in descending or ascending order of cost according to the calculation results; here M is less than or equal to N.

[0083] Among the above, the objects to be optimized (i.e., the steel grades to be optimized) are selected based on the qualification rates of historical rolled products, at least the qualification rates of rolled products in the most recent year; and the objects to be optimized can be selected according to the highest qualification rate, or can be selected based on a qualification rate of 95% as the benchmark, or can also be selected according to the top three qualification rates. The specific rules can be determined according to the actual situation.

[0084] The mechanical property prediction model in the above step S2 at least includes: a yield strength prediction model, a tensile strength prediction model, a fracture elongation prediction model, and an impact work prediction model.

[0085] After establishing the above-mentioned cost optimization solution based on the mechanical property prediction model, it is also considered that since the mechanical property prediction model itself is based on big data, the standard deviation used to measure the model accuracy of the prediction model based on big data will fluctuate with the fluctuation of specific values. In order to eliminate the influence of data fluctuation on the model accuracy and make the data array selected based on the mechanical property prediction model more real and reliable, this technical solution also completes the screening of the calculation results in step S6 based on the standard score idea considering the influence of the fluctuation of specific values of each element on the accuracy of the mechanical property prediction model. The basic idea is: first, compensate and correct the model standard deviation obtained by multiple linear regression under each data array based on the fluctuation of specific element values, and then screen the calculation results according to the compensated and corrected model standard deviation combined with the standard score idea. The specific steps are as follows:

[0086] S61: Determine the fluctuation amplitude of each element in each group of data; it should be noted that: the fluctuation amplitude here is the same as the meaning of another term "half interval" appearing in the text, and the two meanings are equivalent.

[0087] S62: Calculate the influence degree of the fluctuation of each element on the accuracy of the mechanical property prediction model at the current value of the element according to the fluctuation amplitude of each element within N groups of data;

[0088] S63: Sum up the influence degrees of the fluctuations of each element within each group of data on the accuracy of the mechanical property prediction model to determine the influence degree of each group of data on the accuracy of the mechanical property prediction model;

[0089] S64: Perform compensation and correction calculations on the model standard deviation obtained by multiple linear regression based on the influence degree of the fluctuation of each group of data on the accuracy of the mechanical property prediction model;

[0090] S65: Screen the calculation results of each group of data according to the standard score idea based on the compensated and corrected model standard deviation.

[0091] The calculation of the influence degree of the fluctuation of each element on the accuracy of the mechanical property prediction model at the current value of the element in the above step S62 is specifically carried out according to the following formula:

[0092]

[0093] In the above formula,

[0094] Y1: The fluctuation influence degree caused by the current element under the current value on the accuracy of the mechanical property prediction model;

[0095] K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current element under the current value;

[0096] A: The fluctuation amplitude of the current element.

[0097] The sensitivity coefficient k of the accuracy of the mechanical property prediction model to the current element under the current value here is determined according to the following formula:

[0098]

[0099] In the above formula,

[0100] K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current element under the current value;

[0101] Xi_max: The maximum value of the current i-th element;

[0102] P_Y Xi_max : The predicted value of the mechanical property prediction model corresponding to when the current i-th element takes the maximum value;

[0103] Xi_aim: The target value of the current i-th element;

[0104] P_Y Xi_aim : The predicted value of the mechanical property prediction model corresponding to when the current i-th element takes the target value.

[0105] The compensation and correction calculation for the model standard deviation obtained based on multiple linear regression in the above step S64 is specifically carried out according to the following formula:

[0106]

[0107] STD_r: The compensated and corrected model standard deviation under the current array;

[0108] STD: The model standard deviation obtained based on multiple linear regression;

[0109] Y: The fluctuation influence degree caused by the current array on the accuracy of the mechanical property prediction model.

[0110] And the above step S65 is specifically as follows:

[0111] S651: Determine the compliance rate of each mechanical property prediction model based on the idea of standard scores;

[0112] S652: Sort the compliance rates of the mechanical property prediction models for each set of data, select the smallest value of the compliance rate, and compare this value with a set threshold; if this value is greater than or equal to the set threshold, perform screening, otherwise eliminate.

[0113] Step S651 specifically proceeds according to the following formula:

[0114] Qr_Y i = min[(P_Y i - Y i _min) / (α × Y i _STD_r),(Y i _max - P_Y i ) / (α × Y i _STD_r)],

[0115] In the above formula,

[0116] Qr_Y i : The compliance rate of the current i-th mechanical property prediction model;

[0117] P_Y i : The predicted value of the current i-th mechanical property prediction model;

[0118] Y i _min: The minimum value of the current i-th mechanical property prediction model;

[0119] Y i _max: The maximum value of the current i-th mechanical property prediction model;

[0120] Y i _STD_r: The model standard deviation;

[0121] α: The multiple of the standard deviation. The multiple α of the standard deviation is taken to be greater than or equal to 3.

[0122] The above steps S1 to S7 are all based on L4. At the initial stage of cost optimization calculation, the various parameters and numerical requirements of the process requirements corresponding to the steel grades to be optimized are entered into the corresponding storage units of L4 to facilitate the real-time reading and calling of data for subsequent operations. These parameters and numerical values include the range interval values of various steel grade elements, the maximum and minimum values of various mechanical properties, and various target values of process requirements.

[0123] Embodiment

[0124] To better understand the above technical solution, the following takes the hot rolling production line as an example for specific illustration.

[0125] The hot rolling product performance prediction model selected in this embodiment is as follows:

[0126] A centralized description of the following tables is as follows:

[0127] Table 1 Main component and process step length setting table;

[0128] Table 2 Main specification dimension step length setting table;

[0129] Table 3 Unit price calculation of main alloy element costs;

[0130] Table 4 Statistical table of the qualification rates of the properties of various steel grades in a certain hot rolling production line of a certain steel plant in the past year;

[0131] Table 5 Main component design of steel grade ST2;

[0132] Table 6 Main hot rolling process parameter design of steel grade ST2;

[0133] Table 7 Semi-interval of main process parameters of steel grade ST2;

[0134] Table 8 Main mechanical property requirements of steel grade ST2;

[0135] Table 9 Alloy optimization virtual slab information of steel grade ST2;

[0136] Table 10 Performance prediction results of each virtual slab of steel grade ST2;

[0137] Table 11 Sensitivity coefficients of each optimization parameter of steel grade ST2 to each property;

[0138] Table 12 Performance perturbations of each optimization parameter of steel grade ST2 to each property;

[0139] Table 13 Standard deviation of performance after compensation of each index of steel grade ST2;

[0140] Table 14 Alloy optimization results of steel grade ST2;

[0141] Table 15 Recommended alloy element optimization plan for steel grade ST2.

[0142] Optimization calculation step lengths of each component, process, and specification set according to the experience of technical personnel.

[0143] Table 1 Main component and process step length setting table

[0144] Parameter Accuracy Step size Number of calculations C 4 0.005 5 Si 3 0.05 5 Mn 3 0.1 5 P 4 0.05 5 S 4 0.001 5 N 4 0.001 5 H 6 0.0001 5 Al 4 0.01 5 O 4 0.001 5 V 4 0.01 5 Nb 4 0.01 5 Ti 4 0.01 5 Cu 3 0.1 5 Ca 4 0.001 5 Mo 3 0.005 5 Cr 3 0.05 5 Ni 3 0.05 5 B 4 0.0005 5 W 4 0.001 5 Zr 4 0.001 5 As 4 0.001 5 Sn 4 0.001 5 DT 1 10 5 RT 1 10 5 FT 1 10 5 CT 1 10 5 ...

[0145] Table 2 Main specification dimension step length setting table

[0146]

[0147]

[0148] Set the unit price calculation of different elements according to the market price.

[0149] Table 3 Unit Price Calculation Table of Main Alloying Elements

[0150] Alloying element Unit cost SI 1400 MN 230 CR 720 TI 1200 NI 2720 MO 4800 V 28000 NB 130 AL 160 CU 670 B 500 W 4000 CA 170

[0151] The performance qualification rates of the products produced by a certain hot rolling line of a certain steel plant in the recent year are statistically shown as follows.

[0152] Table 4 Statistical Table of Performance Qualification Rates of Various Steel Grades of a Certain Hot Rolling Line of a Certain Steel Plant in the Recent Year

[0153] Steel grade type Performance qualification rate in the past year ST1 93.82％ ST2 97.71％ ST3 96.51％ ... ...

[0154] According to the statistical results of the performance qualification rates, the steel grade type ST2 with the highest qualification rate is selected as the object for alloy cost optimization.

[0155] The steelmaking composition design of the selected steel grade ST2 is as follows:

[0156] Table 5 Main Composition Design of Steel Grade ST2

[0157]

[0158]

[0159] The hot rolling process parameters design of the selected steel grade ST2 is as follows:

[0160] Table 6 Main Hot Rolling Process Parameters Design of Steel Grade ST2

[0161] Process Minimum value Maximum value Target value DT * * 1210 RT * * 1010 FT 870 890 880 CT 630 650 640

[0162] Table 7 Semi-Interval of Main Hot Rolling Parameters of Steel Grade ST2

[0163] Composition Half interval C * SI 0.075 MN 0.075 P * S * AL 0.005 B * N * V 0.005 NB 0.005 TI 0.005 CU * NI * MO * CR * CA * H * W * ZR * O * AS * SN * DT * RT * FT 10 CT 10

[0164] The performance requirement information of the selected steel grade ST2 is as follows:

[0165] Table 8 Main Mechanical Property Requirements of Steel Grade ST2

[0166]

[0167] According to the business requirements, the composition parameters to be optimized are selected as: Si, Mn, Al, V, NB, TI; when the composition parameters change, the corresponding rolling process should also be adjusted appropriately, and the finish rolling temperature and coiling temperature are selected as the optimization parameters according to the business requirements.

[0168] When constructing the virtual slab unit, the non-optimized components take the design target values, and C, P, and S are selected as the design target values.

[0169] Parameter Target value C 0.14 P 0 S 0

[0170] Based on the maximum value, minimum value, and step size information, the optimized component information forms an optimized component information array. The optimized component array is as follows:

[0171] The optimized array of SI is: [0.2, 0.25, 0.3, 0.35];

[0172] The optimized array of MN is: [1.4, 1.5, 1.55];

[0173] The optimized array of AL is: [0.025, 0.035];

[0174] The optimized array of V is: [0.025, 0.035];

[0175] The optimized array of Nb is: [0.025, 0.035];

[0176] The optimized array of Ti is: [0.01, 0.02];

[0177] For the optimized parameters of the hot rolling process parameters, an optimized process information array needs to be formed based on the maximum value, minimum value, and step size information. For other non-optimized process parameters, the design target values are taken. The tapping temperature and rough rolling temperature are set to the target values:

[0178] Parameter Target value DT 1210 RT 1010

[0179] The optimized array of FT is: [870, 880, 890];

[0180] The optimized array of CT is: [630, 640, 650];

[0181] The specification dimensions form a specification information array based on the maximum value, minimum value, and step size information;

[0182] The optimized array of Thick is: [4, 5, 6, 7, 8, 9, 10, 12, 14, 16, 18, 20, 21].

[0183] Combining the target values of non-optimized elements, the optimized element information array, the target values of non-optimized processes, the optimized process information array, the specification information array, etc., a virtual slab unit information array is obtained.

[0184] Table 9 Optimization Virtual Slab Information of Steel Grade ST2 Alloy

[0185]

[0186]

[0187] Using the hot rolling product performance prediction model to predict the mechanical properties of the virtual slab unit information array, the predicted values and standard deviations of each index are obtained.

[0188] Table 10 Prediction Results of the Performance of Each Virtual Slab of Steel Grade ST2

[0189] Serial number Si Mn Al V ... P_Rm P_Re P_El P_Rm_STD P_Re_STD P_El_STD ... <![CDATA[a1]]> 0.2 1.4 0.025 0.025 ... 590.3 503.8 32.3 14.4 20.1 2.9 <![CDATA[a2]]> 0.2 1.4 0.035 0.025 ... 585.7 497.1 33.1 14.5 20 3 <![CDATA[a3]]> 0.2 1.5 0.025 0.025 ... 581.9 491.7 33.7 14.6 19.9 3.1 ...

[0190] Considering that the simultaneous change of different optimization parameters will cause performance disturbance to the prediction performance, it is necessary to compensate for the fluctuation of the performance accuracy caused by the parameter change.

[0191] Calculate the sensitivity coefficient Kxi=(P_Y(Xi_Max,Xi+1,,,)-P_Y(Xi_Aim,Xi+1,,,)) / (Xi_Max-Xi_Aim)) (i.e., the sensitivity coefficient K of the accuracy of the mechanical property prediction model for the current element at the current value) when the performance accuracy changes with different parameter changes.

[0192] Table 11 Sensitivity Coefficients of Each Optimization Parameter of Steel Grade ST2 to Each Performance

[0193] K_RM K_RE K_EL Si 116.25 70.6 -18.1 Mn 56.65 56.65 -8.73 Al ... ... ... V ... ... ... Nb ... ... ... Ti ... ... ... FT ... ... ... CT -0.136 -0.114 0.031

[0194] Combined with the semi-interval Xi_HSpace of each parameter, calculate the performance accuracy fluctuation Bd_Y = Kxi * Kxi * (Xi_HSpace / 3) * (Xi_HSpace / 3) (i.e., the fluctuation influence degree Y1 of the current element at the current value on the accuracy of the mechanical property prediction model. Although it has an extra square in form compared with Y1, the essence of the whole calculation process is the same).

[0195] Table 12 Performance Disturbances of Each Optimization Parameter of Steel Grade ST2 to Each Performance

[0196] BD_RM BD_RE BD_EL Si 8.45 3.12 0.2 Mn 2.01 2.01 0.05 Al ... ... ... V ... ... ... Nb ... ... ... Ti ... ... ... FT ... ... ... CT 0.21 0.14 0.01

[0197] Finally, when obtaining the optimized calculation, the performance accuracy fluctuation BDCoff_Y = Bd1 + Bd2,,, + Bdi (i.e., the above step S63) caused by the simultaneous change of each optimization parameter:

[0198] Table 12 Performance Disturbances Caused by the Common Change of Each Optimization Parameter of Steel Grade ST2

[0199] BDCoff_RM BDCoff_RE BDCoff_EL BDCoff 47.57 47.11 1.54

[0200] Compensate for the performance fluctuation caused by the parameter change to obtain the actual performance standard deviation:

[0201] The standard deviation Yi_STD of the corresponding index = sqrt(P_Yi_STD^2 + BDCoff);

[0202] Table 13 Performance Standard Deviations of Each Index of Steel Grade ST2 After Compensation

[0203] Serial number Si ... P_Rm_STD P_Re_STD P_El_STD Rm_STD RE_STD EL_STD ... <![CDATA[a1]]> 0.2 ... 14.4 20.1 2.9 15.97 21.24 3.15 <![CDATA[a2]]> 0.2 ... 14.5 20 3 16.06 21.14 3.25 <![CDATA[a3]]> 0.2 ... 14.6 19.9 3.1 16.15 21.05 3.34 ...

[0204] Calculate the compliance rate Qr_Yi of each prediction index according to the predicted value, maximum value, and minimum value of the mechanical properties: Qr_Yi = min[(P_Yi - Yi_min) / (3 * Yi_STD), (Y_max - P_Yi) / (3 * Yi_STD)];

[0205] For a single virtual slab unit, take the minimum value of the compliance rates of each mechanical property index min(Qr_Y1, Qr_Y2,..., Qr_Yi) as the comprehensive compliance rate Qr_Com of the virtual slab unit. Screen the combination of virtual slab information that meets the performance requirements according to the comprehensive compliance rate being greater than the given threshold Qr_lim = 0.8. At the same time, the compliance rate Qr_Com here also participates in the formation of the recommended solutions in the following.

[0206] Table 14 Optimization Results of Steel Grade ST2 Alloy

[0207] Serial number Si Mn Al V Nb Ti Thick FT CT Qr_Rm Qr_Re Qr_El Qr_Com <![CDATA[b1]]> 0.2 1.4 0.025 0.025 0.025 0.02 4 870 630 0.911 0.943 1.031 0.911 <![CDATA[b2]]> 0.2 1.4 0.035 0.025 0.025 0.02 4 870 630 0.896 0.938 0.958 0.896 <![CDATA[b3]]> 0.2 1.5 0.025 0.025 0.025 0.02 4 870 630 0.962 0.971 1.233 0.962 ... <![CDATA[b i > 0.2 1.5 0.025 0.035 0.025 0.02 5 870 640 0.989 0.957 1.191 0.957 ... <![CDATA[b j > 0.2 1.5 0.025 0.035 0.035 0.02 6 880 640 0.965 0.859 1.104 0.859 ...

[0208] For the screened combined data, calculate the alloy cost of each combination according to the unit price of alloying elements, sort them from low to high in terms of cost, and provide the final optimized solution for the alloy cost of the steel grade to guide technicians in optimizing the steel grade. Or, sort them from low to high in terms of cost, and then combine with Qr_Com (here, combining Qr_Com follows the idea of combining the predicted values of each mechanical property prediction model) to provide the final optimized solution for the alloy cost of the steel grade to guide technicians in optimizing the steel grade; specifically, when the costs are equal, form the recommended priority order from high to low in terms of the compliance rate.

[0209] Table 15 Recommended Solutions for Optimizing Alloying Elements of Steel Grade ST2

[0210] Recommended order Original cost Optimized cost SI Mn Al V Nb Ti Thick FT CT Qr_Com 1 1467.90 1333.25 0.2 1.4 0.025 0.025 0.025 0.02 4 880 640 0.975 2 1467.90 1333.25 0.2 1.4 0.025 0.025 0.025 0.02 4 870 640 0.968 3 1467.90 1333.25 0.2 1.4 0.025 0.025 0.025 0.02 5 880 640 0.961 4 1467.90 1333.25 0.2 1.4 0.025 0.025 0.025 0.02 6 880 640 0.956 ...

Claims

1. A method for optimizing the alloy cost of a rolled product, characterized in that It includes the following steps: S1: Determine the steel grades with alloy cost to be optimized according to the qualified rate of historical rolled products on the rolling production line; S2: Obtain each mechanical property prediction model based on multiple linear regression; S3: Under the constraints of process requirements, convert the elemental components in the steel grades to be optimized from continuous numbers to discrete numbers; S4: Perform full permutation and combination on the elemental components represented by discrete numbers to form N groups of data after full permutation and combination; S5: For the N groups of data, perform mechanical property calculations based on the mechanical property prediction model in sequence; S6: Screen the calculation results of each group of data according to process requirements; S7: Calculate the total cost under each group of data for the M groups of screened data, and form a recommendation list based on the M groups of sorted data in descending or ascending order of cost according to the calculation results.

2. The method for optimizing the alloy cost of a rolled product according to claim 1, characterized in that: In step S2, the model standard deviation is also obtained based on multiple linear regression; Accordingly, the screening of the calculation results in step S6 is completed based on the standard score idea considering the influence of the fluctuations of the specific values of each element on the accuracy of the mechanical property prediction model.

3. The method for optimizing the alloy cost of a rolled product according to claim 2, characterized in that: For screening the calculation results based on the standard score idea considering the influence of the fluctuations of the specific values of the elements on the accuracy of the mechanical property prediction model, it is completed by first compensating and correcting the model standard deviation obtained based on multiple linear regression under each array based on the fluctuations of the specific values of the elements, and then screening the calculation results based on the compensated and corrected model standard deviation in combination with the standard score idea; The specific steps are as follows: S61: Determine the fluctuation amplitude of each element in each group of data; S62: Calculate the fluctuation influence degree caused by the current value of each element on the accuracy of the mechanical property prediction model for each element in the N groups of data according to the fluctuation amplitude of each element; S63: Perform summation calculation on the fluctuation influence degrees caused by each element in each group of data on the accuracy of the mechanical property prediction model to determine the fluctuation influence degree of each group of data on the accuracy of the mechanical property prediction model; S64: Perform compensation and correction calculation on the model standard deviation obtained based on multiple linear regression based on the fluctuation influence degree of each group of data on the accuracy of the mechanical property prediction model; S65: Screen the calculation results of each group of data based on the compensated and corrected model standard deviation according to the standard score idea.

4. The method for optimizing the alloy cost of a rolled product according to claim 3, characterized in that: The calculation of the fluctuation influence degree caused by the current value of each element on the accuracy of the mechanical property prediction model in step S62 is specifically carried out according to the following formula: In the above formula, Y1: The fluctuation influence degree caused by the current value of the current element on the accuracy of the mechanical property prediction model; K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current value of the current element; A: The fluctuation amplitude of the current element.

5. The method for optimizing the alloy cost of a rolled product according to claim 4, characterized in that: The sensitivity coefficient K of the accuracy of the mechanical property prediction model to the current value of the current element is determined according to the following formula: In the above formula, K: The sensitivity coefficient of the accuracy of the mechanical property prediction model to the current value of the current element; Xi_max: The maximum value of the current i-th element; P_Y Xi_max : The predicted value of the mechanical property prediction model corresponding to the maximum value of the current i-th element; Xi_aim: The target value of the current i-th element; P_Y Xi_aim : The predicted value of the mechanical property prediction model corresponding to the current i-th element taking the target value.

6. The alloy cost optimization method for a rolled product according to claim 3, characterized in that: The compensation and correction calculation of the model standard deviation obtained based on multiple linear regression in step S64 is specifically carried out according to the following formula: STD_r: The compensated and corrected model standard deviation under the current array; STD: The model standard deviation obtained based on multiple linear regression; Y: The fluctuation influence degree of the current array on the accuracy of the mechanical property prediction model.

7. The alloy cost optimization method for a rolled product according to claim 3, characterized in that: Step S65 is specifically as follows: S651: Determine the compliance rate of each mechanical property prediction model based on the idea of standard score; S652: Sort the compliance rates of each mechanical property prediction model under each group of data, select the smallest value of the compliance rate, and compare this value with the set threshold; If this value is greater than or equal to the set threshold, then perform screening, otherwise reject.

8. The alloy cost optimization method for a rolled product according to claim 7, characterized in that: Step S651 is specifically carried out according to the following formula: Qr_Y i = min[(P_Y i - Y i _min) / (α × Y i _STD_r), (Y i _max - P_Y i ) / (α × Y i _STD_r)], In the above formula, Qr_Y i : The compliance rate of the current ith mechanical property prediction model; P_Y i : The predicted value of the current i-th mechanical property prediction model; Y i _min: the set minimum value of the current ith mechanical property; Y i _max: the set maximum value of the current ith mechanical property; Y i _STD_r: Model standard deviation; α: The multiple of the standard deviation.

9. The alloy cost optimization method for a rolled product according to claim 8, characterized in that: The multiple α of the standard deviation is taken as greater than or equal to 3.

10. The alloy cost optimization method for a rolled product according to claim 1, characterized in that: Steps S1 to S7 are all based on L4.

11. The alloy cost optimization method for a rolled product according to claim 1, characterized in that: Each mechanical property prediction model in step S2 includes: yield strength prediction model, tensile strength prediction model, fracture elongation prediction model and impact energy prediction model.

12. The alloy cost optimization method for a rolled product according to claim 1, characterized in that: The recommendation list formed in step S7 is formed on the basis of sorting M groups in descending or ascending order of cost according to the calculation results, and can be directly generated according to the sorting of M groups, or formed according to the sorting of M groups combined with the predicted values of each mechanical property prediction model.

Images