Orthogonal experiment full-data model calculation method

Abstract

The invention discloses an orthogonal experiment full-data model calculation method. Comprising the steps of setting an orthogonal experiment full-data model, solving an unbiased reference quantity V of the orthogonal full-data model, solving an average value xN in the orthogonal full-data model under an N-factor x level condition, solving a deviation degree delta xN of the orthogonal full-data model, solving a data value corresponding to any experimental point of a comprehensive experiment, solving V and xN in calculation of the orthogonal full-data model and the like. Through mathematical model algorithm argumentation and fitting calculation, complete data of comprehensive experiments can be derived without adding any experiment point on the basis of original orthogonal experiment design, hidden optimal experiment conditions can be extracted from fitting results, and optimization condition results better than orthogonal optimization conditions are obtained.

Description

Technical Field

[0001] The invention belongs to the field of scientific experiments, and particularly relates to a method for calculating an orthogonal experiment full data model. Background Art

[0002] Orthogonal experiment is a common design method for studying multi-factor and multi-level experiments. It selects some representative points from the comprehensive experiment based on orthogonality to conduct experiments and obtain experimental conclusions close to those of the comprehensive experiment. Therefore, orthogonal experimental design is a fast and efficient means of experimental optimization. The disadvantage of orthogonal experiment is that only some representative experimental points are selected to obtain experimental data, and the optimal conditions screened out are only orthogonal optimal conditions, while there are very likely to be hidden optimal conditions in the comprehensive experiment, but there is no way to obtain them. Therefore, the results of orthogonal experiment are not the optimal solution of the comprehensive experiment. If you want to obtain the optimal conditions of the comprehensive experiment, you must conduct a comprehensive experiment, but the amount of experiments will increase greatly. Therefore, it is urgent to further study and solve the calculation method of the orthogonal full data model of orthogonal experiment. Summary of the invention

[0003] The purpose of the present invention is to provide a method for calculating the full data model of an orthogonal experiment, which can derive and calculate the complete data of a comprehensive experiment on the basis of the original orthogonal experimental design without adding any experimental points, and the algorithm is simple and the calculated value is accurate, providing important support for obtaining the best conditions for scientific research and engineering optimization applications.

[0004] The technical solution of the present invention is: a method for calculating the full data model of an orthogonal experiment, comprising the following steps:

[0005] Step 1: Set up the full data model of orthogonal experiment;

[0006] Step 2: Solve the unbiased reference quantity V of the orthogonal full data model;

[0007] Step 3: Solve the average value xN under the x-level condition of N factors in the orthogonal full data model;

[0008] Step 4: Solve the deviation ΔxN of the orthogonal full data model;

[0009] Step 5: Solve the data value corresponding to any experimental point of the comprehensive experiment;

[0010] Step 6: Solve V and xN in the orthogonal full data model calculation.

[0011] Furthermore, the method of setting the orthogonal experiment full data model in step 1 includes:

[0012] Suppose there is a complete orthogonal experiment with n factors and m levels, and its corresponding comprehensive experiment contains m n independent experimental groups, i.e., m n Experimental values;

[0013] For a complete orthogonal experimental design with n factors and m levels, there are n factors A, B, C, ..., N, and any factor N contains I, II, III, ..., M, and m levels. Therefore, the value of the corresponding combination (A i B j C k …N s =a ijk…s ) can be expressed as the test result under the combination of A factor i level, B factor j level, C factor k level, ..., N factor s level; where the intermediate function i, j, k, ..., s ≤ m;

[0014] Furthermore, the method for solving the unbiased reference quantity V of the orthogonal full data model in step 2 includes:

[0015] In a complete orthogonal experiment with n factors and m levels, the unbiased benchmark V is defined as the m corresponding to the full experiment. n The average value of the measured values ​​of the independent experimental groups; V represents the average state value finally obtained by considering the influence of all levels of all factors. This value serves as the benchmark for all subsequent calculations of this model. The mathematical expression is as follows:

[0016]

[0017] Furthermore, the method of solving the mean value xN under the condition of N factors x levels in the orthogonal full data model in step 3 includes:

[0018] xN represents the average of all test results at N factor x level (x≤m), which can be expressed as:

[0019]

[0020] Furthermore, the method for solving the orthogonal full data model deviation ΔxN in step 4 includes:

[0021] ΔxN represents the deviation from xN to V, which can be expressed as:

[0022] ΔxN=V-xN (Formula 3)

[0023] Furthermore, the method of solving the data value of any experimental point corresponding to the comprehensive experiment in step 5 includes:

[0024] In the above complete orthogonal experiment with n factors and m levels, the difference between the test value of any experimental point in the comprehensive experiment and the unbiased reference quantity V comes from the superposition of the influence effects brought about by the changes of the corresponding factors and levels under the experimental conditions. Therefore, in the comprehensive orthogonal experiment, the data value of any experimental point can be calculated by the unbiased reference quantity V and the deviation degree under the conditions of the corresponding factors and corresponding levels, which is expressed as follows:

[0025] Res(A i B j C k …N s )=a ijk…s =V-(ΔiA+ΔjB+ΔkC+…+ΔsN) (Formula 4)

[0026] Theoretically, the unbiased benchmark V is the average value of the comprehensive experimental data points of n factors and m levels, and xN represents the average value of all experimental points under N factors and x levels. However, due to the presence of unknown data in Formula 1 and Formula 2, the model calculation cannot be performed according to Formula 1 and Formula 2.

[0027] Furthermore, the method of solving V and xN in the orthogonal full data model calculation in step 6 includes:

[0028] In the orthogonal full data model calculation, the unbiased reference quantity V is directly equal to the mean of all orthogonal measured points, which can be represented by Ave(al). The mean xN of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When calculating, the mean of the orthogonal measured points under the N factor x level can be calculated together.

[0029] V = Ave(al) (Formula 5)

[0030] Similarly:

[0031] xN=Ave(xN) (Formula 6)

[0032] Finally, multiple iterative calculations are performed according to Formula 4 to obtain all data values ​​corresponding to the comprehensive experiment.

[0033] The present invention, through mathematical model algorithm demonstration and fitting calculation, can derive complete data of a comprehensive experiment on the basis of the original orthogonal experimental design without adding any experimental points, and can extract hidden optimal experimental conditions from the fitting results to obtain optimization condition results that are better than orthogonal optimization conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the present invention;

[0035] Figure 2 It is an improvement and innovation of the full data model of orthogonal experiment compared with the traditional orthogonal design;

[0036] Figure 3 Design cube plot for 3 factors and 3 levels. DETAILED DESCRIPTION

[0037] For a traditional orthogonal design with n factors and m levels, L(T)m n According to the orthogonal table, T measured values ​​need to be measured, and the above T measured values ​​are the known data; for a comprehensive experimental design with n factors and m levels, a total of m n experimental data points, of which T experimental values ​​have been determined by orthogonal experiments, which is equivalent to m n -T experimental values ​​have not been determined. Through the model algorithm of the present invention, the above m can be obtained without additional experiments. n -T unknown experimental values, the improvement and innovation of the orthogonal experiment full data model compared with the traditional orthogonal experiment model Figure 2 shown.

[0038] Embodiment 1: A method for calculating an orthogonal experimental full data model comprises the following steps:

[0039] Step 1: Set up the full data model of orthogonal experiment;

[0040] Step 2: Solve the unbiased reference quantity V of the orthogonal full data model;

[0041] Step 3: Solve the average value xN under the x-level condition of N factors in the orthogonal full data model;

[0042] Step 4: Solve the deviation ΔxN of the orthogonal full data model;

[0043] Step 5: Solve the data value corresponding to any experimental point of the comprehensive experiment;

[0044] Step 6: Solve V and xN in the orthogonal full data model calculation.

[0045] Embodiment 2: Based on Embodiment 1, the method of setting the orthogonal experiment full data model in step 1 includes:

[0046] Suppose there is a complete orthogonal experiment with n factors and m levels, and its corresponding comprehensive experiment contains m n independent experimental groups, i.e., m n Experimental values;

[0047] For a complete orthogonal experimental design with n factors and m levels, there are n factors A, B, C, ..., N, and any factor N contains I, II, III, ..., M, and m levels. Therefore, the value of the corresponding combination (A i B j C k …N s=a ijk…s ) can be expressed as the test result under the combination conditions of A factor i level, B factor j level, C factor k level, …, N factor s level; among which, the intermediate function i,j,k,…,s≤m.

[0048] Embodiment 3: Based on Embodiments 1 and 2, the method for solving the unbiased reference quantity V of the orthogonal full data model in step 2 includes:

[0049] In a complete orthogonal experiment with n factors and m levels, the unbiased benchmark V is defined as the m corresponding to the full experiment. n The average value of the measured values ​​of the independent experimental groups; V represents the average state value finally obtained by considering the influence of all levels of all factors. This value serves as the benchmark for all subsequent calculations of this model. The mathematical expression is as follows:

[0050]

[0051] Embodiment 4: Based on Embodiments 1 to 3, the method of solving the average value xN under the condition of N factors x levels in the orthogonal full data model in step 3 includes:

[0052] xN represents the average of all test results at N factor x level (x≤m), which can be expressed as:

[0053]

[0054] Embodiment 5: Based on Embodiments 1 to 4, the method for solving the orthogonal full data model deviation ΔxN in step 4 includes:

[0055] ΔxN represents the deviation from xN to V, which can be expressed as:

[0056] ΔxN=V-xN (Formula 9)

[0057] Embodiment 6: Based on Embodiments 1 to 5, the method for solving the data value of any experimental point corresponding to the comprehensive experiment in step 5 includes:

[0058] In the above complete orthogonal experiment with n factors and m levels, the difference between the test value of any experimental point in the comprehensive experiment and the unbiased reference quantity V comes from the superposition of the influence effects brought about by the changes of the corresponding factors and levels under the experimental conditions. Therefore, in the comprehensive orthogonal experiment, the data value of any experimental point can be calculated by the unbiased reference quantity V and the deviation degree under the conditions of the corresponding factors and corresponding levels, which is expressed as follows:

[0059] Res(A i B j C k …N s )=aijk…s =V-(ΔiA+ΔjB+ΔkC+…+ΔsN) (Formula 10)

[0060] Theoretically, the unbiased reference quantity V is the average value of comprehensive experimental data points with n factors and m levels, and xN represents the average value of all experimental points with N factors and x levels. However, due to the presence of unknown data in Formula 7 and Formula 8, model calculations cannot be performed based on Formula 7 and Formula 8.

[0061] Embodiment 7: Based on Embodiments 1 to 6, the method of solving V and xN in the orthogonal full data model calculation in step 6 includes:

[0062] In the orthogonal full data model calculation, the unbiased reference quantity V is directly equal to the mean of all orthogonal measured points, which can be represented by Ave(al). The mean xN of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When calculating, the mean of the orthogonal measured points under the N factor x level can be calculated together.

[0063] V = Ave(al) (Formula 11)

[0064] Similarly:

[0065] xN=Ave(xN) (Formula 12)

[0066] Finally, multiple iterative calculations are performed according to Formula 4 to obtain all data values ​​corresponding to the comprehensive experiment.

[0067] Example 8: A method for calculating a full data model of a 3-factor 3-level orthogonal experiment, a 3-factor 3-level orthogonal design as follows Figure 3 As shown, the following steps are included:

[0068] Specifically, in L(9)3 3 In an orthogonal design, the unbiased reference quantity is defined as:

[0069]

[0070] xN represents the average value of all test results at N factor x level (x≤m), which can be expressed as:

[0071]

[0072] Specifically, in L(9)3 3 In orthogonal design, IA represents the average of all test results at level 1 of factor A, which can be expressed as:

[0073]

[0074] That is Figure 3 The sum of the shaded area data;

[0075] Similarly, we can get:

[0076]

[0077] ΔxN represents the deviation from xN to V, which can be expressed as:

[0078] ΔxN=V-xN (Formula 16)

[0079] Specifically, in L(9)3 3 In orthogonal design, ΔIA represents the deviation from IA to V, which can be expressed as:

[0080] ΔIA=V-IA (Formula 17)

[0081] Similarly, we can get:

[0082] ΔIIA=V-IIA, ΔIIIA=V-IIIA

[0083] ΔIB=V-IB, ΔIIB=V-IIB, ΔIIIB=V-IIIB

[0084] ΔIC=V-IC, ΔIIC=V-IIC, ΔIIIC=V-IIIC

[0085] The difference between the test value of any experimental point and the unbiased reference quantity V in the above complete orthogonal experiment with n factors and m levels comes from the superposition of the influence effects brought about by the changes in the corresponding factors and levels under the experimental conditions. Therefore, the data values ​​of the experimental points in any complete orthogonal experiment with n factors and m levels can be calculated by the unbiased reference quantity V and the deviation between the corresponding factors and the corresponding levels, which can be expressed as:

[0086] Res(A i B j C k …N s )=a ijk…s =V-(ΔiA+ΔjB+ΔkC+…+ΔsN) (Formula 18)

[0087] Specifically, in L(9)3 3 In orthogonal design, the data value of any experimental point can be expressed as:

[0088] A i B j C k =aijk=V-(ΔiA+ΔjB+ΔkC) (Formula 19)

[0089] The unbiased reference quantity V is the average value of the comprehensive experimental data points of n factors and m levels, and xN represents the average value of all experimental points under N factors and x levels, but only a few representative orthogonal design points are known; for example, in a complete orthogonal experiment with 6 factors and 5 levels, the comprehensive experiment contains 15,625 data points, but the corresponding orthogonal table only requires the measurement of 25 representative data points; therefore, in theory, V and xN cannot be calculated according to Formula 1 and Formula 2 in this embodiment, and Formula 8 and Formula 9 are required for calculation. Due to the "evenly dispersed, neatly comparable" characteristics of orthogonal design, the mean of the corresponding representative orthogonal data points is the same as the mean of the comprehensive experimental data points, and this conclusion has been confirmed by data verification;

[0090] In the orthogonal experimental design, the mean V of all orthogonal measured points can be represented by Ave(al). When the software is calculating, the orthogonal measured points can be simply averaged. The mean of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When the software is calculating, the orthogonal measured points under the condition of N factor x level can be simply averaged together.

[0091] but:

[0092] V = Ave(all) (Formula 20)

[0093] Similarly:

[0094] xN=Ave(xN) (Formula 21)

[0095] Finally, according to Formula 18 in this embodiment, multiple iterative calculations are performed to obtain all data values ​​corresponding to the comprehensive experiment.

[0096] Specifically, in L(9)3 3 In orthogonal design, the unbiased reference quantity V is Figure 3 The average value of 27 comprehensive experimental data points, but only 9 orthogonal design points are known (① to ⑨). Due to the characteristics of orthogonal design "evenly dispersed, neatly comparable", the mean of 27 comprehensive experimental data points and 9 representative orthogonal measured data points is the same, and the calculation of V can be expressed as:

[0097]

[0098] Specifically, in L(9)3 3 In orthogonal design, similar calculation process of V, IA, IIA, IIIA, IB, IIB, IIIB, IC, IIC, IIIC can all be calculated from representative measured data points (Appendix Figure 3 In ①-⑨), the process is as follows:

[0099]

[0100] Specifically, in L(9)3 3 In an orthogonal design, according to Formula 19, any unmeasured data value in the corresponding comprehensive experiment can be calculated as follows:

[0101] Res(A1B1C1)=V-(ΔIA+ΔIB+ΔIC)

[0102] Res(A1B2C3)=V-(ΔIA+ΔIIB+ΔIIIC)

[0103] ······

[0104] Res(A3B3C3)=V-(ΔIIIA+ΔIIIB+ΔIIIC)

[0105] Embodiment 9: A method for calculating a full data model of a 4-factor 3-level orthogonal experiment, comprising the following steps:

[0106] Specifically, in L(9)3 4 In orthogonal design, the unbiased reference quantity V is defined as:

[0107]

[0108] Specifically, in L(9)3 4 In orthogonal design, IA represents the average of all test results at level 1 of factor A, which can be expressed as:

[0109]

[0110] Similarly, we can get:

[0111]

[0112] Specifically, in L(9)3 4 In orthogonal design, ΔIA represents the deviation from IA to V, which can be expressed as formula 3:

[0113] ΔIA=V-IA (Formula 25)

[0114] Similarly, we can get:

[0115] ΔIIA=V-IIA, ΔIIIA=V-IIIA

[0116] ΔIB=V-IB, ΔIIB=V-IIB, ΔIIIB=V-IIIB

[0117] ΔIC=V-IC, ΔIIC=V-IIC, ΔIIIC=V-IIIC

[0118] ΔID=V-ID, ΔIID=V-IID, ΔIIID=V-IIID

[0119] Specifically, in L(9)3 4 In orthogonal design, the data value of any experimental point can be expressed as:

[0120] A i B j C k D l =aijkl=V-(ΔiA+ΔjB+ΔkC++ΔlD) (Formula 26)

[0121] In actual model calculations, the mean V of all orthogonal measured points can be represented by Ave(al). When the software is calculating, the orthogonal measured points can be simply averaged. The mean of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When the software is calculating, the orthogonal measured points under the condition of N factor x level can be simply averaged together.

[0122] but:

[0123] V = Ave(al) (Formula 27)

[0124] Similarly:

[0125] xN=Ave(xN) (Formula 28)

[0126] Specifically, in L(9)3 4 In an orthogonal design, according to Formula 4, any unmeasured data value in the corresponding comprehensive experiment can be calculated as follows:

[0127] Res(A1B1C1D1)=V-(ΔIA+ΔIB+ΔIC+ΔID)

[0128] Res(A1B2C3D3)=V-(ΔIA+ΔIIB+ΔIIIC+ΔIIID)

[0129] ······

[0130] Res(A3B3C3D3)=V-(ΔIIIA+ΔIIIB+ΔIIIC+ΔIIID)

[0131] Embodiment 10: A method for calculating a full data model of a 5-factor 4-level orthogonal experiment, comprising the following steps:

[0132] Specifically, in L(16)4 5 In orthogonal design, the unbiased reference quantity V is defined as:

[0133]

[0134] Specifically, in L(16)4 5In orthogonal design, IA represents the average of all test results at level 1 of factor A, which can be expressed as:

[0135]

[0136] Similarly, we can get:

[0137]

[0138]

[0139] Specifically, in L(9)4 5 In orthogonal design, ΔIA represents the deviation from IA to V, which can be expressed as:

[0140] ΔIA=V-IA (Formula 31)

[0141] Similarly, we can get:

[0142] ΔIIA=V-IIA, ΔIIIA=V-IIIA, ΔIVA=V-IVA

[0143] ΔIIB=V-IB, ΔIIB=V-IIB, ΔIIIB=V-IIIB, ΔIVB=V-IVB

[0144] ΔIC=V-IC, ΔIIC=V-IIC, ΔIIIC=V-IIIC, ΔIVC=V-IVC

[0145] ΔID=V-ID, ΔIID=V-IID, ΔIIID=V-IIID, ΔIVD=V-IVD

[0146] ΔIE=V-IE, ΔIIE=V-IIE, ΔIIIE=V-IIIE, ΔIVE=V-IVE

[0147] Specifically, in L(16)4 5 In orthogonal design, the data value of any experimental point can be expressed as:

[0148] A i B j C k D l E p =aijklp=V-(ΔiA+ΔjB+ΔkC++ΔID+ΔpE) (Formula 32)

[0149] In actual model calculation, the mean V of all orthogonal measured points can be represented by Ave(al). In software calculation, the orthogonal measured points can be simply averaged. The mean of all orthogonal measured points under N factor x level conditions can be represented by Ave(xN). In software calculation, the orthogonal measured points under N factor x level conditions can be averaged together.

[0150] but:

[0151] V = Ave(al) (Formula 33)

[0152] Similarly:

[0153] xN=Ave(xN) (Formula 34)

[0154] Specifically, in L(16)4 5 In an orthogonal design, according to formula 32, any unmeasured data value in the corresponding comprehensive experiment can be calculated as follows:

[0155] Res(A1B1C1D1E1)=V-(ΔIA+ΔIB+ΔIC+ΔID+ΔIE)

[0156] Res(A1B2C3D4E4)=V-(ΔIA+ΔIIB+ΔIIIC+ΔIVD+ΔIVE)

[0157] ······

[0158] Res(A4B4C4D4E4)=V-(ΔIVA+ΔIVB+ΔIVC+ΔIVD+ΔIVE)

[0159] Embodiment 5: A method for calculating a full data model of a 6-factor 5-level orthogonal experiment, comprising the following steps:

[0160] Specifically, in L(25)5 6 In orthogonal design, the unbiased reference quantity V is defined as:

[0161]

[0162] Specifically, in L(25)5 6 In orthogonal design, IA represents the average of all test results at level 1 of factor A, which can be expressed as:

[0163]

[0164] Similarly, we can get:

[0165]

[0166]

[0167] Specifically, in L(25)5 6 In orthogonal design, ΔIA represents the deviation from IA to V, which can be expressed as:

[0168] ΔIA=V-IA (Formula 37)

[0169] Similarly, we can get:

[0170] ΔIIA=V-IIA, ΔIIIA=V-IIIA, ΔIVA=V-IVA, ΔVA=V-VA,

[0171] ΔIB=V-IB, ΔIIB=V-IIB, ΔIIIB=V-IIIB, ΔIVB=V-IVB, ΔVB=V-VB,

[0172] ΔIC=V-IC, ΔIIC=V-IIC, ΔIIIC=V-IIC, ΔIIC=V-IVC, ΔVC=V-VC,

[0173] ΔID=V-ID, ΔIID=V-IID, ΔIIID=V-IIID, ΔIVD=V-IVD, ΔVD=V-VD,

[0174] ΔIE=V-IE, ΔIIE=V-IIE, ΔIIIE=V-IIIE, ΔIVE=V-IVE, ΔVE=V-VE,

[0175] ΔIF=V-IF, ΔIIF=V-IIF, ΔIIIF=V-IIIF, ΔIVF=V-IVF, ΔVF=V-VF,

[0176] Specifically, in L(25)5 6 In orthogonal design, the data value of any experimental point can be expressed as:

[0177] A i B j C k D l E p F q =aijklpq=V-(ΔiA+ΔjB+ΔkC++ΔlD+ΔpE+ΔqF) (Formula 38)

[0178] In actual model calculations, the mean V of all orthogonal measured points can be represented by Ave(al). When the software is calculating, the orthogonal measured points can be simply averaged. The mean of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When the software is calculating, the orthogonal measured points under the condition of N factor x level can be simply averaged together.

[0179] but

[0180] V = Ave(al) (Formula 39)

[0181] Similarly:

[0182] xN=Ave(xN)(Formula 40)

[0183] Specifically, in L(25)5 6 In an orthogonal design, according to formula 38, any unmeasured data value in the corresponding comprehensive experiment can be calculated as follows:

[0184] Res(A1B1C1D1E1F1)=V-(ΔIA+ΔIB+ΔIC+ΔID+ΔIE+ΔIF)

[0185] Res(A1B2C3D4E5F5)=V-(ΔIA+ΔIIB+ΔIIIC+ΔIVD+ΔVE+ΔVF)

[0186] ······

[0187] Res(A5B5C5D5E5F5)=V-(ΔVA+ΔVB+ΔVC+ΔVD+ΔVE+ΔVF)

[0188] The protection of the present invention is not limited to the above embodiments, and the above formulas are applicable to multiple factors and multiple levels, and all belong to the protection scope of the present invention.

Claims

1. A method for calculating an orthogonal experiment full data model, characterized in that: The following steps are involved: Step 1: Set up the full data model of orthogonal experiment; Step 2: Solve the unbiased reference quantity V of the orthogonal full data model; Step 3: Solve the average value xN under the x-level condition of N factors in the orthogonal full data model; Step 4: Solve the deviation ΔxN of the orthogonal full data model; Step 5: Solve the data value corresponding to any experimental point of the comprehensive experiment; Step 6: Solve V and xN in the orthogonal full data model calculation.

2. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method of setting the orthogonal experiment full data model in step 1 includes: Suppose there is a complete orthogonal experiment with n factors and m levels, and its corresponding comprehensive experiment contains m n independent experimental groups, i.e., m n Experimental values; For a complete orthogonal experimental design with n factors and m levels, there are n factors A, B, C, ..., N, and any factor N contains I, II, III, ..., M, and m levels. Therefore, the value of the corresponding combination (A i B j C k …N s =α ijk…s ) can be expressed as the test result under the combination conditions of A factor i level, B factor j level, C factor k level, …, N factor s level; among which, the intermediate function i, j, k, …, s ≤ m.

3. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method for solving the unbiased reference quantity V of the orthogonal full data model in step 2 includes: In a complete orthogonal experiment with n factors and m levels, the unbiased benchmark V is defined as the m corresponding to the full experiment. n The average value of the measured values ​​of the independent experimental groups; V represents the average state value finally obtained by considering the influence of all levels of all factors. This value serves as the benchmark for all subsequent calculations of this model. The mathematical expression is as follows:

4. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method of solving the mean value xN under the condition of N factors x levels in the orthogonal full data model in step 3 includes: xN represents the average of all test results at N factor x level (x≤m), which can be expressed as:

5. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method for solving the orthogonal full data model deviation ΔxN in step 4 includes: ΔxN represents the deviation from xN to V, which can be expressed as: ΔxN=V-xN (Formula 3).

6. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method for solving the data value of any experimental point corresponding to the comprehensive experiment in step 5 includes: In the above complete orthogonal experiment with n factors and m levels, the difference between the test value of any experimental point in the comprehensive experiment and the unbiased reference quantity V comes from the superposition of the influence effects brought about by the changes of the corresponding factors and levels under the experimental conditions. Therefore, in the comprehensive orthogonal experiment, the data value of any experimental point can be calculated by the unbiased reference quantity V and the deviation degree under the conditions of the corresponding factors and corresponding levels, which is expressed as follows: Res(A i B j C k …N s ) = a ijk...s = V - (ΔiA + ΔjB + ΔkC + … + ΔsN) (Formula 4) Theoretically, the unbiased benchmark V is the average of the comprehensive experimental data points of n factors and m levels, and xN represents the average of all experimental points under N factors and x levels. However, due to the existence of unknown data in Formula 1 and Formula 2, the model calculation cannot be performed based on Formula 1 and Formula 2.

7. The orthogonal experiment full data model calculation method according to claim 1, characterized in that: The method for solving V and xN in the orthogonal full data model calculation in step 6 includes: In the orthogonal full data model calculation, the unbiased reference quantity V is directly equal to the mean of all orthogonal measured points, which can be represented by Ave(all). The mean xN of all orthogonal measured points under the condition of N factor x level can be represented by Ave(xN). When calculating, the orthogonal measured points under the N factor x level can be averaged together. V = Ave(all) (Formula 5) Similarly: xN=Ave(xN) (Formula 6) Finally, multiple iterative calculations are performed according to Formula 4 to obtain all data values ​​corresponding to the comprehensive experiment.

Images