CN106709670A - Sorting method for accessory suppliers of automatic control valve enterprises - Google Patents

Abstract

The invention discloses a method for sorting parts suppliers of automatic control valve enterprises. Firstly, construct an evaluation index system for the ranking of parts suppliers of automatic control valve enterprises, in which the evaluation index is divided into three categories: quantitative, comprehensive qualitative and direct qualitative; obtain the initial value of each evaluation index, and use the membership degree method to calculate the tendency of the initial value of each evaluation index The classification model based on the support vector machine is used to screen the candidate suppliers to reduce the number; the weighting method is used to deal with the uncertainty of the importance of the evaluation index, and the two-stage evidence theory model is used to identify the screened candidate suppliers. Realize the sorting of parts suppliers of automatic control valve enterprises. The invention is easy to apply and easy to implement, and can solve the problems of uncertainty and incompleteness of evaluation index values and unfixed importance of evaluation index values in the ranking of parts suppliers of automatic control valve enterprises.

Description

A sorting method for parts suppliers of automatic control valve enterprises

technical field

The invention relates to the field of supplier ordering in supply chain optimization, in particular to a method for ordering suppliers of parts suppliers of automatic control valve enterprises.

Background technique

As an important part of raw material procurement decision-making of automatic control valve enterprises, the ranking of parts suppliers is a key link in building supply chain relationships. Reasonable selection of parts suppliers will directly affect the production costs and competitiveness of automatic control valve enterprises. Due to the uncertainty of the market, the asymmetry of information between cooperative enterprises and other random factors, automatic control valve enterprises face great risks in the process of sorting parts suppliers. On the hot issue of component supplier ranking, the existing representative methods include fuzzy multi-objective integer programming method, Bayesian network, fuzzy TOPSIS and multi-choice objective programming integration method, etc. The main problems in these existing methods are: the evaluation indicators involved in the ranking of parts suppliers of automatic control valve enterprises are very extensive. Due to the limitations of decision makers’ understanding of the parts supplier market, the obtained evaluation index values are often uncertain. and incompleteness; in addition, the importance of each evaluation indicator is obviously not fixed when the needs are different or the preferences of decision makers are different.

Contents of the invention

In order to solve the problems of the uncertainty and incompleteness of evaluation index values and the unfixed importance of evaluation indexes in the ranking of parts suppliers of automatic control valve enterprises, the present invention provides a sorting method of parts suppliers of automatic control valve enterprises. Firstly, construct an evaluation index system for the ranking of parts suppliers of automatic control valve enterprises, in which the evaluation index is divided into three categories: quantitative, comprehensive qualitative and direct qualitative; obtain the initial value of each evaluation index, and use the membership degree method to calculate the tendency of the initial value of each evaluation index The classification model based on the support vector machine is used to screen the candidate suppliers to reduce the number; the weighting method is used to deal with the uncertainty of the importance of the evaluation index, and the two-stage evidence theory model is used to identify the screened candidate suppliers. Realize the sorting of parts suppliers of automatic control valve enterprises.

Technical scheme of the present invention is:

The method for sorting parts suppliers of an automatic control valve enterprise is characterized in that it comprises the following steps:

Step 1: Construct the evaluation index system for the ranking of parts suppliers of automatic control valve enterprises, in which the evaluation index is divided into three categories: quantitative, comprehensive qualitative and direct qualitative.

The evaluation index system for the sorting of the automatic control valve enterprise accessories suppliers includes product competitiveness (expressed by I _ProComp ), internal competitiveness (expressed by I _InComp ), external competitiveness (expressed by I _OutComp ) and collaboration ability (expressed by I _CoopAbil Indicates) four indicators. I _InComp , I _OutComp , and I _CoopAbil are quantitative indicators, which are scored by decision makers; the index value of product competitiveness I _ProComp is difficult to determine, and is a comprehensive qualitative indicator, so it is decomposed into price level (expressed by I _Cost ), quality Level (indicated by I _Quality ), service level (indicated by I _Service ) and flexibility level (indicated by I _Flexibility ) are four sub-indices. I _Cost and I _Quality are quantitative indicators, and I _Service and I _Flexibility are direct qualitative indicators.

Step 2: After inspecting the actual situation, the decision makers give the initial values of the quantitative and direct qualitative evaluation indicators of the candidate parts suppliers. Assign definite values to the determined quantitative evaluation indicators and direct qualitative evaluation indicators, assign value intervals to relatively vague quantitative evaluation indicators, and assign null values to completely unknown evaluation indicators. The tendency degree of the initial value of each evaluation index is calculated by the method of membership degree.

For the initial value of each evaluation index, set five levels of comments: {G ₁ , G ₂ , G ₃ , G ₄ , G ₅ }={very bad, bad, medium, excellent, very good}, where G ₁ and G ₅ are the evaluation grades corresponding to the lowest limit value D ₁ and the highest limit value D ₅ respectively, then under a certain evaluation index, the evaluation index values equivalent to the evaluation grade are {D ₁ , D ₂ , D ₃ , D ₄ , D ₅ }. It should be noted here that the lower the I _Cost , the better, and the higher the other, the better. The conversion method is just the opposite.

Assume that the numerical values corresponding to each comment level are: E(G ₁ )=0, E(G ₂ )=0.25, E(G ₃ )=0.5, E(G ₄ )=0.75, E(G ₅ )=1. β _i is the membership degree of the evaluation index value to the comment level G _i , then the tendency degree P _t (A _i ) of each component supplier under the evaluation index t can be determined according to two different situations of quantitative index and direct qualitative index.

The calculation method of the tendency degree P _t (A _i ) of the quantitative index is: when the evaluation index value is a certain value a, it can be directly calculated. When the evaluation index value is the value interval [a,b], there are three cases: if D _i ≤a≤D _i+1 or D _i ≤a≤b≤D _i+1 , then P _t (A _i )=β _i ·E(G _i )+β _i+1 ·E(G _i+1 ); if D _i ≤a≤D _i+1 , D _i+1 ≤b≤D _i+2 , then P _t (A _i )=β _i ·E(G _i )+β _i+1 ·E(G _i+1 )+β _i+2 ·E(G _i+2 ); if D _i ≤a≤D _i+1 , D _j ≤b≤D _j+1 , then: P _t (A _i )=β _i ·E(G _i )+...+β _j ·E(G _j+1 ).

The calculation method of the tendency degree P _t (A _i ) of the direct qualitative index is: it can be obtained directly according to the corresponding comment grade, that is, P _t (A _i )=E(G _i ).

Step 3: Screen candidate suppliers to narrow down the number using a support vector machine based classification model.

In the case of linear separability, the basic idea of support vector machine can be described as: Assume two types of samples (x ₁ ,z ₁ ),…,(x _l ,z _l ), x∈R ^u , l is the number of samples, u is the input dimension, define the hyperplane w·x+f=0 to separate the two types of samples, and the classification result is In the formula, w is an adjustable weight vector, f is the offset of the hyperplane, and w·x represents the inner product of the vector w∈R ^u and x _i ∈R ^u . In order for the classification hyperplane to correctly classify all samples, the classification interval 2/||w|| must be maximized on both sides. For the training sample set, find the optimal value of weight w and offset b to minimize the weight cost function, that is And the constraint condition is satisfied: z _i (w· _xi +f)-1≥0, i=1, 2, . . . , l. Introducing the Lagrange multiplier ξ _i ≥ 0, i=1,2,…,l, we get The extreme point of Γ is a saddle point, take the minimum value of Γ for w and f: w=w ^* and f=f ^* , and the maximum value for ξ: ξ=ξ ^* . The optimal hyperplane can be determined by solving the quadratic programming after deriving Γ. Only samples with ξ=0 act on w ^* and determine the classification result, and such samples are defined as support vectors. ξ ^* and w ^* can be obtained explicitly, namely Select a support vector sample x _i : f ^* = z _i -w· _xi , for any input sample x, calculate the classification function Finally, according to the sign of the classification function d(x), the attribution of the sample x to be classified is determined.

Randomly extract 50 historical data of automatic control valve enterprises on the ranking of suppliers of the same type of parts as the training samples of the support vector machine classification model; Vector, determine the classification function d(x), if d(x)=w ^* x+f ^* ≥ 0, then x passes the preliminary screening; if d(x)=w ^* x+f ^* <0, then x Be eliminated.

Step 4: Use the weighting method to deal with the uncertainty of the importance of the evaluation index, use the two-stage evidence theory model to identify the candidate suppliers after screening, and realize the ranking of the parts suppliers of the automatic control valve enterprise.

Step 4.1 defines the set of screened candidate accessory suppliers as the recognition frame, that is, Θ={y ₁ ,y ₂ ,...,y _N }, where y _i (i=1,2,...,N) is The i-th screened candidate accessory supplier. All possible sets on Θ are represented by the power set 2 ^Θ . When there are N elements in Θ and each element is incompatible with each other, the number of elements in the power set 2 ^Θ of Θ is 2 ^N . Let A be a single factor or a combination of factors that affect the assembly quality, m(A) is the basic probability assignment function of A on the recognition frame Θ, which represents the degree of trust in A, satisfying m(A) can be expressed as m:2 ^Θ →[0,1], and A that satisfies m(A)>0 is called the focal element.

Step 4.2 Consider the actual needs and the degree of preference of decision makers to determine the weight coefficients of each evaluation index and sub-indicators.

Step 4.3 For the four sub-indices under I _ProComp : I _Cost , I _Quality , I _Service and I _Flexibility , calculate the weighted basic probability distribution values of all focal elements which is where l<2 ^N .

Step 4.4 For I _InComp , I _OutComp , I _CoopAbil , calculate the weighted basic probability distribution values of all focal elements The calculation method is the same as step 4.3.

Step 4.5 Assign values according to the weighted basic probabilities corresponding to the four sub-indices under I _ProComp As evidence input, evidence fusion is performed, that is, where K is the normalization constant, Solve the basic probability distribution value m _ProComp (A _i ) of the candidate parts suppliers screened under I _ProComp and Θ.

Step 4.6 Normalize the candidate parts suppliers screened under I _ProComp and the basic probability distribution value m _ProComp (A _i ) of Θ and add the weight coefficient of I _ProComp to obtain the weighted basic probability distribution value The weighted normalization method is as follows: the basic probability distribution value m _ProComp (A _i ) of the focal element Θ produced by the first fusion is regarded as the basic probability distribution value of the non-Θ focal element, and m _ProComp (Θ) = m _ProComp ( A _l ), the weighted normalization calculation formula is

Step 4.7 to As evidence input, carry out secondary evidence fusion, the fusion method is the same as step 4.4, and obtain the comprehensive basic probability distribution value m(A _i ) of the screened candidate parts suppliers.

Step 4.8 uses the identification rules based on the trust interval to sort the parts suppliers of the automatic control valve enterprise.

Calculate the trust function value Bel(A _i ) and the likelihood function value Pl(A _i ) of all screened candidate parts suppliers respectively, where the trust function value Indicates the total trust in A _i , the likelihood function value Indicates the uncertainty of A _i , where Construct the trust interval [Bel(A _i ), Pl(A _i )], and the specific identification rules are as follows: Assume that the degree to which candidate accessory supplier A _i is better than candidate accessory supplier A _j is P(A _i >A _j ), if The trust intervals of A _i and A _j are [Bel(A _i ),Pl(A _i )] and [Bel(A _j ),Pl(A _j )] respectively, then we have where P(A _i >A _j )∈[0,1]. Then, if P(A _i >A _j )>0.5, then A _i is better than A _j , recorded as If P(A _i >A _j )<0.5, then A _i is not as good as A _j , recorded as If P(A _i >A _j )＝0.5, there is no difference between A _i and A _j , recorded as A _i ～A _j ; for any A _i , A _j and A _k , if P(A _i >A _j )> 0.5 and P(A _j >A _k )>0.5, then A _i is better than A _k , recorded as So as to realize the sorting of parts suppliers by automatic control valve enterprises, and obtain the best parts suppliers.

The beneficial effects of the present invention are:

The method is simple and reasonable, and easy to realize. Taking full account of the uncertainty and incompleteness of the evaluation index value and the instability of the importance of the evaluation index in the ranking of the parts suppliers of the automatic control valve enterprise, the support vector machine classification model is used to screen the candidate suppliers to reduce the number; through the weighting method To deal with the uncertainty of the importance of the evaluation index, use the two-stage evidence theory model to identify the selected candidate suppliers, and realize the ranking of the parts suppliers of the automatic control valve enterprise, which can overcome the shortcomings of the existing supplier ranking methods.

Description of drawings

Figure 1 is a schematic flow chart of the ranking method of parts suppliers of automatic control valve enterprises.

detailed description

The present invention is described below in conjunction with specific embodiments, so that the advantages and features of the present invention can be more easily understood by those skilled in the art, so as to define the protection scope of the present invention more clearly.

Example:

When a self-control valve company sorts the suppliers of sleeve accessories, there are 20 qualified suppliers of sleeve accessories to choose from. The implementation steps are as follows:

Step 1: Construct the evaluation index system for the ranking of sleeve accessories suppliers of automatic control valve enterprises, including product competitiveness (expressed by I _ProComp ), internal competitiveness (expressed by I _InComp ), external competitiveness (expressed by I _OutComp ) and collaboration Capability (indicated by I _CoopAbil ) four indicators. I _InComp , I _OutComp , and I _CoopAbil are quantitative indicators, which are scored by decision makers; product competitiveness I _ProComp is a comprehensive qualitative indicator, which is decomposed into price level (expressed by I _Cost ), quality level (expressed by I _Quality ), Service level (indicated by I _Service ) and flexibility level (indicated by I _Flexibility ) are four sub-indices, I _Cost and I _Quality are quantitative indicators, and I _Service and I _Flexibility are direct qualitative indicators. Here, I _Cost : unit price (unit: yuan), I _Quality : mean error value (unit: mm) obtained according to the provided accessory samples.

Step 2: After inspecting the actual situation, the decision makers give the initial values of the quantitative and direct qualitative evaluation indicators of the candidate sleeve accessories suppliers, as shown in Table 1, and the empty values are represented by "/". Here, in order to briefly describe the specific implementation of the present invention so that it can be understood by those skilled in the art, only the initial values of the evaluation indicators of the first three candidate sleeve accessory suppliers are listed.

Table 1 Initial values of evaluation indicators for candidate sleeve accessories suppliers

For the quantitative indicators I _Cost , I _Quality , I _InComp , I _OutComp and I _CoopAbil , the comment levels {G ₁ , G ₂ , G ₃ , G ₄ , G ₅ }={very bad, bad, medium, good, very good} The corresponding reference index values are: G ^Cost = {100000, 10000, 1000, 100, 10}; G ^Quality = {0.05, 0.04, 0.03, 0.02, 0.01}; G ^InComp = {17, 13, 9, 5, 1 }; G ^OutComp = {1, 3, 5, 7, 9}; G ^CoopAbil = {0, 0.25, 0.5, 0.75, 1}. From this, the membership degree of the initial value of the evaluation index in Table 1 to the evaluation grade is calculated, and then the information in Table 1 is converted into the form of evaluation grade-membership degree, as shown in Table 2.

Table 2 Comment grades of the initial value of evaluation indicators for candidate sleeve accessories suppliers - membership degree form

Calculate the tendency of the initial value of each evaluation index, as shown in Table 3.

Table 3 The tendency of the initial value of the evaluation index of candidate sleeve accessories suppliers

Step 3: Randomly extract 50 historical data of the self-control valve company on the ranking of suppliers of the same type sleeve accessories as the training samples of the support vector machine classification model; then use the data in Table 3 as the input vector to determine the classification function d( x), if d(x)=w ^* ·x+f ^* ≥0, then x passes the preliminary screening; if d(x)=w ^* ·x+f ^* <0, then x is eliminated, and the candidate sleeve accessories Suppliers 1, 2 and 3 pass the screening.

Step 4: Use the weighting method to deal with the uncertainty of the importance of the evaluation index, and use the two-stage evidence theory model to identify the sleeve accessories suppliers 1, 2 and 3.

Step 4.1 defines the set of screened candidate accessory suppliers as the recognition frame, that is, Θ={y ₁ , y ₂ , y ₃ }, where y ₁ , y ₂ , y ₃ represent sleeve accessory suppliers 1 and 2 in turn and 3.

Step 4.2 Considering the actual demand and the degree of preference of decision makers, determine the weight coefficients of each evaluation index and sub-index as (ω _Cost , ω _Quality , ω _Service , ω Flexibility )=(0.7, 0.9, _0.8 , 0.9), (ω _ProComp , ω _InComp , ω _OutComp , ω _CoopAbil )=(0.8, 0.8, 0.9, 0.7).

Step 4.3 For the four sub-indices I _Cost , I _Quality , I _Service and I _Flexibility under I _ProComp , according to the inclination and weight of socket accessories suppliers 1, 2 and 3 (y ₁ , y ₂ , y ₃ ) in Table 3 Coefficients (ω _Cost , ω _Quality , ω _Service , ω _Flexibility )=(0.7, 0.9, 0.8, 0.9), calculate the weighted basic probability distribution values of all focal elements. Among them, the price level I _Cost : Quality level I _Quality : Service level I _Service : Flexibility level I _Flexibility :

Step 4.4 For I _InComp , I _OutComp , I _CoopAbil , according to the _{inclination and weight coefficients (ω InComp , ω OutComp ,} ω _CoopAbil )=(0.8, 0.9, 0.7), calculating the weighted basic probability distribution values of all focal elements. Among them, internal competitiveness I _InComp : External competitiveness I _OutComp : Collaborative Ability I _CoopAbil :

Step 4.5 Assign values according to the weighted basic probabilities corresponding to the four sub-indices under I _ProComp As evidence input, evidence fusion is carried out, and the basic probability distribution value m _ProComp (A _i ) of each focal element under I _ProComp is solved as m _ProComp (y ₁ )=0.0771; m _ProComp (y ₂ )=0.8005; m _ProComp (y ₃ )=0.0887; m _ProComp (y ₁ ,y ₂ )=0.0126; m _ProComp (y ₂ ,y ₃ )=0.0189; m _ProComp (Θ)=0.0021.

Step 4.6 Normalize the candidate parts suppliers screened under I _ProComp and the basic probability distribution value m _ProComp (A _i ) of Θ and add the weight coefficient of I _ProComp to obtain the weighted basic probability distribution value for

Step 4.7 to As evidence input, the second evidence fusion is carried out, and the comprehensive basic probability distribution value m(A _i ) is obtained, among which, m(y ₁ )=0.1601; m(y ₂ )=0.7988; m(y ₃ )=0.0165; m (y ₁ ,y ₂ )=0.0158; m(y ₂ ,y ₃ )=0.0053; m(Θ)=0.0035.

Step 4.8 uses the identification rules based on the trust interval to sort the parts suppliers of the automatic control valve enterprise.

Calculate the trust function value Bel(A _i ) and the likelihood function value Pl(A _i ) of sleeve accessories suppliers 1, 2 and 3 (y ₁ , y ₂ , y ₃ ) respectively, and construct the trust interval [Bel(A _i ), Pl(A _i )] is y ₁ : [0.1601, 0.1794]; y ₂ : [0.7988, 0.8234]; y ₃ : [0.0165, 0.0253]. According to the identification rules based on trust intervals, the trust intervals of sleeve accessories suppliers 1, 2 and 3 (y ₁ , y ₂ , y ₃ ) are compared, and the result is: Thus, the sorting of sleeve fitting suppliers by the automatic control valve enterprise is realized, and the sleeve fitting supplier 2 is the best.

Claims (1)

1. a kind of internally piloted valve enterprise fittings supply business's sort method, it is characterised in that comprise the following steps：

Step 1：Build the evaluation index system of internally piloted valve enterprise fittings supply business's sequence, wherein evaluation index be divided into it is quantitative, Aggregate qualitative and direct qualitative three class.

The evaluation index system of the internally piloted valve enterprise fittings supply business sequence (uses I comprising product competitiveness_ProCompRepresent), it is interior Portion's competitiveness (uses I_InCompRepresent), external competitive power (use I_OutCompRepresent) and collaboration capabilities (use I_CoopAbilRepresent) four refer to Mark.I_InComp、I_OutComp、I_CoopAbilIt is quantitative target, is given a mark by policymaker and obtained；Product competitiveness I_ProCompDesired value compared with Hardly possible determines, is aggregate qualitative index, therefore be broken down into price level and (use I_CostRepresent), quality level (use I_QualityRepresent), Service level (uses I_ServiceRepresent) and degree of flexibility level (use I_FlexibilityRepresent) four sub- indexs, I_Cost、I_QualityIt is quantitative Index, I_ServiceAnd I_FlexibilityIt is direct qualitative index.

Step 2：After policymaker investigates actual conditions, the initial of candidate's fittings supply amount of deciding through consultation and direct qualitative evaluation index is given Value.Pair determine qualitative assessment index and direct qualitative evaluation index assign determination value, to the qualitative assessment index of Relative Fuzzy Interval is assigned, null value is assigned to totally unknown evaluation index.Each evaluation index initial value is calculated with subordination method Tendency degree.

For the initial value of each evaluation index, 5 grades of comments are set：{G₁, G₂, G₃, G₄, G₅}={ is very bad, bad, medium, excellent, very It is excellent }, wherein G₁And G₅Respectively liminal value D₁With Up limit D₅Corresponding evaluation approach, then in a certain evaluation index Under be equivalent to evaluation approach evaluation index value be { D₁, D₂, D₃, D₄, D₅}.Here should be noted I_CostMore low more excellent, other are then got over High more excellent, conversion regime is just the opposite.

If the corresponding numerical value of each evaluation approach is respectively：E(G₁)=0, E (G₂)=0.25, E (G₃)=0.5, E (G₄)=0.75, E (G₅)=1.β_iIt is evaluation index value for evaluation approach G_iDegree of membership, then under evaluation index t each fittings supply business tendency degree P_t(A_i) can respectively be determined according to two kinds of different situations of quantitative target and direct qualitative index.

The tendency degree P of quantitative target_t(A_i) computational methods are：When evaluation index value is determination value a, can directly calculate.Work as assessment When desired value is interval [a, b], it is divided into three kinds of situations：If D_i≤a≤D_i+1Or D_i≤a≤b≤D_i+1, then P_t(A_i)=β_i· E(G_i)+β_i+1·E(G_i+1)；If D_i≤a≤D_i+1, D_i+1≤b≤D_i+2, then P_t(A_i)=β_i·E(G_i)+β_i+1·E(G_i+1)+ β_i+2·E(G_i+2)；If D_i≤a≤D_i+1, D_j≤b≤D_j+1, then：P_t(A_i)=β_i·E(G_i)+...+β_j·E(G_j+1)。

The tendency degree P of direct qualitative index_t(A_i) computational methods are：Can directly be tried to achieve according to corresponding evaluation approach, i.e. P_t(A_i) =E (G_i)。

Step 3：Candidate supplier is screened with the disaggregated model based on SVMs reduce quantity.

In the case of linear separability, the basic thought of SVMs can be described as：Assuming that two class sample (x₁,z₁),…, (x_l,z_l), x ∈ R^u, l is sample number, and u is input dimension, defines hyperplane wx+f=0 and separates this two classes sample, classification knot It is reallyW is adjustable weight vector in formula, and f is the amount of bias of hyperplane, and wx represents vectorial w ∈ R^uWith x_i∈R^uInner product.To make Optimal Separating Hyperplane correctly classify all samples, it is necessary to make the class interval 2/ of its both sides | | w | | it is maximum.For training sample set, the optimal value of weight w and skew b is found, make weights cost function minimum, i.e., And meet constraints：z_i(w·x_i+ f) -1 >=0, i=1,2 ..., l.Introduce Lagrange Multiplier ξ_i>=0, i=1,2 ..., l are obtainedThe extreme point of Γ is saddle point, takes Γ To the minimum value of w and f：W=w^*And f=f^*, and to the maximum of ξ：ξ=ξ^*.Be can determine that to solving quadratic programming after Γ derivations Optimal hyperlane.Only the sample of ξ=0 is to w^*Classification results are worked and determine, such sample is defined as supporting vector. ξ^*And w^*Can explicitly try to achieve, i.e.,Choose a supporting vector sample x_i:f^*=z_i-w·x_i, for any input Sample x, calculates classification functionFinally according to classification function d (x) Symbol determines the ownership of sample x to be sorted.

50 historical datas of the internally piloted valve enterprise in homotype fittings supply business's sequence are randomly selected, as support vector cassification The training sample of model；The tendency degree of each evaluation index initial value of this supplier sequence for then being obtained using step 2 as Input vector, determines classification function d (x), if d (x)=w^*·x+f^*>=0, then x is by preliminary screening；If d (x)=w^*·x+f^* <0, then x be eliminated.

Step 4：The not stationarity of evaluation index significance level is processed by weighting method, with double rank Evidence theory models to screening Candidate supplier afterwards is identified, and realizes the sequence of internally piloted valve enterprise fittings supply business.

The set of the candidate fittings supply business after screening is set to identification framework, i.e. Θ={ y by step 4.1₁,y₂,...,y_N, its Middle y_i(i=1,2 ..., N) is the candidate fittings supply business after i-th screening.Set power set 2 is possible on Θ^ΘCome Represent, when the unit in Θ have N number of and each element objectionable intermingling, the power set 2 of Θ^ΘElement number be 2^N.If A is shadow Single factor test or the factor combination of assembling quality are rung, m (A) is the basic probability assignment function of A on identification framework Θ, is represented to A's Degree of belief, meetsAndM (A) is represented by m:2^Θ→ [0,1], meets m (A)>0 A is referred to as Jiao Unit.

Step 4.2 considers actual demand and decisionmaker's preference degree, determines the weight coefficient of each evaluation index and sub- index.

Step 4.3 is for I_ProCompLower four sub- indexs：I_Cost、I_Quality、I_ServiceAnd I_Flexibility, calculate all burnt units Weighting basic probability assignment valueI.e.Wherein l<2^N。

Step 4.4 is for I_InComp、I_OutComp、I_CoopAbil, calculate the weighting basic probability assignment value of all burnt unitsCalculate Method is with step 4.3.

Step 4.5 is with I_ProCompLower four sub- indexs are corresponding to weight basic probability assignment value It is input into as evidence, carries out evidence fusion, i.e.,Wherein K is normal for normalization Number, hasSolve I_ProCompUnder The basic probability assignment value m of candidate's fittings supply business and Θ after screening_ProComp(A_i)。

Step 4.6 is by I_ProCompThe basic probability assignment value m of candidate's fittings supply business and Θ after lower screening_ProComp(A_i) carry out Normalize and add I_ProCompWeight coefficient, obtain weighting basic probability assignment valueWeighting normalization method is： The basic probability assignment value m of Jiao unit Θ that first time fusion is produced_ProComp(A_i) it is considered as the basic probability assignment of the burnt units of non-Θ Value, if m_ProComp(Θ)=m_ProComp(A_l), weighting normalization computing formula is

Step 4.7 withIt is input into as evidence, carries out secondary card According to fusion, fusion method solves the comprehensive basic probability assignment value m (A of the candidate fittings supply business after screening with step 4.4_i)。

Step 4.8 carries out the sequence of internally piloted valve enterprise fittings supply business with the recognition rule based on trust interval.

The belief function value Bel (A of the candidate fittings supply business after all screenings are calculated respectively_i) and likelihood function value Pl (A_i), its Middle belief function valueRepresent to A_iTotal degree of belief, likelihood function valueRepresent to A_i Uncertainty, hereConstruction trusts interval [Bel (A_i),Pl(A_i)], specific recognition rule is：Assuming that candidate's accessory Supplier A_iBetter than candidate fittings supply business A_jDegree be P (A_i>A_j), if A_iAnd A_jTrust interval be respectively [Bel (A_i), Pl(A_i)] and [Bel (A_j),Pl(A_j)], then have Wherein P (A_i>A_j)∈[0,1].So, if P (A_i>A_j)>0.5, then A_iCompare A_jIt is outstanding, it is designated asIf P (A_i>A_j)< 0.5, then A_iNot as A_jIt is outstanding, it is designated asIf P (A_i>A_j)=0.5, then A_iWith A_jThere is no difference, be designated as A_i~A_j；For Any A_i、A_jAnd A_kIf, P (A_i>A_j)>0.5 and P (A_j>A_k)>0.5, then A_iCompare A_kIt is outstanding, it is designated asSo as to realize certainly Sequence of the Kong Fa enterprises to fittings supply business, obtains optimal fittings supply business.