Conflict management in Dempster-Shafer theory (D-S theory) is a hot topic in information fusion. In this paper, a novel weighted evidence combination rule based on evidence distance and uncertainty measure is proposed. The proposed approach consists of two steps. First, the weight is determined based on the evidence distance. Then, the weight value obtained in first step is modified by taking advantage of uncertainty. Our proposed method can efficiently handle high conflicting evidences with better performance of convergence. A numerical example and an application based on sensor fusion in fault diagnosis are given to demonstrate the efficiency of our proposed method.
Doctoral Program of Higher Education of China20130191110027Research and Construction of Self-Regulated Learning Platform of Software Technology Specialty in Higher Vocational CollegesKJ1502901Research and demonstration application based on MOOCS Independent Learning Platform of Training Support Technology for Industry 4 talentsGZTG2016051. Introduction
Information fusion technology (IFT) is utilized to analyze multisource uncertain information comprehensively. Through reserving the common information, IFT can decrease indeterminacy greatly. The Dempster-Shafer theory of evidence [1] (D-S theory, also known as evidence theory or theory of belief functions) is regarded as an efficient model to fuse information in intelligent systems [2]. And Dempster’s combination rule is the most crucial instrument of D-S theory. The theory was firstly proposed by Dempster in 1967 [1] and then developed to its present form by Shafer et al. in 1976 [3]. The Dempster’s combination rule possesses several interesting mathematical properties, such as commutativity and associativity, and it plays a very significant role in evidence theory [4]. Nowadays, the evidence theory is applied widely in many fields, like supplier selection ([5, 6]), target recognition ([7–9]), decision making ([4, 10, 11]), reliability analysis ([12–15]), and so on.
Although D-S theory has a lot of advantages, there also exist some basic problems that still are not completely clarified. One of the most significant issues is that D-S theory will become invalid when using it to fuse highly conflicting evidences, and the counter-intuitive results ([10, 16, 17]) will be generated. To solve such a problem, two major methodologies are popular. One is to preprocess the bodies of evidence (BOEs) ([18–20]), and the other is to modify the combined rule ([21–23]). There are mainly three alternative combination rules belonging to the second type and they are, respectively, Dubois and Prade’s disjunctive combination rule [24], Smets’ unnormalized combination rule [25], and Yager’s combination rule [21]. These three alternatives mentioned above are examined and they all propose a general combination framework. The main work of preprocessing bodies of evidences (BOEs) includes Murphy’s simple average in [19], Yong et al.’s weighted average on the basis of distance of evidence in [26], and Han et al.’s modified weighted average in [27]. In [19], a simple averaging approach of the primitive BOEs is proposed, and in that case all BOEs are seen equally important, which is unreasonable in practice. Yong et al. [26] get a better combination result according to combining the weight average of the masses for n-1 times. The approach proposed by Han et al. [27] is a novel weighted evidence combination approach based on the evidence distance and ambiguity measure (AM), which actually modifies Yong et al.’s work [26].
In this paper, a novel weighted evidence combination rule based on evidence distance and uncertainty measure is proposed to address the combination of conflicting evidences. The numerical example and an application in fault diagnosis are given to sufficiently prove the efficiency of our proposed method.
The remaining paper is organized as follows: Section 2 starts with a brief introduction of the Dempster-Shafer theory and evidence distance; the proposed method is presented in Section 3; Sections 4 and 5 give a numerical example and an application in fault diagnosis, respectively, to prove the efficiency of our proposed approach; finally, the conclusion is made in Section 6.
2. Preliminaries
In this section, some preliminaries are briefly introduced below.
2.1. Basics of Evidence Theory
Dempster-Shafer theory of evidence (D-S theory) is used for dealing with uncertainty information as an efficient mathematical model in intelligent systems [1]. In 1967, The definition, D-S theory was proposed by Dempster [1] and then his student Shafer et al. developed this theory [3] in 1976.
([29, 30]) Let Ω be a nonempty finite set and 2Ω be the set of all subsets of Ω, denoted Ω={{θ1},{θ2},…,{θn}} and 2Ω={∅,{θ1},{θ2},…,{θn},{θ1,θ2},…,{θ1,θ2,…,θn}}.
In Dempster-Shafer theory of evidence [3], a basic probability assignment (BPA) is a mapping: 2Ω→[0,1] satisfying(1)∑E⊆ΩmE=1,m∅=0.
If m(E)>0, E is called a focal element, and the set consisting of all the focal elements is called one body of evidences (BOEs). When there are more than one independent body of evidences, Dempster’s combination rule, (2) which is a powerful and crucial tool in D-S theory, can be utilized to combine these evidences.(2)mE=∑E1,E2⊆Ω,E1∩E2=Em1E1m2E21-k,where k=∑E1∩E2=∅m1(E1)m2(E2) stands for conflict degree, also called normalization constant. What is noted is that if m⊕(∅)=1, this combination rule will make no sense. Here, we give a specific example about combination rule and show the corresponding results in Table 1.
An example of combination rule.
BOEs
m(E1)
m(E2)
m(E3)
S1
0.5
0.2
0.3
S2
0.6
0.2
0.2
Combined outcome
0.75
0.1
0.15
Example 1.
Suppose that the frame of discernment Ω={E1,E2,E3} is complete and there are two BOEs listed as follows:
S1:m1(E1)=0.5, m1(E2)=0.2, m1E3=0.3;
S2:m2(E1)=0.6, m2(E2)=0.2, m2(E3)=0.2.
In the frame of discernment Ω, there are two BOES, m1 and m2. When m1 and m2 are both reliable, to generate a new BPA, we can conjunctive rule denoted (3) [31]. When only one of them is totally reliable and we are not sure about another one, then we should apply disjunctive combination rule [31] denoted (4).(3)m⊓E=∑E1,E2⊆Ω,E1∩E2=Em1E1m2E2,∀E⊆Ω,(4)m⊔E=∑E1,E2⊆Ω,E1∪E2=Em1E1m2E2,∀E⊆Ω.
([32, 33]) Given a proposition E1∈2Ω, the belief function of E1, Bel(E1), is defined in (5), which represents the total belief that the object is in E1. The plausibility function of E1, Pl(E1), is defined in (6), which measures the total belief that can move into E1. In D-S theory, Bel(E1) and Pl(E1) are called lower bound function and upper bound function, respectively, denoted [Bel(E1),Pl(E1)]. And Bel(E1) and Pl(E1) must satisfy the following relations of (7) and (8):(5)BelE1=∑E2⊆E1mE2,(6)PlE1=∑E1∩E2≠∅mE2,(7)PlE1=1-BelE1¯,(8)PlE1≥BelE1.
For any proposition E1, its uncertainty can be represented by using its belief function and plausibility function as (9).(9)uE1=PlE1-BelE1.
Here, we give an example about belief function and plausibility function and show its results in Table 2.
An example of Bel and Pl.
Function
{E1}
{E2}
{E3}
{E1,E2}
{E1,E3}
{E2,E3}
{E1,E2,E3}
m
0.4
0
0
0.3
0
0.2
0.1
Bel
0.4
0
0
0.7
0.4
0.2
1
Pl
0.8
0.6
0.3
1
1
0.9
1
u
0.4
0.6
0.3
0.3
0.6
0.7
0
Example 2.
Assume Ω={E1,E2,E3}; a BPA is given where m({E1})=0.4, m({E1,E2})=0.3, m({E2,E3})=0.2, m({E1,E2,E3})=0.1.
But the classical Dempster combination rule is not efficient all the time. When BOEs are in high conflict, illogical results will be generated [31]. Nowadays, there are mainly two kinds of methodologies. One is to modify the combined rule, and the other is to preprocess evidences. Smets’ unnormalized combination rule [25], Dubois and Prade’s disjunctive combination rule [24], and Yager’s combination rule [21] belong to the first category. These three alternatives mentioned above are examined and they all proposed a general combination framework. To preprocess data, Murphy’s simple average in [19], Yong et al.’s weighted average [26], and Han et al.’s modified weighted average in [27] are popular. In [19], a simple averaging approach of the primitive BOEs is proposed. And in that case all BOEs are seen equally important, which is unreasonable in real life. In [26], Yong et al. can get a better combination result by combining the weight average of the masses for n-1 times. In [27], a novel weighted evidence combination approach based on the distance of evidence and AM is proposed, which is based on Yong et al.’s work [26] actually.
2.2. Evidence Distance
With D-S theory applying widely, the study about evidence distance has attracted more and more interests [34]. The dissimilarity measure of evidences can represent the lack of similarity between two BOEs. Performance evaluation [35], reliability evaluation [36], conflict evidence combination [26], target association ([37, 38]), and a lot of methods regarding evidence distance are brought up as an appropriate measure of the difference. And several definitions of distance in evidence theory are also proposed, like Jousselme et al. distance [39], Wen et al.’s cosine similarity [40], Ristic and Smets’ transferable belief model (TBM) global distance measure ([37, 38]), Sunberg and Rogers’s belief function distance metric [41], and so on. Among those definitions on distance of evidence, the most frequently used is Jousselme et al.’s distance [39].
The Jousselme et al. distance [39] is identified based on Cuzzolin’s the geometric interpretation of evidence theory [42]. The power set of the frame of discernment 2Ω is regarded as a 2N-linear space. A distance and vectors are defined with the BPA as a particular case of vectors. The Jousselme et al. distance is defined as follows: (10)dij=12mi→-mj→TDmi→-mj→mi, mj are two BPAs under the frame of discernment Ω, and D is a 2N×2N matrix. The element in D is defined as D(E1,E2)=|E1∩E2|/|E1∪E2|,E1,E2∈P(Ω); · represents cardinality. This Jousselme et al. distance satisfies all four requirements (nonnegativity, nondegeneracy, symmetry, and triangle inequality) [39] of a strict distance metric. This distance is an efficient tool to quantify the dissimilarity of two BOEs. An example of the Jousselme et al. distance is shown below.
The value inside the BOE vectors m1→ and m2→ and the distance matrix D are given by(11)m1→=0.40.20.10.3,m2→=0.50.20.10.2,D=10121301121312121231313231.
It follows that (m1→-m2→)T=-0.1000.1 and (m1→-m2→)=-0.1000.1:(12)dm1,m2=12-0.1000.110121301121312121231313231-0.1000.1=0.1125.
3. The Proposed Method
We followed the methods of Wang et al. [43]. Assume that there are in total n BOEs mi, i=1,…,k, collected; we can use (13) to precalculate these k pieces of BOE ([44–46]).(13)mWAE=∑i=1nwi×mi,where wi stands for the corresponding weight degree of each BOE mi and mWAE is the weighted average BPA of k BOEs. By means of the classical Dempster’s rule ([1, 3]) to combine mWAE for n-1 times, we can get the final combined result. But, to find an appropriate weight wi is a little difficult. Actually there are many related works including Yong et al. [26] and Han et al.’s approach [27]. In this paper, we present a new modified weighted evidence combination rule on the basis of evidence distance and a novel uncertainty measure. The flow graph of our proposed method is described in Figure 1.
The flow graph of the proposed method.
Step 1.
In this step, based on evidence distance, we can determine all weights of the primitive BOEs.
As seen from the definition of evidence distance, the less the distance between two BOEs is, the more the similarity of those two is. The similarity measure SIMij between two BOEs is defined [20] in(14)SIMmi,mj=1-dmi,mj.
After getting all the degrees of similarity between BOEs, we can construct a similarity measure matrix (Smm) [26]. The Smm will give us insight into the agreement between BOEs.(15)Smm=1S12⋯S1j⋯S1k⋮⋮⋮⋮⋮⋮Si1Si2⋯Sij⋯Sin⋮⋮⋮⋮⋮⋮Sn1Sn2⋯Snj⋯1.We define the support degree of each BOE mi(i=1,2,…,k) [26] as follows: (16)SUPmi=∑j=1,j≠ikSIMmi,mj.
Then, we can obtain the credibility degree of each BOE mi(i=1,2,…,k) based on SUP(mi) [26]:(17)CRDi=SUPmi∑j=1kSUPmj.
In fact, for ∑i=1kCRDi=1 it is evident that we can regard the credibility degree CRDi as a weight directly. So, CRDi can be used to replace wi in (13) to modify original BPAs. As mentioned in [27], the uncertainty degree can also be utilized to construct weights. A combination result can come up by using AM and modifying CRDi obtained based on evidence theory. Here, we propose another uncertainty measure to modify CRDi obtained in Step 1, and the results perform better.
Step 2.
In this step, the weight will be modified on the basis of uncertainty.
Supposing that one of some BOEs with relatively high credibility degree generated in the first step has less uncertainty degree than the others, we believe that the BOE is more credible and it should possess more weight because of its good quality. On the contrary, if a BOE has both more uncertainty degree and a low credibility degree, such a BOE is relatively incredible and even causes a wrong result perhaps. Thus, such a BOE should possess less weight.
Based on the thoughts mentioned above, the weight determined based on evidence distance will be modified through the following steps.
Step 2-1. As for each given original BOE mi(i=1,2,…,k), compute the belief function Bel(A) and the plausibility function Pl(A) of each proposition A∈2Ω using (5) and (6).
Step 2-2. We take advantage of (9) to measure uncertainty denoted u(A) for each proposition A∈2Ω. Adding up all u(A)(A∈2Ω) by using (18), we can obtain total uncertainty measure for one BOE mi(i=1,2,…,k), denoted Ui.(18)Ui=∑A∈2ΩuA.
Step 2-3. We utilize the following equation to modify the weight CRDi(i=1,2,…,k) obtained in Step 1.(19)CRDmi=CRDi×eUi.
Step 2-4. The final modified weight CRDmni(i=1,2,…,k) will be obtained after normalizing all CRDmi(i=1,2,…,k) with (20).(20)CRDmni=CRDmi∑j=1kCRDmj.
Step 2-5. The modified weighted averaged BOE denoted mWAE is obtained as(21)mWAE=∑i=1kCRDmnmi×mii=1,2,…,k.
If k pieces of evidence are supplied, we can use the classical Dempster’s rule [1] to combine mWAE for k-1 times [19]. After that, we could get a better combined outcome to make a better decision.
4. Experiments
In this section, a numerical example is illustrated to demonstrate the efficiency of our proposed method.
Example 4.
In a multisensor-based automatic target recognition system, suppose that the frame of discernment Ω={E1,E2,E3} is complete and E1 is the real target. From five different sensors, the system collects five BOEs listed as follows:
S1:m1A=0.41, m1B=0.29, m1C=0.3;
S2:m2(A)=0, m2(B)=0.9, m2(C)=0.1;
S3:m3(A)=0.58, m3(B)=0.07, m3(A,C)=0.35;
S4:m4(A)=0.55, m4(B)=0.1, m4(A,C)=0.35;
S5:m5(A)=0.6, m5(B)=0.1, m5(A,C)=0.3.
Step 1.
Through the first step of our proposed method, the credibility degree CRDi(i=1,2,…,5) will be obtained, that is, the weight of each BOE can be determined based on evidence distance. The evidence distance between BOEs and the credibility degree of each BOE are shown in Tables 3 and 4, respectively.
The evidence distance between BOEs.
Evidence distance
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
d(m1,m2)
0.5386
0.5386
0.5386
0.5386
d(m1,m3)
-
0.3495
0.3495
0.3495
d(m1,m4)
-
-
0.3257
0.3257
d(m1,m5)
-
-
-
0.3311
d(m2,m3)
-
0.8142
0.8142
0.8142
d(m2,m4)
-
-
0.7850
0.7850
d(m2,m5)
-
-
-
0.7906
d(m3,m4)
-
-
0.0300
0.0300
d(m3,m5)
-
-
-
0.0374
d(m4,m5)
-
-
-
0.0354
The weight determined by the evidence distance.
Weight
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
CRD(m1)
0.5000
0.4284
0.2829
0.2059
CRD(m2)
0.5000
0.2494
0.1365
0.0899
CRD(m3)
-
0.3222
0.2861
0.2321
CRD(m4)
-
-
0.2945
0.2368
CRD(m5)
-
-
-
0.2353
Step 2.
Through this step, we can get the modified weight.
We can calculate total uncertainty Ui for each BOE i(i=1,2,…,5) using (9) and (18). The results are listed in Table 5.
The total uncertainty for each BOE.
Item
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
U1
0
0
0
0
U2
0
0
0
0
U3
-
1.4
1.4
1.4
U4
-
-
1.4
1.4
U5
-
-
-
1.2
Then by making use of (19) and (20), the final modified weight CRDmni(i=1,2,…,5) of each BOE mi(i=1,2,…,5) can be obtained. The results are shown in Table 6.
The final modified weight for each BOE.
Item
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
CRDmn(m1)
0.5000
0.2159
0.1020
0.0691
CRDmn(m2)
0.5000
0.1257
0.0492
0.0302
CRDmn(m3)
-
0.6584
0.4183
0.3160
CRDmn(m4)
-
-
0.4305
0.3224
CRDmn(m5)
-
-
-
0.2623
Next, the mWAE can be obtained by means of replacing wi of (13) with CRDmni(i=1,2,…,5). The results are listed in Table 7.
mWAE on the basis of the novel modified weight averaging approach.
Item
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
mWAE
m(A)=0.2050
m(A)=0.4704
m(A)=0.5212
m(A)=0.5463
m(B)=0.5950
m(B)=0.2218
m(B)=0.1462
m(B)=0.1278
m(C)=0.2000
m(C)=0.0773
m(C)=0.0365
m(C)=0.0238
m(AC)=0
m(AC)=0.2305
m(AC)=0.2971
m(AC)=0.3021
Finally, we make advantage of the classical Dempster-Shafer combination rule ([1, 3]) to combine mWAE shown in Table 7 for 4 times [19]. After that, the final combined outcomes will be got and shown in Table 8 and Figure 2.
The final BPA by using the proposed method.
Item
m1, m2
m1, m2, m3
m1, m2, m3, m4
m1, m2, m3, m4, m5
Final BPA
m(A)=0.0964
m(A)=0.8923
m(A)=0.9788
m(A)=0.9916
m(B)=0.8119
m(B)=0.0293
m(B)=0.0010
m(B)=0.0001
m(C)=0.0917
m(C)=0.0455
m(C)=0.0102
m(C)=0.0026
m(AC)=0
m(AC)=0.0329
m(AC)=0.0173
m(AC)=0.0057
The final BPAs based on the proposed method.
We also make use of different combination rules to calculate the Example 4, and the results are shown in Table 9 and the comparing figures are shown in Figures 3–6.
Evidence combination outcomes based on different combination rules.
Approach
Combination outcomes
m1, m2
m1, m2, m3
m1, m2,m3,m4
m1, m2, m3, m4, m5
Dempster’s rule [1]
m(A)=0
m(A)=0
m(A)=0
m(A)=0
m(B)=0.8969
m(B)=0.6575
m(B)=0.3321
m(B)=0.1422
m(C)=0.1031
m(C)=0.3425
m(C)=0.6679
m(C)=0.8575
Murphy’s simple average [19]
m(A)=0.0964
m(A)=0.4619
m(A)=0.8362
m(A)=0.9620
m(B)=0.8119
m(B)=0.4497
m(B)=0.1147
m(B)=0.0210
m(C)=0.0917
m(C)=0.0794
m(C)=0.0410
m(C)=0.0138
m(AC)=0
m(AC)=0.0090
m(AC)=0.0081
m(AC)=0.0032
Yong et al.’s weighted average [26]
m(A)=0.0964
m(A)=0.4974
m(A)=0.9089
m(A)=0.9820
m(B)=0.8119
m(B)=0.4054
m(B)=0.0444
m(B)=0.0039
m(C)=0.0917
m(C)=0.0888
m(C)=0.0379
m(C)=0.0107
m(AC)=0
m(AC)=0.0084
m(AC)=0.0089
m(AC)=0.0034
Han et al.’s novel weighted average [27]
m(A)=0.0964
m(A)=0.5188
m(A)=0.9246
m(A)=0.9844
m(B)=0.8119
m(B)=0.3802
m(B)=0.0300
m(B)=0.0023
m(C)=0.0917
m(C)=0.0926
m(C)=0.0362
m(C)=0.0099
m(AC)=0
m(AC)=0.0084
m(AC)=0.0092
m(AC)=0.0034
Our proposed method
m(A)=0.0964
m(A)=0.8923
m(A)=0.9788
m(A)=0.9916
m(B)=0.8119
m(B)=0.0293
m(B)=0.0010
m(B)=0.0001
m(C)=0.0917
m(C)=0.0455
m(C)=0.0102
m(C)=0.0026
m(AC)=0
m(AC)=0.0329
m(AC)=0.0173
m(AC)=0.0057
The results using different combination rules when m1, m2 are given.
The results using different combination rules when m1, m2, m3 are given.
The results using different combination rules when m1, m2, m3, m4 are given.
The results using different combination rules when m1, m2, m3, m4, m5 are given.
As seen from Table 9 and Figures 3–6, when evidences are in high conflict, counter-intuitive results will be produced by using the classical Dempster’s combination rule and they do not reflect the truth. With incremental BOEs, although Murphy’s simple average [19], Yong et al.’s weighted average [26], and Han et al.’s novel weight average [27] all can give reasonable results, their results are all inferior to the outcomes of our proposed approach. Moreover, the performance of convergence of our proposed method is better than other existing methods. The main reason for these phenomena mentioned above is that, by making use of the evidence distance [39] and uncertainty measure, the final weights of bad evidences are decreased greatly (e.g., in experiment section, consider the case S1, S2, S3, S4, S5; the weight of conflicting evidence S2 falls to 0.0899 from 0.5). So its effect on the final combined outcomes will be weakened extremely. Similarly, because the final weights of good evidences are increased (e.g., in experiment section, consider the cases S1, S2, S3; the total weight of credible evidences S1 and S3 adds to 0.8743 from 0.5), the effect of these credible evidences on the final combined results is strengthened extremely. The numerical example illustrates adequately that our proposed method is efficient.
5. Application
As all know, sensor data fusion plays a crucial role in fault diagnosis. In this section, our proposed method is also applied to solve such a problem and the results show that our proposed method is as feasible and efficient as the other existing approaches, and the most important thing is that, by utilizing our proposed method, the accuracy of fault diagnosis will be improved drastically. The example in this application section is cited from [28].
Suppose a machine has three gears G1, G2, and G3, and the failure fault modes F1, F2, F3 represent that there are faults in G1, G2, and G3, respectively. The fault hypothesis set is Ω={F1,F2,F3}. Assume S1, S2, S3 are three types of sensors, respectively. The evidence set derived from different sensors is denoted as E={E1,E2,E3} and the BPAs are all listed in Table 10. Assume the sufficiency indexes of those three pieces of evidences are 1, 0.6, and 1 and the importance indexes are 1, 0.34, and 1, respectively.
The basic probability assignments in the application.
Evidence number
m({F1})
m({F2})
m({F2,F3})
m(Ω)
E1
0.6
0.1
0.1
0.2
E2
0.05
0.8
0.05
0.1
E3
0.7
0.1
0.1
0.1
According to the equation K=∑B∩C=∅m1(B)m2(C), we can easily get the conflict degree of each pair of evidences and they are k1,2=0.52, k1,3=0.26, and k2,3=0.605, respectively, which obviously shows that the second evidence E2 has conflicts highly with others. To solve such a fault diagnosis problem, the statistic sensor reliability and the dynamic sensor reliability will be considered. The statistic sensor reliability mainly depends on the technical factors, which can be evaluated by experts’ assessment or comparing the detection value with the real value in the long-term practice. The dynamic sensor reliability is influenced by the surrounding conditions changing with time, which can be measured by comparing the consistency of the outputs with other sensors that aim at the same point.
In this application section, the statistic sensor reliability rs is measured based on the evidence sufficiency index μ and evidence importance index ν in Fan and Zuo’s approach [28] and we define ris=μi×νi,j. The dynamic sensor reliability rd is measured based on the newly proposed method in this paper and we define rid=Crdmni. The comprehensive sensor reliability r=rs×rd is defined to modify the highly conflicting evidences and finally through the classical Dempster’s combination rule, we can obtain the final results with a high accuracy to address such a fault diagnosis problem. The detailed steps are shown as follows.
Step 1.
Calculate the statistic sensor reliability rs of each evidence and the results are listed in Table 11.
The statistic sensor reliability rs of each evidence.
E1
E2
E3
rs
1
0.204
1
Step 2.
Based on the newly proposed method, calculate the dynamic sensor reliability rd of each evidence and the results are listed in Table 12.
The dynamic sensor reliability rd of each evidence.
E1
E2
E3
rd
0.4701
0.1830
0.3469
Step 3.
According to the equation r=rs×rd, calculate the comprehensive sensor reliability r and then normalize. The results are shown in Table 13.
The comprehensive reliability r of each evidence.
E1
E2
E3
r
0.4701
0.0373
0.3469
rn
0.5503
0.0437
0.4060
Step 4.
The normalized comprehensive sensor reliability rn is applied to replace wi in (13) to modify BPAs, and through using the classical Dempster’s combination rule for two times, we can obtain the final results shown in Table 14.
The final results in the application.
m({F1})
m({F2})
m({F2,F3})
m(Ω)
Value
0.8899
0.0785
0.0243
0.0073
As seen from Table 14, based on our method, the belief degree of the fault F1 is 88.99%, while the fault F2 only has a belief degree of 7.39%. It is evident that mF1>mF2. Thus, through our approach, we can correctly find the fault F1; that is, Gear 1 has a fault and so our proposed method is efficient in fault diagnosis.
We also compare our proposed method with others and the comparison results are shown in Table 15. In D-S evidence theory, the belief degree of the fault F1 is 0.4519, while that of F2 is 0.5048. Due to the existence of conflicting evidence F2, D-S evidence theory comes to the counterintuitive results that m(F2)>m(F1), which will lead to a wrong decision. However, the remaining two methods can handle the conflicting evidence F2, so that they can both reach the right result. In Fan and Zuos method, mF1=81.19, while our newly proposed approach has a higher belief degree of 88.99. The main reason is that the proposed method takes into consideration not only the static reliability represented by evidence sufficiency and evidence importance, but also the dynamic reliability, the information volume of the sensor itself, measured by evidence distance and uncertainty measurement, which decreases the weight of conflicting evidence F2 and then makes F2’s influence on the final result less.
The comparison results between our proposed approach and other methods.
m({F1})
m({F2})
m({F2,F3})
m(Ω)
Dempster-Shafer evidence theory [1]
0.4519
0.5048
0.0336
0.0096
Fan and Zuo’s [28]
0.8119
0.1096
0.0526
0.0259
Ours
0.8899
0.0785
0.0243
0.0073
6. Conclusion
In this paper, a novel weighted evidence combination rule based on evidence distance ([20, 39]) and uncertainty measure is put forward. The proposed approach can address the combination of the conflicting evidences efficiently. By comparing other existing method, our proposed approach performs with the fastest convergence when managing high conflicting evidences. In addition, we apply our newly proposed method to the fault diagnosis. The results of the application sufficiently demonstrate the efficiency of our approach. The process of our proposed method is simple and with better convergence, so we believe our proposed method has a promising future.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
This work was supported by Specialized Research Fund for the Doctoral Program of Higher Education of China (SRFDP) (no. 20130191110027) and by Research and Construction of Self-Regulated Learning Platform of Software Technology Specialty in Higher Vocational Colleges (no. KJ1502901) and by Research and demonstration application based on MOOCS Independent Learning Platform of Training Support Technology for Industry 4 talents (no. GZTG201605).
DempsterA. P.Upper and lower probabilities induced by a multivalued mapping196738325339MR020700110.1214/aoms/1177698950SongY.WangX.ZhangH.A distance measure between intuitionistic fuzzy belief functions2015862882982-s2.0-8493841807510.1016/j.knosys.2015.06.011ShaferG.1976Princeton, NJ, USAPrinceton University PressMR0464340BeynonM.CurryB.MorganP.The Dempster-Shafer theory of evidence: an alternative approach to multicriteria decision modelling2000281375010.1016/s0305-0483(99)00033-x2-s2.0-0034134990DouZ.XuX.LinY.ZhouR.Application of D-S evidence fusion method in the fault detection of temperature sensor20142014610.1155/2014/3950573950572-s2.0-84901037078DengX.HuY.DengY.MahadevanS.Supplier selection using AHP methodology extended by D numbers20144115616710.1016/j.eswa.2013.07.0182-s2.0-84880825976XuX.LiS.SongX.WenC.XuD.The optimal design of industrial alarm systems based on evidence theory2016461421562-s2.0-8494681164610.1016/j.conengprac.2015.10.014ChenY.CremersA. B.CaoZ.Interactive color image segmentation via iterative evidential labeling20142012923042-s2.0-8490163494010.1016/j.inffus.2014.03.007SuhD.YookJ.A method to determine basic probability assignment in context awareness of a moving object20132013797264110.1155/2013/9726412-s2.0-84884218922ZadehL. A.A simple view of the Dempster-Shafer theory of evidence and its implication for the rule of combination19867285902-s2.0-0022733771LeungY.JiN.-N.MaJ.-H.An integrated information fusion approach based on the theory of evidence and group decision-making201314441042210.1016/j.inffus.2012.08.0022-s2.0-84887827053KhamsehS. A.SedighA. K.MoshiriB.FatehiA.Control performance assessment based on sensor fusion techniques20164914282-s2.0-8495693998710.1016/j.conengprac.2016.01.008NiuD.WeiY.ShiY.KarimiH. R.A novel evaluation model for hybrid power system based on vague set and Dempster-Shafer evidence theory201220122-s2.0-8487017450110.1155/2012/784389784389ZhouQ.ZhouH.ZhouQ.YangF.LuoL.LiT.Structural damage detection based on posteriori probability support vector machine and Dempster-Shafer evidence theory2015363683742-s2.0-8493914462710.1016/j.asoc.2015.06.057XuJ.ZhongZ.XuL.ISHM-oriented adaptive fault diagnostics for avionics based on a distributed intelligent agent system20154613228723022-s2.0-8494647302210.1080/00207721.2014.971090ZadehL. A.Review of a mathematical theory of evidence198453, article 81LefevreE.ColotO.VannoorenbergheP.Belief function combination and conflict management2002321491622-s2.0-003660884510.1016/S1566-2535(02)00053-2TorraV.NarukawaY.MiyamotoS.Modeling Decisions for Artificial Intelligence: Theory, Tools and Applications20053558Berlin, HeidelbergSpringer18Lecture Notes in Computer Science10.1007/11526018_1Zbl1121.68429MurphyC. K.Combining belief functions when evidence conflicts2000291192-s2.0-003372878110.1016/S0167-9236(99)00084-6VoorbraakF.On the justification of Dempster's rule of combination1991482171197MR110319610.1016/0004-3702(91)90060-W2-s2.0-0026122354YagerR. R.On the Dempster-Shafer framework and new combination rules198741293137MR88678610.1016/0020-0255(87)90007-7Zbl0629.680922-s2.0-0023311178YamadaK.A new combination of evidence based on compromise200815913168917082-s2.0-4344911610110.1016/j.fss.2007.11.019HanD.HanC.YangY.A modified evidence combination approach based on ambiguity measureProceedings of the 11th International Conference on Information Fusion, FUSION '08July 2008Cologne, Germany10.1109/ICIF.2008.46323672-s2.0-56749143861DuboisD.PradeH.Representation and combination of uncertainty with belief functions and possibility measures19884324426410.1111/j.1467-8640.1988.tb00279.xSmetsP.Combination of evidence in the transferable belief model199012544745810.1109/34.551042-s2.0-0025434991YongD.WenKangS.ZhenFuZ.QiL.Combining belief functions based on distance of evidence20043834894932-s2.0-454435725510.1016/j.dss.2004.04.015HanD.DengY.HanC.HouZ.Weighted evidence combination based on distance of evidence and uncertainty measure201130539640010.3724/SP.J.1010.2011.00396FanX.ZuoM. J.Fault diagnosis of machines based on D-S evidence theory. Part 1: D-S evidence theory and its improvement20062753663762-s2.0-3264443707310.1016/j.patrec.2005.08.025ZhuW.YangH.JinY.LiuB.A Method for Recognizing Fatigue Driving Based on Dempster-Shafer Theory and Fuzzy Neural Network201720172-s2.0-8501332479210.1155/2017/61910356191035JiangW.WeiB.QinX.ZhanJ.TangY.Sensor Data Fusion Based on a New Conflict Measure201620162-s2.0-8496503663410.1155/2016/57690615769061LiuW.Analyzing the degree of conflict among belief functions200617011909924MR224643210.1016/j.artint.2006.05.0022-s2.0-33745854779ZhengH.DengY.HuY.Fuzzy evidential influence diagram and its evaluation algorithm2017131284510.1016/j.knosys.2017.05.024QianJ.GuoX.DengY.A novel method for combining conflicting evidences based on information entropy20174648768882-s2.0-8500409725710.1007/s10489-016-0875-yJousselmeA.-L.MaupinP.On some properties of distances in evidence theory, Belief, Brest, FranceFixsenD.MahlerR. P. S.The modified Dempster-Shafer approach to classification1997271961042-s2.0-003087109910.1109/3468.553228GuoH.ShiW.DengY.Evaluating sensor reliability in classification problems based on evidence theory20063659709812-s2.0-3374939727910.1109/TSMCB.2006.872269RisticB.SmetsP.The TBM global distance measure for the association of uncertain combat ID declarations20067327628410.1016/j.inffus.2005.04.0042-s2.0-33744944512RisticB.SmetsP.Global cost of assignment in the TBM framework for association of uncertain ID reports20071143033092-s2.0-3424757088010.1016/j.ast.2006.10.008JousselmeA. L.GrenierD.BosséÉ.A new distance between two bodies of evidence2001229110110.1016/s1566-2535(01)00026-42-s2.0-0010623928WenC.WangY.XuX.Fuzzy Information Fusion Algorithm of Fault Diagnosis Based on Similarity Measure of Evidence2008BerlinSpringer506515SunbergZ.RogersJ.A belief function distance metric for orderable sets201314436137310.1016/j.inffus.2013.03.0032-s2.0-84876354757CuzzolinF.A geometric approach to the theory of evidence20083845225342-s2.0-4684910427710.1109/TSMCC.2008.919174WangJ.HuY.XiaoF.DengX.DengY.A novel method to use fuzzy soft sets in decision making based on ambiguity measure and Dempster-Shafer theory of evidence: an application in medical diagnosis20166911110.1016/j.artmed.2016.04.0042-s2.0-84965046342BaldwinJ. F.A Calculus for Mass Assignments in Evidential Reasoning1994New Jersey, NJ, USAJohn Wiley & Sons, Inc.513531LefevreE.ColotO.VannoorenbergheP.De BrucqD.Contribution des mesures d’information à la modélisation crédibiliste de connaissances20001728797FabreS.AppriouA.BriottetX.Presentation and description of two classification methods using data fusion based on sensor management200121497110.1016/S1566-2535(00)00018-X2-s2.0-0002450033