Influencing Factors and Evaluation Model of Quality Risks in Intelligent Manufacturing Mobile Supply Chain

2.1 Identification Framework and EIS

In IM enterprises, the production cycle lasts long, and the manufacturing process has a high complexity, involving many quality formation links. With the development of the Internet and the Internet of Things (IoT), the traditional mode of individual IM production has gradually shifted towards the collaborative production mode of SC. Meanwhile, the focus of its product quality monitoring and management has moved from the internal of the enterprise to the SC parties. In other words, the product quality of an IM enterprise that adopts the MSC depends on the internal quality supervision of the enterprise, as well as the quality supervision by MSC parties. The natural environment and resources are the foundation and support for the operation of IM MSC, which constrain and pressurize MSC parties. Therefore, this paper sets up a quality risk identification framework for IM MSC based on natural environment and resource conditions, from the perspective of quality chain (see Figure 1).

1.png

Figure 1. The identification framework for IM MSC quality risks

Note: R&D is short for research and development.

IM MSC quality risks are transferrable, poorly recognizable, and harmful, with a certain time lag. Through the MSC operation, the quality risks of IM enterprise products will be transmitted, transformed, concealed, accumulated, and interactive.

Therefore, it is crucial to identify the factors affecting the quality risks in a scientific and reasonable manner. Otherwise, the quality risks cannot be identified or evaluated accurately. This paper designs a scientific, systematic, comprehensive, and hierarchical EIS for quality risks based on the influencing factors. The EIS divides quality risks into four dimensions: natural environment and resource conditions, SC parties involved in quality formation, quality formation process, and MSC financial risks:

Layer 1 (dimensions of quality risks)

QR={QR₁, QR₂, QR₃, QR₄}={natural environment and resource conditions, SC parties involved in quality formation, quality formation process, MSC financial risks}

Specifically, QR₁ and QR₄ reflect the basic environment and economic background of MSC, respectively; QR₂ and QR₃ correspond to the quality chain of MSC. The four dimensions jointly characterize how much different enterprises on the quality chain, and different quality formation processes influence the quality risks of final IM products. Under each dimensions, detailed risk items were designed as follows:

Layer 2 (quality risks)

QR₁={QR₁₁, QR₁₂, QR₁₃, QR₁₄}={support resources risk, social environment risk, supply and processing technology risk, legal regulation risk}

QR₂={QR₂₁, QR₂₂, QR₂₃, QR₂₄}={supplier risk, logistics company risk, retailer risk, final product risk}

QR₃={QR₃₁, QR₃₂, QR₃₃, QR₃₄, QR₃₅, QR₃₆}={market research risk, R&D and trial production risk, procurement risk, manufacturing and assembly risk, distribution and after-sales risk, optimization and maintenance risk}

QR₄={QR₄₁, QR₄₂, QR₄₃}={MSC platform technology risk, MSC platform payment risk, MSC platform supervision risk}

For each risk item, several driving factors were designed as the metrics of quality risk identification and evaluation:

Layer 3 (influencing factors of quality risks)

QR₁₁={QR₁₁₁, QR₁₁₂, QR₁₁₃, QR₁₁₄}={talent quality risk, infrastructure risk, working condition risk, natural resource utilization rate}

QR₁₂={QR₁₂₁, QR₁₂₂, QR₁₂₃}={industrial policy adjustment risk, market operation mechanism risk, public opinion risk}

QR₁₃={QR₁₃₁, QR₁₃₂, QR₁₃₃, QR₁₃₄}={technology introduction risk, technology adaptation and conversion risk, technology maturity risk, technology proficiency risk}

QR₁₄={QR₁₄₁, QR₁₄₂, QR₁₄₃, QR₁₄₄}={legal supervision risk, quality inspector risk, quality standard level risk, penalty intensity risk}

QR₂₁={QR₂₁₁, QR₂₁₂, QR₂₁₃}={raw materials or parts risk, delayed delivery risk, out of stock risk}

QR₂₂={QR₂₂₁, QR₂₂₂}={untimely transportation risk, in-transit damage risk}

QR₂₃={QR₂₃₁, QR₂₃₂}={after-sales quality risk, sales strategy risk}

QR₂₄={QR₂₄₁, QR₂₄₂}={utilization method risk, user maintenance risk}

QR₃₁={QR₃₁₁, QR₃₁₂, QR₃₁₃, QR₃₁₄}={survey scope risk, survey information accuracy risk, market analysis accuracy risk, demand analysis accuracy risk}

QR₃₂={QR₃₂₁, QR₃₂₂, QR₃₂₃, QR₃₂₄}={R&D and trial production planning risk, R&D and trial production input risk, R&D and trial production output risk, R&D and trial production review risk}

QR₃₃={QR₃₃₁, QR₃₃₂, QR₃₃₃, QR₃₃₄}={procurement plan risk, supplier selection risk, procurement contract formulation risk, product acceptance risk}

QR₃₄={QR₃₄₁, QR₃₄₂, QR₃₄₃, QR₃₄₄}={manufacturing and assembly preparation risk, manufacturing and assembly capability risk, manufacturing and assembly process risk, manufacturing and assembly environment risk}

QR₃₅={QR₃₅₁, QR₃₅₂, QR₃₅₃}={packaging risk, after-sales risk, after-sales response and feedback risk}

QR₃₆={QR₃₄₁, QR₃₆₂, QR₃₆₃}={product quality control risk, product repair risk and customer relationship maintenance risk, product continuous optimization risk}

QR₄₁={QR₄₁₁, QR₄₁₂, QR₄₁₃}={platform operational capability risk, software and hardware security risk, data transmission security risk}

QR₄₂={QR₄₂₁, QR₄₂₂, QR₄₂₃}={payment information security risk, participating enterprise privacy security risk, payment method security risk}

QR₄₃={QR₄₃₁, QR₄₃₂, QR₄₃₃}={supervision system risk, supervision personnel risk}

2.1 Features of quality risks

During the identification and evaluation of IM MSC quality risks, it is difficult to quantify the risk features, owing to the multiple dimensions of the influencing factors. Figure 2 presents the causality between multiple dimensions of quality risks. To eliminate the effect of subjectivity of the evaluators, this paper converts the descriptions of risk features into intuitive triangular fuzzy numbers: ITF=([x,y,z];μ_ITF,γ_ITF), where μ_ITF is the membership, and γ_ITF is the non-membership. The membership function and non-membership function can be respectively expressed as:

${{\mu }_{ITF}}\left( i \right)=\left\{ \begin{align}  & \frac{i-x}{y-x}{{\mu }_{ITF}},a\le i<y \\ & \frac{z-i}{z-y}{{\mu }_{ITF}},y\le i\le z \\ & 0       ,   others \\\end{align} \right.$     (1)

${{\gamma }_{ITF}}\left( i \right)=\left\{ \begin{align}  & \frac{y-i+v_{p}^{\tilde{\ }}\left( i-x \right)}{y-x},x\le i<y \\ & \frac{i-y+v_{p}^{\tilde{\ }}\left( z-i \right)}{z-y},y\le i\le z \\ & 0     ,             others \\\end{align} \right.$     (2)

2.png

Figure 2. The causality between multiple dimensions of quality risks

Specifically, μ_ITF and γ_ITF mean the evaluator agrees or disagrees with the evaluation result; 1-μ_ITF-γ_ITF reflects the degree of hesitation θ_ITF of the evaluator in giving the evaluation result. The greater the θ_ITF value, the more uncertain the evaluator is about his/her evaluation result. The triangular fuzzy number [x,y,z] that characterizes the eigenvalue of quality risks can be converted from the level of quality risk features. The natural language describing the evaluation level of quality risk features can be transformed into the following triangular fuzzy number [x_k,y_k,z_k]:

$\left\{ \begin{align}  & {{x}_{0}}=0 \\ & {{x}_{j}}=\frac{j-1}{k-1}  (1\le j\le k-1) \\ & {{y}_{j}}=\frac{j}{k-1} \left( 0\le j\le k-1 \right) \\ & {{z}_{j}}=\frac{j+1}{k-1} \left( 0\le j\le k-2 \right) \\ & {{z}_{j-1}}=1 \\\end{align} \right.$     (3)

When k takes different values, the transformation equation of triangular fuzzy number can be obtained for quality risk features on each evaluation level. Here, the quality risk features of IM MSC were divided into 7 levels. The levels, attributes, and triangular fuzzy numbers of the four dimensions of quality risks are listed in Table 1.

Table 1. The levels and attributes of quality risk features

This paper combines the entropy method and the intuitionistic fuzzy entropy to determine the weight of each quality risk feature. By the entropy method, the evaluation result of quality risk q under influencing factor p, and the corresponding expected triangular fuzzy number can be solved by:

${{e}_{pq}}=\frac{{{x}_{pq}}+2{{y}_{pq}}+{{z}_{pq}}}{4}$     (4)

The entropy of the feature μ_p of quality risk p can be expressed as:

${{E}_{p}}=-\frac{\sum\limits_{q=1}^{N}{\frac{{{d}_{qp}}}{\sum\limits_{i=1}^{N}{{{d}_{qp}}}}\ln \left( \frac{{{d}_{qp}}}{\sum\limits_{q=1}^{N}{{{d}_{qp}}}} \right)}}{\ln N}$     (5)

The corresponding entropy weight ω_p can be expressed as:

${{\omega }_{p}}=\frac{1-{{E}_{p}}}{\sum\limits_{p=1}^{M}{\left( p-{{E}_{p}} \right)}}$     (6)

When the quality risk feature is weighed by intuitionistic fuzzy entropy, the intuitionistic fuzzy entropy of μ_p can be expressed as:

$E_{p}^{*}=\frac{1}{N}\sum\limits_{p=1}^{N}{\frac{\min \left( {{\mu }_{qp}},\text{  }{{\gamma }_{qp}} \right)+{{\theta }_{qp}}}{\max \left( {{\mu }_{qp}},\text{  }{{\gamma }_{qp}} \right)+{{\theta }_{qp}}}}$     (7)

The larger the E_p^*, the greater the fuzziness of the evaluation results on μ_p under influencing factor p, and the larger the weight of the corresponding quality risk feature. The corresponding entropy weight ω_p^* can be expressed as:

$\omega _{p}^{*}=\frac{1-E_{p}^{*}}{M-\sum\limits_{p=1}^{M}{E_{p}^{*}}}$     (8)

Combining ω_p and ω_p^* with geometric mean, the final weight can be obtained as:

${{\hat{\omega }}_{p}}=\frac{\sqrt{{{\omega }_{p}}\omega _{p}^{*}}}{\sum\limits_{p=1}^{M}{\sqrt{{{\omega }_{p}}\omega _{p}^{*}}}}$     (9)

3.png

Figure 3. The workflow of the comprehensive evaluation process

The proposed IM MSC quality risk evaluation model was designed based on three multi-attribute decision-making methods, namely, AHP, technique for order preference by similarity to ideal solution (TOPSIS), and data envelopment analysis (DEA). The quality risk evaluation results obtained by the three methods were converted into intuitive triangular fuzzy numbers. After passing the consistency test of Kendall’s Concordance (W) Coefficient, the three quality risk evaluation results were comprehensively evaluated by fuzzy Borda method. Figure 3 shows the workflow of the comprehensive evaluation process of the proposed model.

IM MSC quality risks may arise in various links, ranging from the supply of raw materials and parts, manufacturing, transportation, warehousing, mobile marketing to after-sales. The traditional mode, which monitors and manages the quality of a single IM enterprise or marketing platform, cannot control the quality risks of the entire MSC. Drawing on the contract theory, the quality risks were controlled from the angle of MSC after they have been accurately evaluated. The relevant parameters were configured as follows:

In the contract model, variable g represents the quality level of raw materials and parts provided by MSC; UC_SC(g) is the unit cost of the MSC to make the quality of raw materials and parts reach a certain level; ζ is the probability of defected products; UC_E(ζ) is the cost of the IM enterprise to ensure the level of product quality inspection; z is the level of quality control effort of the IM enterprise; UC_E(z) is the unit cost of the IM enterprise to make product quality reach a certain level; c_RMi is the unit price of raw materials or parts; c_P is the sales price of IM enterprise products; L₁ is the internal loss induced by production delay and selection of new supplier; L₂ is the additional loss of reputation and claims caused by the sales of defected products; λ is the probability of claims; λ[(1-g)(1-ζ)+g(1-z)] is the probability of additional loss. Then, the revenues functions of MSC supplier and IM enterprise can be respectively established as:

$\left\{ \begin{align}  & \prod{_{SC}=\sum\limits_{i=1}^{K}{{{c}_{RMi}}}[1-\zeta (1-g)]-\zeta (1-g)\rho {{L}_{1}}-\lambda [(1-g)(1-\zeta )+g(1-z)]\beta ({{c}_{P}}+{{L}_{2}})-U{{C}_{SC}}(g)} \\ & \prod{_{E}=\sum\limits_{i=1}^{K}{({{c}_{P}}-{{c}_{RMi}})}[1-\zeta (1-g)]-\zeta [(1-g)(1-\zeta ){{L}_{1}}-\lambda [(1-g)(1-\zeta )+g(1-z)](1-\beta )({{c}_{P}}+{{L}_{2}})-U{{C}_{E}}(\zeta )-U{{C}_{E}}(z)} \\\end{align} \right.$     (16)

Considering the possibility of the unilateral moral hazard of MSC supplier, the supplier must ensure its retained revenue under the risk control constraints. To maximize its revenue, the IM enterprise needs to control the product quality inspection at the optimal level. Then, formula (16) can be converted into the Lagrangian form:

$\begin{align}  & \Phi =\sum\limits_{i=1}^{K}{({{c}_{P}}-{{c}_{RMi}})}[1-\zeta (1-g)]-\zeta (1-g)(1-\rho ){{L}_{1}}\text{-}\lambda [(1-g)(1-\zeta )+g(1-z)](1-\beta )({{c}_{P}}+{{L}_{2}})-U{{C}_{E}}(\zeta )-U{{C}_{E}}(z) \\ & \text{    }+\tau \{\sum\limits_{i=1}^{K}{{{c}_{RMi}}}[1-\zeta (1-g)]-\zeta (1-g)\rho {{L}_{1}}-\lambda (1-g)(1-\zeta )+g(1-z)]\beta ({{c}_{P}}+{{L}_{2}})-U{{C}_{SC}}(g)-U{{C}_{0}}\} \\ & \text{    }+\upsilon [\sum\limits_{i=1}^{K}{{{c}_{RMi}}}\zeta +\rho {{L}_{1}}\zeta -\beta \lambda (\zeta -z)({{c}_{P}}+{{L}_{2}})-U{{{{C}'}}_{SC}}(g)]=\prod{_{SC}+\tau (}\prod{_{E}-U{{C}_{0}})+\upsilon \frac{\partial \prod{_{SC}}}{\partial g}} \\\end{align}$     (17)

where, τ is the Lagrangian factor for MSC to obey the risk control constraints; υ is the Lagrangian factor for the incentive compatible constraint of the IM enterprise. Taking the partial derivatives of the loss risk sharing ratios β and ρ of the supplier and IM enterprise and making them equal to zero:

$\begin{align}  & \frac{\partial \Phi }{\partial \beta }=(1-\tau )\lambda ({{c}_{P}}+{{L}_{2}})[(1-g)(1-\zeta )+g(1-z)] \\ & \text{        }-\upsilon \lambda (\zeta -z)({{c}_{P}}+{{L}_{2}})=0 \\\end{align}$     (18)

$\frac{\partial \Phi }{\partial \rho }=(1-\tau )\zeta (1-g){{L}_{1}}+\upsilon \zeta {{L}_{1}}=0$     (19)

Then, formulas (18) and (19) were solved, and the solutions were substituted into the Lagrangian function. Next, finding the first-order partial derivatives of g, ζ, and z and making them equal to zero:

$\left\{ \begin{align}  & \frac{\partial \Phi }{\partial g}=({{c}_{P}}+{{L}_{1}})\zeta -\lambda (\zeta -z)({{c}_{P}}+{{L}_{2}})-U{{{{C}'}}_{SC}}(g)=0 \\ & \frac{\partial \phi }{\partial \zeta }=(1-g)[\lambda ({{c}_{P}}+{{L}_{2}})-({{c}_{P}}+{{L}_{1}})]-U{{{{C}'}}_{E}}(\zeta )=0 \\ & \frac{\partial \phi }{\partial z}=\lambda ({{c}_{P}}+{{L}_{2}})g-U{{{{C}'}}_{E}}(z)=0 \\\end{align} \right.$     (20)

To maximize the overall benefits of MSC, the risk control constraints for the MSC supplier was converted into an equation, and combined with the incentive compatible constraint of the IM enterprise:

$\left\{ \begin{matrix}   {{c}_{P}}[1-\zeta (1-g)]\text{-}\zeta (1-g)\rho {{L}_{1}}-\lambda [(1-g)(1-\zeta )+g(1-z)]\beta ({{c}_{P}}+{{L}_{2}})-U{{C}_{SC}}(g)=U{{C}_{0}}  \\   {{c}_{P}}\zeta +\rho \zeta {{L}_{1}}\text{-}\beta \lambda (\zeta -z)({{c}_{P}}+{{L}_{2}})-U{{{{C}'}}_{SC}}(g)=0\text{                                                           }  \\\end{matrix} \right.$     (21)

Solving ρ, and β:

$\left\{ \begin{matrix}   \beta =\frac{{\sum\limits_{i=1}^{K}{{{c}_{RMi}}}}/{K}\;\text{-}U{{{{C}'}}_{SC}}(g)(1-g)-U{{C}_{SC}}(g)-U{{C}_{0}}}{\lambda (1-z)({{c}_{P}}+{{L}_{2}})}\text{                                                   }  \\   \rho =\frac{{(\zeta -1)z\sum\limits_{i=1}^{K}{{{c}_{RMi}}}}/{K}\;-[(1-g)(\zeta -z)+(z-1)]U{{{{C}'}}_{SC}}(g)-(\zeta -z)[U{{C}_{SC}}(g)+U{{C}_{0}}]}{\zeta (1-z){{L}_{1}}}  \\\end{matrix} \right.$     (22)

Under the risk control contract, the internal and external losses of the MSC are shared by the supplier and the IM enterprise. As long as the β value is reasonable, the overall benefits of the SC will be optimized, and the IM MSC quality risks can be effectively controlled.