ABSTRACT
This paper addresses the optimal design problems for constant-stress accelerated degradation test (CSADT) based on gamma processes with fixed effect and random effect. For three optimization criteria, we prove that optimal CSADT plans with multiple stress levels degenerate to two-stress-level test plans only using the minimum and maximum stress levels under model assumptions. Under each optimization criterion, the optimal sample size allocation proportions for the minimum and maximum stress levels are determined theoretically. The effect of the stress level on the objective functions is also discussed. A numerical example and a simulation study are provided to illustrate the obtained results.
MATHEMATICS SUBJECT CLASSIFICATION:
Acknowledgments
This work was supported by the National Natural Science Foundation of China under Grant No. 11671080, the Jiangsu Provincial Key Laboratory of Networked Collective Intelligence under Grant No. BM2017002, and the Innovation Project for College Graduates of Jiangsu Province of China under Grand No. KYLX16 0183.
Appendix
A.1. Proof of Theorem 1
Proof.
Based on Criterion 1: For fixed N, M, Δt, maximizing det[I(θ)|N1, …, NJ − 1] in (Equation13(13) (13) ) is equivalent to maximizing where Pk = Nk/N is the allocation proportion of units under stress level sk, and . Denoting , we further have For j = 1, set ∂D(P1, …, PJ − 1)/∂P1 = 0, then we have . For j = 2, …, J − 1, we have ∂D(P1, …, PJ − 1)/∂Pj = (sj − sJ)(sj − s1) < 0. Hence, D(P1, …, PJ − 1) is a monotone decreasing function of Pj for j = 2, …, J − 1. Since 0 ⩽ Pj ⩽ 1, D(P1, …, PJ − 1) is maximized at P*j = 0 for j = 2, …, J − 1. As a result, the test units will not be arranged at the stress level sk, k = 2, …, J − 1 with the optimal test plan. That is the optimal CSADT plan is a two-stress-level CSADT plan based on Criterion 1 as claimed. Based on Criterion 2: For fixed N, M, Δt, minimizing in (Equation14(14) (14) ) is equivalent to minimizing Denote . We further have It can be seen that Avar(P1, …, PJ − 1) is an increasing function of V(P1, …, PJ − 1), since αΔtψ1(αΔt) > 1. Therefore, minimizing Avar(P1, …, PJ − 1) is equivalent to minimizing V(P1, …, PJ − 1). Furthermore, For j = 1, set ∂V(P1, …, PJ − 1)/∂P1 = 0, then we have . For j = 2, …, J − 1, we have ∂V(P1, …, PJ − 1)/∂Pj = 4(sj − sJ)(s1 − sj)/(s1 + sJ)2 > 0. Hence, V(P1, …, PJ − 1) is a monotone increasing function of Pj for j = 2, …, J − 1. Since 0 ⩽ Pj ⩽ 1, V(P1, …, PJ − 1) is minimized at P*j = 0 for j = 2, …, J − 1. As a result, the test units will not be arranged at the stress level sk, k = 2, …, J − 1 with the optimal test plan. That is the optimal CSADT plan is a two-stress-level CSADT plan based on Criterion 2 as claimed. Based on Criterion 3: For fixed N, M, Δt, minimizing tr[I− 1(θ)|N1, …, NJ − 1] in (Equation15(15) (15) ) is equivalent to minimizing A(P1, …, PJ − 1):
We further have
For j = 1, set ∂A(P1, …, PJ − 1)/∂P1 = 0, then we have (A.1) (A.1) From (EquationA.1(A.1) (A.1) ), one can have that , since the numerator and denominator of the left side of (EquationA.1(A.1) (A.1) ) is positive, and the numerator is greater than the denominator at both sides of (EquationA.1(A.1) (A.1) ). For j = 2, …, J − 1, we have Hence, A(P1, …, PJ − 1) is a monotone increasing function of Pj for j = 2, …, J − 1. Since 0 ⩽ Pj ⩽ 1, A(P1, …, PJ − 1) is minimized at P*j = 0 for j = 2, …, J − 1. As a result, the test units will not be arranged at the stress level sk, k = 2, …, J − 1 with the optimal test plan. That is the optimal CSADT plan is a two-stress-level CSADT plan based on Criterion 3 as claimed.
A.2. Proof of Theorem 2
Proof.
Based on Criterion 1: By the proof of Theorem 1 for Criterion 1, we know that to maximize the objective function of Criterion 1 is equivalent to maximizing s21P1 + s2J(1 − P1) − [s1P1 + sJ(1 − P1)]2. Setting the first derivative of the objective function with respect to P1 to zero, it is straightforward to find that the maximal determinant of the Fisher information matrix is reached by setting the allocation proportion as P1 = N1/N = 1 − NJ/N = 1/2. Based on Criterion 2: By the proof of Theorem 1 for Criterion 2, we know that to minimize the objective function of Criterion 2 is equivalent to minimizing [s1P1 + sJ(1 − P1)]2/[s21P1 + s2J(1 − P1)]. Taking the first derivative of the objective function with respect to P1 and setting the result to zero, it is straightforward to find that the optimal allocation proportion is P1 = N1/N = 1 − NJ/N = sJ/(s1 + sJ). Based on Criterion 3: By the proof of Theorem 1 for Criterion 3, we know that to minimize the objective function of Criterion 3 is equivalent to minimizing Setting the first derivative of the objective function with respect to P1 to zero, it is straightforward to find that the optimal allocation proportion is
A.3. Proof of Lemma 1
Proof.
Since X|β0 ∼ Ga(Cβ0, A) and β0 ∼ Ga(γ, δ), the joint PDF of (X, β0) is
Then
By using integral transformation, it can be derived that where B(p, q) = ∫10tp − 1(1 − t)q − 1dt is the beta function, and B(p, q) = Γ(p)Γ(q)/Γ(p + q). Finally, we can obtain that Other results in Lemma 1 can be derived by using the similar method.