Chapter 23 Accelerated Destructive Degradation Test Data, Models, and Analysis William Q. Meeker and Luis A. Escobar Iowa State University and Louisiana State University Copyright 2002-2003 W. Q. Meeker and L. A. Escobar. Complements to the authors’ text Statistical Methods for Reliability Data, John Wiley & Sons Inc., 1998. & Printing Group, Hewlett-Packard) Weeks Aged 2 4 6 12 16 8 6 4 18 7 6 0 13 0 0 6 6 8 6 9 23 8 6 6 20 Weeks 8 31 24 25 88 Totals 15 20 23 - 1 23 - 3 Accelerated Destructive Degradation Tests Data, Models, and Data Analysis Chapter 23 Objectives • Describe useful accelerated destructive degradation test (ADDT) reliability models. • Show the connection between degradation reliability models and failure-time reliability models. • Present methods of data analysis and reliability inference for ADDT. 4 6 9 6 23 - 4 23 - 2 • Discuss the use of ADDT data to estimate acceleration factors. • Illustrate methods with examples. 6 6 AdhesiveBondB ADDT Plan Number of Units Measured at Each (Time, Temperature) Combination 70 6 6 60 7 16 8 12 Weeks 6 8 4 8 8 2 50 25 0 0 AdhesiveBondB Data Resp:Linear,Time:Linear, Dist:Normal 50DegreesC 80 60 100 40 60 20 40 15 20 0 70DegreesC 10 15 20 5 0 60 80 100 40 20 0 0 20 10 100 5 Weeks 80 0 5 60DegreesC 10 15 23 - 6 20 AdhesiveBondB ADDT Data Scatter Plot at Individual Conditions of Temperature Linear–Linear Axes DegreesC Work done jointly with Danny L. Kugler and Laura L. Kramer (Imaging December 14, 2015 8h 10min AdhesiveBondB ADDT Data • Objective: Assess the strength of an AdhesiveBond over time; estimate the proportion of devices with a strength below 40 Newtons after 5 years of operation (approximately 260 weeks) at 25◦C. • The test is destructive; each unit can be measured only once. There were 6 right censored observations. 0 Temp ◦C 8 50DegreesC 60DegreesC 70DegreesC 5 10 AdhesiveBondB Data Destructive Degradation Scatter Plot Resp:Linear,Time:Linear AdhesiveBondB ADDT Data Scatter Plot Linear–Linear Axes 8 — 50 60 70 Totals 100 80 60 40 20 0 0 23 - 5 Newtons • Test plan: 8 units with no ageing were measured at the start of the experiment. A total of 80 additional units were aged and measured according to the following temperatures and time schedule. Newtons 0 0 80 40 20 10 0 AdhesiveBondB Data Resp:Log,Time:Square root, Dist:Normal 10 15 50DegreesC 5 10 15 70DegreesC 5 10 15 60DegreesC 5 23 - 7 AdhesiveBondB ADDT Data Scatter Plot at Individual Conditions of Temperature Square Root–Log Axes 80 40 20 10 80 40 20 10 Weeks Statistical Model for ADDT Data = µij + ǫijk , k = 1, . . . , nij = hd(Dij ) + ǫijk • Sample path model: For observation k at time τi and accelerating variable(s) level xj the model is yijk where µij = hd(Dij ) and yijk are, respectively, monotone increasing transformations of Dij and the observed degradation. nij is the number of observations at time ti and AccVarj vector level xj . ǫijk is a residual deviation which describes unit-to-unit variability with (ǫijk /σ) ∼ Φ(z). 23 - 9 • The transformations for the observed degradation, Dij , ti,, and the AccVarj might be suggested by physical/chemical theory, past experience, or the data. 10 20 30 70 DegreesC 60 DegreesC 50 DegreesC 40 50 60 23 - 11 Degradation Path Model: Multiple Temperature Levels D(τ, x, β ) = exp[β0 + β1 exp(β2x)τ ] √ τ = Weeks, x is Arrhenius-Transformed Temp Linear–Linear Axes 100 80 60 40 20 0 0 Weeks Model for Accelerated Degradation Path • Actual degradation path model: Actual path of a unit at ti and accelerating variable(s) level (combinations) AccVarj (e.g., temperature or temperature and relative humidity) is Dij = D(τi, xj , β ) where τi = ht(ti) is a known monotone increasing transformations of ti and xj = ha(AccVarj ) are known transformations of the accelerating variables at the jth level (or combination of levels) AccVarj . When there is no possibility of confusion, τi and xj are called the time and the AccVar, respectively. • Rates in the model are with respect to τ = ht(t). • Path parameters: the elements of β are fixed but unknown. 23 - 8 • Though no commonly used in accelerated testing it is possible to have models with interactions among the accelerating variables. 100 80 60 40 20 0 0 0 10 20 70 DegreesC 60 DegreesC 50 DegreesC 10 30 40 70 DegreesC Weeks 20 30 50 40 50 60 60 Using Transformations to Linearize a Degradation Model log[D(τ, x, β )] = β0 + β1 exp(β2x)τ √ Weeks, x is Arrhenius-Transformed Temp Square Root–Log Axes 100 50 20 10 Weeks 23 - 12 23 - 10 Degradation Path Model: Single Temperature Level D(τ, x, β ) = exp[β0 + β1 exp(β2x)τ ] √ Weeks and x is Arrhenius-Transformed Temp Linear–Linear Axes τ = τ = Newtons Newtons Newtons Newtons k = 1, . . . , nij A Class of Linear Degradation Models • These models are of the form yijk = µij + ǫijk ′ x )τ + ǫ , = β0 + β1 exp(β 2 j i ijk where yijk , τi, and xj may be transformations of the measured degradation, ti, and the accelerating variable(s) AccVarj , respectively. • The model is linear in the sense that for specified AccVar vector xj , the degradation is linear in τi. i=2 23 - 13 ◮ For a single (scalar) AccVar x , β ′ x = β x . 2 j j j 2 ′ = (β , β , . . . ), ◮ For multiple AccVar, xj , xj = (x2, . . . , xp)′, β 2 2 3 ′ x is a linear combination of the AccVar, i.e., and β 2 j p X βixji = β2xj2 + β3xj3 + · · · + βpxjp. β 2′ xj = Linear Degradation Model for the AdhesiveBondB Data yijk = β0 + β1 exp(β2xj )τi + ǫijk For the AdhesiveBondB data, the strength of the adhesive as a function of time and temperature is modeled by where yijk = log(Newtonsijk ) p √ ti = Weeksi τi = xj = −11605/(◦Cj + 273.15) (ǫijk /σ) ∼ Φnor (z). 23 - 15 The ǫijk term contains model and measurement errors. AdhesiveBondB ADDT Data Normal Model ML Individual Line Fits • For each temperature level three individual ML estimates [j] b [j]. are obtained: βb0 , υb [j], and σ 4.490 −0.1088 −0.3626 −0.2089 υb [j] 0.01944 0.02214 0.01494 sceυb[j] ML Estimates Upper −0.08309 −0.16969 −0.1424 −0.32643 −0.2571 −0.4028 Lower 95% Approximate Confidence Interval for υ [j] • A summary of the Normal model individual ML fits for the AdhesiveBondB data is 4.489 [j] 50◦C βb0 60◦C 4.400 AccVarj 70◦C 23 - 17 Linear Degradation Model Interpretation of the Parameters ′ x )τ + ǫ yijk = β0 + β1 exp(β 2 j i ijk For the linear degradation model √ ti then β0 is the intercept at time t = 0. • β0 is path intercept parameter when τ = 0. For example, ◮ If τi = ◮ If τi = log(ti) then β0 is intercept at time t = 1. ′ x ). • The degradation rate at xj is υ(xj )(AccVarj ) = β1 exp(β 2 j • The sign (±) of β1 determines if the degradation is increasing or decreasing in time. 23 - 14 • For a power transformation of time τ = tκ, the components of the vector parameter β 2 are related to the amount of acceleration obtained by increasing the accelerating variables AccVar. ADDT Individual Analysis Likelihood for Fixed AccVar level xj i k=1 σ nij Y Y 1 φ 1−δijk δ ijk y y ijk − µij ijk − µij × 1−Φ σ σ • For the data at fixed xj of the AccVar with exact failure times and right-censored observations, the likelihood is Lj (θ |DATA) = ′ x )τ , δ where µij ≡ µ(τi, xj , β ) = β0 + β1 exp(β 2 j i ijk indicates whether observation yijk is a failure (δijk = 1) or a right ′ , σ), and n censored observation (δijk = 0), θ = (β0, β1, β 2 ij is the number of observations at (τi, xj ). 80 40 20 10 80 40 20 10 0 0 10 15 50DegreesC 5 10 15 70DegreesC 5 Weeks 80 40 20 10 0 AdhesiveBondB Data Resp:Log,Time:Square root, Dist:Normal 5 10 15 60DegreesC AdhesiveBondB ADDT Data Individual Normal Model ML Fits Square Root–Log Axes [j] b [j]τ b [j] = βb0 + υ µ 23 - 18 23 - 16 • For fixed xj , the model parameters are: the spread σ, the ′ x ). intercept β0, and the slope of the line υj = β1 exp(β 2 j The parameter υj can be interpreted as the degradation rate of µ(τ, xj , β ) with respect to the transformed time τ. Newtons 50DegreesC 60DegreesC 70DegreesC Weeks 5 10 15 AdhesiveBondB Data Destructive Degradation Individual Regression Analyses Resp:Log,Time:Square root, Dist:Normal 50 55 60 65 70 Degradation rate versus DegreesC on Arrhenius Scale for AdhesiveBondB Data Resp:Log,Time:Square root,x:arrhenius, Dist:Normal Arrhenius Plot of |υb [j]| Individual Degradation Rates Normal Model ML Estimates 0 AdhesiveBondB ADDT Data Overlay of Individual Normal Model Fits Square Root–Log Axes [j] b [j] = βb0 + υ b [j]τ µ 100 80 60 50 40 30 20 10 0.40 0.30 0.25 0.20 0.15 0.10 0.05 45 DegreesC on Arrhenius Scale Standard Error 0.03864 20 75 23 - 19 23 - 21 95% Approximate Confidence Interval Lower Upper 4.396 4.547 AdhesiveBondB Data ML Estimates for the Acceleration Model Fit ML Estimate 4.471 • Parameter estimates Parameter β0 0.7536 2.261 × 109 0.5375 0.1841 −3.989 × 109 0.05488 0.1356 1.595 × 109 0.6364 0.01233 −8.641 × 108 β2 0.1580 β1 σ b υ(50) = −0.1025 b υ(70) = −0.3883 23 - 23 • Normal model ML for the slopes (degradation rates) at b ◦Cj ) = βb1 exp(βb2xj ), each temperature are obtained from υ( where xj = −11605/(◦Cj + 273.15). In this case for the four temperatures of interest the ML estimates are b υ(25) = −0.0151, b υ(60) = −0.2035, Individual Degradation Rate Estimates • The ML estimates υb [j] (slopes of the individual lines) can be used to identify the relationship between the degradation rate and the AccVar. • When the degradation is decreasing, use absolute values of the degradation rate. • Because ′x log(| υj |) = log(| β1 |) + β 2 j the surface log(| υb [j] |) versus the AccVar xj should be approximately linear in the xj if the model relating degradation rate and the AccVar is adequate. Then ◮ For a single AccVar xj , the plot of log(| υb [j] |) versus xj should be approximately linear. 23 - 20 ◮ For an AccVar vector xj , the plot of log(| υb [j] |) versus any of the accelerating variables, conditional on fixed values of the remaining accelerating variables, should be approximately linear. Likelihood for All Data With Right Censored Data σ φ σ × 1−Φ σ yijk − µij 1−δijk • For a sample of n units consisting of exact failure times and right-censored observations, the likelihood can be expressed as ijk Y Lj (θ |DATA) L(θ |DATA) = j Y 1 yijk − µij δijk L(θ |DATA) = ′ , σ)′ , and δ where θ = (β0, β1, β 2 ijk indicates whether observation ijk is a failure (δijk = 1) or a right censored observation (δijk = 0). µij = β0 + β1 exp(β2xj )τi 11605 . xj = − ◦ C + 273.15 j 50DegreesC 60DegreesC 70DegreesC 25DegreesC 5 10 15 AdhesiveBondB data Destructive Degradation Regression Analyses Resp:Log,Time:Square root,DegreesC:Arrhenius, Dist:Normal AdhesiveBondB ADDT Data Normal Model Arrhenius ML Fit b ij = βb0 + βb1 exp(βb2xj )τi µ 0 Weeks 20 23 - 24 23 - 22 • Note that µij is a nonlinear function of the accelerating variable xj . 10 20 30 40 60 50 80 100 • For the AdhesiveBondB data where xj is a scalar, Newtons Newtons Slope Model Checking Residual Plots • Residuals versus fitted values. 60 DegreesC 65 70 .05 .50 .95 23 - 25 This is useful when • Residuals versus accelerating variables. • Residuals versus time of exposure. • Residuals versus observation order. observations are taken sequentially. • Residual probability plot. 55 AdhesiveBondB Data Destructive Degradation Residuals versus DegreesC Resp:Log,Time:Square root,DegreesC:Arrhenius, Dist:Normal AdhesiveBondB ADDT Data Residuals Versus Temperature Conditions 4 2 0 -2 -4 -6 50 23 - 27 23 - 29 • There appears to be some evidence of nonconstant variance, but it is not systematic with temperature or times. • The vertical line at 0 is the median life of the standardized distribution. Then approximately 50% of the residuals should be below that line. • The standardized residuals look approximately like a random sample from a NOR(0, 1). Some Comments on the AdhesiveBondB Residuals Standardized Residuals .002 .0005 .0001 .01 .2 .1 .05 .4 .6 .98 .95 .9 .8 .995 .999 .9999 Standardized Residuals 4 2 0 -2 -4 -6 3.2 3.4 3.6 Fitted Values 3.8 4.0 4.2 4.4 AdhesiveBondB Data Destructive Degradation Residuals versus Fitted Values Resp:Log,Time:Square root,DegreesC:Arrhenius, Dist:Normal AdhesiveBondB ADDT Data Residuals Versus Fitted Values 3.0 AdhesiveBondB ADDT Data Sev Residual Probability Plot -2 -1 Standardized Residuals 0 1 2 3 .95 .50 .05 23 - 28 23 - 26 4.6 AdhesiveBondB Data Destructive Degradation Residual Probability Plot with 95% Simultaneous Confidence Bands Resp:Log,Time:Square root,DegreesC:Arrhenius, Dist:Normal Normal Probability Plot -3 b y−µ b σ # 23 - 30 b x)τ , τ = h (t), x = h (AccVar), b = βb0 + βb1 exp(β where µ a t 2 b ,σ and βb0, βb1, β 2 b are ML estimates of the corresponding parameters. When AccVar is a vector, ha is generally a different function for each of its elements. FbY (y; τ, x) = Φ • The ML estimate of the degradation distribution for given (t, x) is • The p quantiles of the distribution is yp = µ(t, x, β )+σΦ−1(p). y − µ(τ, x, β ) FY (y; τ, x) = P (Y ≤ y; τ, x) = Φ σ ′ x)τ. where y = hd(degradation), µ(τ, x, β ) = β0 + β1 exp(β 2 " • The degradation distribution, for given time t and AccVar vector x is Distribution of Degradation at (Possibly Transformed) Time and AccVar (t, AccVar) Probability AdhesiveBondB ADDT Data Distribution of Degradation at Given Time and Temperature (t, Temp) • For the AdhesiveBondB data, the Normal model ML esti mate of FY (y; τ, x) at time t and temperature ◦C is −1 (p). b+σ b Φnor ybp = µ 100 200 Weeks 500 1% 0.1% 1000 23 - 33 23 - 31 b y−µ FbY (y; τ, x) = Φnor b σ √ b = βb0+βb1 exp(βb2x)τ , τ = Weeks, x = −11605/(◦C+ where µ 273.15), βb0 = 4.471, βb1 = −864064160, βb2 = 0.6364, b = 0.1580 σ 20 ML Estimate Showing Proportion Failing as a Function of Time at 25◦C −1 b Φnor ybp = βb0 + βb1 exp(βb2x)τ + σ (p) 0 • For p ≥ Φ [(β0 − µf )/σ] i h tbp = ht−1 νb + ςbΦ−1(p) t̃p] = [tbp/w, tbp × w] Zlog(bt ) = [log(tbp) − log(tp)]/scelog(bt ) is p p e [tp, p where w = exp[z(1−α/2)scebt /tbp]. 23 - 35 • An approximate 100(1−α)% confidence interval for tp based on the large-sample approximate NOR(0, 1) distribution of where ht−1(·) is the inverse of the ht(·) function. ML Estimates for Quantiles of the Induced Failure Time Distribution at AccVar x and Critical Degradation Df Decreasing Linear Degradation Newtons • The ML estimate of the p quantile (log Newtons) is 100 50 Induced Failure Time Distribution at Fixed Values of (AccVar, Df ) for Decreasing Linear Degradation " ς= for t ≥ 0 and # ′ x) σ exp(−β 2 . | β1 | µf − µ(τ, x, β ) σ • Observe that T ≤ t [i.e., ht(T ) ≤ τ ] is equivalent to observed degradation less than Df (i.e., Y ≤ µf ,) where µf = hd(Df ). Then τ −ν ς FT (t; x, β ) = FY (µf ; x, β ) = Φ = Φ ′ x) (β0 − µf ) exp(−β 2 | β1 | where τ = ht(t), ν= • The failure time distribution is a mixed distribution with a spike of Pr(T = 0) = Φ [(β0 − µf )/σ] at t = 0. 23 - 32 For t > 0 the cdf is continuous and it agrees with the cdf of a log-location-scale variable with standardized cdf Φ(z), location ν and scale ς. Quantiles for the Induced Failure Time Distribution at AccVar vector x and Critical Condition Df Decreasing Linear Degradation h ′ x) (β0 − µf ) exp(−β 2 | β1 | and i ς= ′ x) σ exp(−β 2 . | β1 | ht(tp) = τp = ν + ςΦ−1(p) • For p ≥ Φ [(β0 − µf )/σ] where ν= Thus tp = ht−1 ν + ςΦ−1(p) . " (β0 − µf ) + σΦ−1(p) . | β1 | # • Substituting the expressions for ν and ς and after simplifications, shows that ′ x + log log[ht(tp)] = log(τp) = −β 2 | βb1 | # . # √ t. 23 - 34 Thus, the log of the transformed failure-time distribution quantiles are linear in the transformed AccVar vector x. AdhesiveBondB ADDT Data Failure Time Distribution Quantile Estimates " " b Φ−1(p) (βb0 − µf ) + σ • For the AdhesiveBondB accelerated test data, ht(t) = Then q log[ tbp] = −βb2x + log This implies that b Φ−1(p) (βb0 − µf ) + σ log(tbp) = −2βb2x + 2 log . | βb1 | Thus the logs of the quantiles of the failure-time distribution are linear in the AccVar vector x. 23 - 36 50 60 70 10% 90% 50% Model plot for AdhesiveBondB data Resp:log,Time:Square root,DegreesC:arrhenius, Dist:normal Failure−time distribution for degradation failure level of 40 Newtons 40 80 23 - 37 AdhesiveBondB Data Model Plot ML Estimate of Failure Time Distribution as Functionn of Temperature o b Φ−1(p)]/ | βb1 | log(tbp) = −2βb2x + 2 log [(βb0 − µf ) + σ 10000 1000 100 30 Probability spike at time zero = 3.6473e−007 20 ′ x )τ + ǫ = β0 + β1 exp(β 2 j i ijk !# !# = β0 + β1 exp(β2xj2 + β3xj3)τi + ǫijk 23 - 41 Acceleration Factors • Here we consider time power transformations, i.e., τ = ht(t) = tκ, where κ 6= 0. • To obtain the accelerating effect of the AccVar x, let τ (x) and τ (xU ) be the (transformed) times to reach the critical degradation Df when the (transformed) accelerating variable take values x and xU , respectively. Df = D[τ (x), x, β ] = D[τ (xU ), xU , β ] • Solving for τ (x) and τxU the equation gives ht[t(xU )] τ (x U ) ′ (x − x )]. = = exp[β 2 U τ (x ) ht[t(x)] 1.2728 30 35 40 45 50 55 60 23 - 40 23 - 38 Using τ (x) = ht[t(x)] = [t(x)]κ and solving for t(xU )/t(x) yields t(xU ) 1 ′ AF (x) = = exp β 2 (x − x U ) . t(x) κ 25 Arrhenius relationship with activation energy in units of eV, AdhesiveBondB Data Normal Model Acceleration Factors as a Function of Temperature 200 100 50 20 10 5 2 1 Degrees C AdhesiveBondD ADDT Data • Experimental factors ◮ AccVar: Temp (Temperature): RH (Relative Humidity): 23 - 42 • Sample size: 6 units allocated at each (Weeks, RH, Temp) combination. This gives a sample size equal to: 6 × 8 × 2 × 3 = 288 units. ◮ Weeks : Measurements after 0, 1, 2, 4, 6, 8, 12, 16 Weeks 50◦C, 60◦C, 70◦C 20%, 80% • The test is destructive; each unit can be measured only once. There is interest in estimating the proportion of devices with a strength below 25 Newtons after 5 years of operation at room temperature of 25◦C and relative humidity of 50%. • Objective: Assess the strength of an AdhesiveBond over time. Acceleration Factor relative to 25 Degrees C 10 1 DegreesC on Arrhenius Scale AdhesiveBondB ADDT Data ML Estimates of Acceleration Factors " AF (temp) = exp 2 β2 " 11605 11605 − tempU K temp K • The Normal model ML estimate are d (temp) = exp 2 βb AF 2 where βb2 = 0.6364. • The ML estimate of the Arrhenius activation energy is 2 × βb2 = 2 × 0.6364 = 1.2728 yijk yijk = log(Newtonsijk ) p √ ti = Weeksi τi = (ǫijk /σ) ∼ Φnor (z). 11605 xj2 = − ◦ Cj + 273.15 RH xj3 = log . 1 − RH • The AccVar transformations are Arrhenius for Temp (Temperature) and logit for RH (Relative Humidity), i.e., where • AdhesiveBondD Model A Two-Variables Linear Degradation Model With Temperature and Relative Humidity as AccVar d (temp) is done easily using SPLIDA. • The computation of AF d (60) = 182.12 For example, AF 23 - 39 where temp K = temp ◦C + 273.15 is temperature in the absolute Kelvin scale. 11605 11605 − tempU K temp K √ • For the AdhesiveBondB, ht(t) = t, then κ = 1/2 and the acceleration factor for temperature is Weeks AdhesiveBondD ADDT Data Structure of the Data 1 Weeks Aged 2 4 6 8 12 16 6 6 6 6 6 6 36 6 6 6 6 6 6 36 6 6 6 6 6 6 36 6 6 6 6 6 6 36 6 6 6 6 6 6 36 6 6 6 6 6 6 36 23 - 43 48 48 48 48 48 48 288 Totals The test plan is a completely balanced 8 × 3 × 2 full factorial arrangement in Weeks, Temp, and RH, with 6 units allocated at each combination of the experimental factors. 0 6 6 6 6 6 6 36 Temp ◦C 6 6 6 6 6 6 36 RH % 20 50 60 70 80 50 60 70 Totals Another Linear Degradation Model for an Insulation Breakdown yijk = β0 + β1 exp(β2xj )τi + ǫijk Nelson 1990, Chapter 11, discusses the following linear degradation model for time to an insulation Breakdown. where yijk = log[(Breakdown Voltage)ijk ] τi = ti = Weeksi xj = −11605/(◦Cj + 273.15) (ǫijk /σ) ∼ Φnor (z). Notice that in this instance the variable time is used in the linear scale. 23 - 45 Future Research in ADDT Data, Models, and Analysis • Nonlinear degradation models. • Coarse data. • Stochastic variability in the degradation response. • Prediction in non-constant environments. • Use of prior information. • Random initiation times. 23 - 47 AdhesiveBondD ADDT Data Acceleration Factors t(xU ) 1 ′ = exp β 2 (x − x U ) t(x) κ √ • For the AdhesiveBondD, ht(t) = t, then acceleration factor for temperature and relative humidity is AF (x) = ( ! where κ = 1/2 because the square root transformation on time. Then 11605 11605 − + AF (temp, RH) = exp 2 β2 temp K temp K U " !#) RH RHU − log 1 − RH 1 − RHU 2 β3 log where tempU K = 25◦C + 273.15 and RHU = 50% are the AccVar at use conditions. 23 - 44 • The estimation of AF (temp, RH) will be programmed in SPLIDA. ADT Models with Interactions • Most accelerated test models used in practice do not contain interaction terms. • Interaction terms imply curvature in the degradation rate versus the AccVar surface, i.e., µij = β0 + β1 exp(β2xj2 + β3xj3 + β4xj2xj3) τi. • Extrapolation is hazardous, especially with surfaces involving curvature. • A physical model could suggest interactions. 23 - 46 • Current version of the SPLIDA software does not allow for interactions. Technical Details The following slides give technical details used in SPLIDA to implement the methodology. 23 - 48 A Reparameterization of the Linear ADDT Model for Numerical Stability • The model is as before ′ x )τ + ǫ . yij = β0 + β1 exp(β 2 j i ij exp xj − x̄ γ 2 τi − τ̄ + ǫij . ′ • Suppose that x̄ is the centroid of the stress variables [i.e., x̄ = (x̄1, . . . , x̄k )′] and τ̄ is an average transformed time. Then the model can be reparameterized as yij = γ0 + γ1 where γ0 is the intercept for the average stress line (i.e., degradation line for x̄) at τ̄ ; γ1 is the slope of the average stress line; and γ 2 = β 2 are the regression coefficients corresponding to the x variables. ′ x) σ exp(−β 2 for t ≥ 0 ς= | β1 | . 23 - 49 • It can be shown that γ1 is the geometric mean of the slopes ′ x ), for the values of x stress variables in the data β1 exp(β 2 j j set. and τ −ν ς Induced Failure Time Distribution at Use Conditions x and Critical Level Df Decreasing Linear Degradation FT (t, x, β ) = Φ ′ x) (β0 − µf ) exp(−β 2 • In this case where ν= | β1 | • The failure time distribution is a mixed distribution with a spike of Φ [(µf − β0)/σ] = Φ (− ν/ς) at t = 0. For t > 0 the cdf is continuous and it agrees with the cdf of a location-scale variable with standardized cdf Φ(z), location ν and scale ς. 23 - 51 Density to Plot Induced Failure Time Distribution at Use Conditions x and Critical Level Df Decreasing Linear Degradation dt t=exp(w) dτ × ς= for − ∞ < w < ∞ ′ x) σ exp(−β 2 . | β1 | if square-root transformation on time τ −ν exp(w) ×φ ς ς Let W = log(T ), one needs to plot fW (w) which is giving by fW (w, β ) = where 1 1 2τ and τ = ht[exp(w)] if not transformation on time = ′ x) (β0 − µf ) exp(−β 2 | β1 | dt t=exp(w) dτ and ν= 23 - 53 " t > 0. # 23 - 50 Relationship Between Stable and Original Parameters • Then ′ x̄)τ̄ γ0 = β0 + β1 exp(β 2 ′ x̄) γ1 = β1 exp(β 2 γ 2 = β 2. • Solving for the βs β2 = γ 2 β1 = γ1 exp(−x̄′γ 2) β0 = γ0 − γ1τ̄ . τ −ν , ς y − µ(τ, x, β ) σ Induced Failure Time Distribution at Use Conditions x and Critical Level Df Decreasing Linear Degradation (Continued) Y = hd(degradation) ∼ Φ • In particular, the following follows: and ht(T ) ∼ Φ ς= for t ≥ 0 ′ x) σ exp(−β 2 . | β1 | 23 - 52 • This implies that ht(T ) is location-scale distributed with parameters (ν, ς) and standardized distribution Φ(·). and −τ − ν ς Induced Failure Time Distribution at Use Conditions x and Critical Level Df Increasing Linear Degradation • In this case FT (t, x, β ) = 1 − Φ ′ x) (β0 − µf ) exp(−β 2 | β1 | where as before ν= • The failure time distribution is a mixed distribution with a spike of 1 − Φ [(µf − β0)/σ] = 1 − Φ (− ν/ς) at t = 0. For t > 0 the cdf is continuous and it agrees with the cdf of continuous random variable. 23 - 54 " t > 0. y − µ(τ, x, β ) σ # Induced Failure Time Distribution at Use Conditions x and Critical Level Df Increasing Linear Degradation Y = hd(degradation) ∼ Φ • In particular the following follows: and −τ − ν , −ht(T ) ∼ Φ ς 23 - 55 • This implies that that −ht(T ) is location-scale distributed with parameters (ν, ς) and standardized distribution Φ(·). Quantiles of the Failure Time Distribution 0 h i h−1 ν + ςΦ−1(p) t ν if p ≤ Φ − ς otherwise • For decreasing degradation, the quantiles of the failure time distribution are tp = 0 n h io −1 ht − ν + ςΦ−1(1 − p) otherwise if p ≤ 1 − Φ − ν ς • For increasing degradation, the quantiles of the failure time distribution are tp = 23 - 57 Density to Plot Induced Failure Time Distribution at Use Conditions x and Critical Level Df Increasing Linear Degradation −τ − ν exp(w) ×φ ς ς × Let W = log(T ), you need to plot fW (w) which is giving by dτ for − ∞ < w < ∞ dt t=exp(w) fW (w, β ) = where 1 2τ and ς= ′ x) σ exp(−β 2 . | β1 | τ = ( ht[exp(w)] 1 if not transformation on time if square-root transformation on time = | β1 | ′ x) (β0 − µf ) exp(−β 2 dt t=exp(w) dτ and ν= 23 - 56