restart

Reset remaining useful life degradation model

Syntax

restart(mdl)

restart(mdl,resetPrior)

restart(___,Name,Value)

Description

restart(mdl) resets the internally stored statistics of the degradation process accumulated by the previous calls to update and resets the InitialLifeTimeValue and CurrentLifeTimeValue properties of the model. If the SlopeDetectionLevel property of the model is not empty, then the slope test is also restarted, ignoring any previous detections.

restart(mdl,resetPrior) sets the prior parameter values in mdl to their corresponding posterior values when resetPrior is true.

restart(___,Name,Value) specifies properties of mdl using one or more name-value pair arguments.

Examples

Reset Degradation Model

Load training data, which is a degradation feature profile for a component.

load('expRealTime.mat')

For this example, assume that the training data is not historical data. When there is no historical data, you can update your degradation model in real time using observed data.

Create an exponential degradation model with the following settings:

• $\theta$ prior distribution with a mean of 2.4 and a variance of 0.006

• $\beta$ prior distribution with a mean of 0.07 and a variance of 3e-5

• Noise variance of 0.003

mdl = exponentialDegradationModel('Theta',2.4,'ThetaVariance',0.006,...
                                  'Beta',0.07,'BetaVariance',3e-5,...
                                  'NoiseVariance',0.003);

Since there is no life time variable in the training data, create an arbitrary life time vector for fitting.

lifeTime = [1:length(expRealTime)];

Observe the degradation feature for 100 iterations. Update the degradation model after each iteration.

for i=1:100
    update(mdl,[lifeTime(i) expRealTime(i)])
end

Reset the model, which clears the accumulated statistics from the previous observations and resets the posterior distributions to the prior distributions.

restart(mdl)

Update Exponential Degradation Model in Real Time

Load training data, which is a degradation feature profile for a component.

load('expRealTime.mat')

For this example, assume that the training data is not historical data. When there is no historical data, you can update your degradation model in real time using observed data.

Create an exponential degradation model with the following settings:

• Arbitrary $\theta$ and $\beta$ prior distributions with large variances so that the model relies mostly on observed data

• Noise variance of 0.003

mdl = exponentialDegradationModel('Theta',1,'ThetaVariance',1e6,...
                                  'Beta',1,'BetaVariance',1e6,...
                                  'NoiseVariance',0.003);

Since there is no life time variable in the training data, create an arbitrary life time vector for fitting.

lifeTime = [1:length(expRealTime)];

Observe the degradation feature for 10 iterations. Update the degradation model after each iteration.

for i=1:10
    update(mdl,[lifeTime(i) expRealTime(i)])
end

After observing the model for some time, for example at a steady-state operating point, you can restart the model and save the current posterior distribution as a prior distribution.

restart(mdl,true)

View the updated prior distribution parameters.

mdl.Prior

ans = struct with fields:
            Theta: 2.3555
    ThetaVariance: 0.0058
             Beta: 0.0722
     BetaVariance: 3.6362e-05
              Rho: -0.8429

Input Arguments

mdl — Degradation RUL model

linearDegradationModel object | exponentialDegradationModel object

Degradation RUL model, specified as a linearDegradationModel object or an exponentialDegradationModel object. restart clears the accumulated statistics in mdl and resets the InitialLifeTimeValue and CurrentLifeTimeValue properties of mdl.

resetPrior — Flag for resetting prior parameter values

false (default) | true

Flag for resetting prior parameter information, specified as a logical value. When resetPrior is:

• true, then restart sets the prior parameter values of mdl to their corresponding current posterior parameter values. For example, mdl.Prior.Theta is set to mdl.Theta.

• false or omitted, then restart does not update the prior.

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: nv1,'value'

Theta — Mean value of model θ parameter

scalar

This property is read-only.

Mean value of model θ parameter, specified as the comma-separated pair 'Theta' and a scalar. Use this argument to set the Theta property of mdl and the corresponding field of the Prior property of mdl.

ThetaVariance — Variance of model θ parameter

nonnegative scalar

This property is read-only.

Variance of the θ parameter in the degradation model, specified as the comma-separated pair 'ThetaVariance' and a nonnegative scalar. Use this argument to set the ThetaVariance property of mdl and the corresponding field of the Prior property of mdl.

Beta — Mean value of model β parameter

scalar

This property is read-only.

Mean value of model β parameter, specified as the comma-separated pair 'Beta' and a scalar. Use this argument to set the Beta property of mdl and the corresponding field of the Prior property of mdl.

This argument applies only when mdl is an exponentialDegradationModel.

BetaVariance — Variance of model β parameter

nonnegative scalar

This property is read-only.

Variance of model β parameter, specified as the comma-separated pair 'BetaVariance' and a nonnegative scalar. Use this argument to set the BetaVariance property of mdl and the corresponding field of the Prior property of mdl.

This argument applies only when mdl is an exponentialDegradationModel.

Rho — Correlation between θ and β

scalar value in the range [-1,1]

This property is read-only.

Correlation between θ and β, specified as the comma-separated pair 'Rho' and a scalar value in the range [-1,1]. Use this argument to set the Rho property of mdl and the corresponding field of the Prior property of mdl.

This argument applies only when mdl is an exponentialDegradationModel.

NoiseVariance — Model additive noise variance

nonnegative scalar

Model additive noise variance, specified as the comma-separated pair 'NoiseVariance' and a nonnegative scalar. Use this argument to set the NoiseVariance property of mdl.

SlopeDetectionLevel — Slope detection level

scalar value in the range [0,1] | []

Slope detection level for determining the start of the degradation process, specified as the comma-separated pair 'SlopeDetectionLevel' and a scalar in the range [0,1]. Use this argument to set the SlopeDetectionLevel property of mdl.

To disable the slope detection test, set SlopeDetectionLevel to [].

UseParallel — Flag for using parallel computing

false (default) | true

Flag for using parallel computing when fitting prior values from data, specified as the comma-separated pair 'UseParallel' and either true or false. Use this argument to set the UseParallel property of mdl.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• This command supports code generation with MATLAB^® Coder™. Before generating code that uses an RUL model, you must save the model using saveRULModelForCoder. For an example, see Generate Code for Predicting Remaining Useful Life.

Version History

Introduced in R2018a