matlab手册_RUL model

1. a basic example - [exponentialDegradationModel]

– (illustrating RUL prediction with a degradation model)
Update RUL Prediction as Data Arrives.

load UpdateRULExampleData   #Data

exponentialDegradationModel
语法：
$mdl=exponentialDegradationModel$ all properities default
$mdl=exponentialDegradationModel(Name,Value)$specify properities
共5 parameters

通用function command

语法：
$update(mdl,data)$
$head(datax,5)$ %datax是table才可

2. Similarity-Based RUL估计

workflow:

1. preprocessing,
2. selecting trendable features, 趋势特征
3. constructing a health indicator by sensor fusion, 构建健康指标
4. training similarity RUL estimators,
5. validating prognostics performance. 验证预测性能

1. data准备：
   训练数据包含218组从运行到故障的模拟。每一组/ensemble，有26列，为ID, 时间戳记录、3种operating conditions设置，21个传感器测量的数据。

i) 将data进行划分： 训练集 验证集

linearDegradationModel
语法：
$mdl=linearDegradationModel$ all properities default
$mdl=linearDegradationModel(Name,Value)$ specify

residualSimilarityModel
语法：
$mdl=residualSimilarityModel$ all properities default
$mdl=residualSimilarityModel(initModel)$
$mdl=residualSimilarityModel{(}_{,}Name,Value)$

%e.g:
load('residualTrainTables.mat')
mdl = residualSimilarityModel('Method',"arma3",'Distance',"absolute");
fit(mdl,residualTrainTables,"Time","Condition")

语法：
$C=vertcat(A,B)$ %垂直串联
$Y=splitapply(func,X,G)$ %将 X 划分为 G 指定的组，并向每个组应用函数 e.g:splitapply(@mean,Height,G)
$A=cellfun(func,C)$ %对元胞数组中的每个元胞应用函数
A = cellfun(___,Name,Value )

centerstats.SD.(char(v)) = splitapply(@std, trainDataUnwrap.(char(v)), clusterIndex);

C = {1:10, [2; 4; 6], []}
A = cellfun(@mean,C)

C = {'Monday','Tuesday','Wednesday','Thursday','Friday'}
A = cellfun(@(x) x(1:3),C,'UniformOutput',false)  %@(x):func作用对象  x(1:3)：function/作用对象的范围圈定  C：具体对象   'UniformOutput',false：A形式为cell

A = cellfun(@mean,data)
A = cellfun(@(x),mean(x),data) %x形参，data实参

语法：
$B=rowfun(func,A)$ %将函数应用于table或timetable行

B = rowfun(@hypot,A,'OutputVariableNames','z')  % function：hypot, to each row of A.

语法：
$C=cellstr(A)$ A:string 转cell
$M=movmean(A,k)$ % k 个数据点的均值组成的数组 Moving mean
$M=movmean(A,[kb,kf])$ %当前位置的元素、后面的 kb 个元素和前面的 kf 个元素，算均值

语法：
$compare(mdl,data)$ %data: (一条不完整线)test component degradation data &(多条全退化线) fit(mdl, 训练用的数据)中’训练用的all数据’ =>重叠画
$compare{(}_{,}Name,Value)$

语法：
$histogram{(}_{,}Name,Value)$

e.g: x = randn(1000,1);
h = histogram(x,‘Normalization’,‘probability’)

语法：
$linkaxes(ax)$ 同步all图的坐标范围，以最大的为对齐

*0.

Matlab反复提及：Degradation models estimate the RUL by predicting when a monitored signal will cross a predefined threshold.

*1. 好code

rowNormalized(isnan(rowNormalized) | isinf(rowNormalized)) = 0;

tmp = cellfun(@(tbl) tbl(:, cellstr(dataVariables(ct))), trainDataNormalized, 'UniformOutput', false); %e.g: ct=1

*2.总指导