Overview of Results

Machine Learning Heart Date: September 10, 2025

CNN-Based Remaining Useful Life Prediction for Aircraft Engines (MATLAB Implementation)

This article will introduce how to implement a Remaining Useful Life (RUL) prediction model for aircraft engines based on Convolutional Neural Networks (CNN) using MATLAB.

The prediction of the remaining useful life of aircraft engines is a key technology in predictive maintenance. By analyzing data, models can be established to predict when the engine may require maintenance or replacement.

MATLAB Implementation Code

% CNN-Based Remaining Useful Life Prediction for Aircraft Engines
% Assuming pre-processed training data is available
% Load data
load('engine_data.mat'); % Contains trainData, trainRUL, testData, testRUL
% Data preprocessing
% Reshape data to the format required by CNN [number of samples, number of features, time steps, 1]
trainData = reshape(trainData, [size(trainData,1), size(trainData,2), size(trainData,3), 1]);
testData = reshape(testData, [size(testData,1), size(testData,2), size(testData,3), 1]);
% Create CNN model
layers = [    imageInputLayer([size(trainData,2) size(trainData,3) 1]) % Input layer
    convolution2dLayer(3, 32, 'Padding', 'same') % Convolutional layer
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2) % Pooling layer
    convolution2dLayer(3, 64, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    maxPooling2dLayer(2, 'Stride', 2)
    convolution2dLayer(3, 128, 'Padding', 'same')
    batchNormalizationLayer
    reluLayer
    fullyConnectedLayer(256) % Fully connected layer
    reluLayer
    dropoutLayer(0.5)
    fullyConnectedLayer(128)
    reluLayer
    dropoutLayer(0.5)
    fullyConnectedLayer(1) % Output layer
    regressionLayer];
% Set training options
options = trainingOptions('adam', ...
    'MaxEpochs', 100, ...
    'MiniBatchSize', 32, ...
    'InitialLearnRate', 0.001, ...
    'LearnRateSchedule', 'piecewise', ...
    'LearnRateDropFactor', 0.5, ...
    'LearnRateDropPeriod', 20, ...
    'Shuffle', 'every-epoch', ...
    'ValidationData', {testData, testRUL}, ...
    'ValidationFrequency', 30, ...
    'Verbose', 1, ...
    'Plots', 'training-progress');
% Train model
net = trainNetwork(trainData, trainRUL, layers, options);
% Predict
predictedRUL = predict(net, testData);
% Evaluate model
rmse = sqrt(mean((predictedRUL - testRUL).^2));
fprintf('Test set RMSE: %.2f\n', rmse);
% Visualize results

Complete Code

Code access (not for public use, please do not misuse)