Prediction And Classification Of Iot Sensor Faults Using Hybrid Deep Learning Model

Abstract

The Quality And Reliability Of Iot Ecosystems Heavily Rely On Accurate And Dependable Sensor Data. However, Sensors Can Produce Erroneous And Faulty Measurements Due To Various Factors Like Environmental Disturbances, Electrical Noise And Others. These Can Have Significant Consequences Across Different Domains, Including A Threat To Safety In Critical Systems. Though Many Researches Have Been Conducted, The Existing Literature Primarily Focuses On Fault Detection In The Sensor Data, While Fault Detection Is Useful, It Is A Reactive Approach That Identifies Faults After They Have Occurred, Meaning That Actions Are Taken After The Fault Has Already Impacted The System, Potentially Leading To Negative Consequences. In This Study, We take A Proactive Approach By Developing A Multi-Stage Solution Using Hybrid Deep Learning Models. In The First Stage, We Trained A CNN-LSTM Model To Forecast Multi-Step Future Sensor Measurements Based On Historical Data. In The Second Stage, We Trained A CNN-MLP Model Using Fault-Injected Sensor Data To Learn Patterns Associated With Different Fault Types And Classify New Measurements Accordingly. By Passing The Forecasted Sensor Values As Input To The Classification Model And Categorizing Them As Normal, Bias, Drift, Random Or Poly-Drift, We Anticipate Potential Faults Before They Manifest. The Raw Dataset Used Is The Publicly Available Intel Lab Data, Which Has Been Annotated And Fault-Injected By De Brun, El. To Identify The Most Effective Models, We Experimented With Different Models. For Regression, We Evaluatedgru, LSTM, Bilstm, CNN-GRU, CNN-LSTM, And CNN-Bilstm, Comparing Their Performance Using Root Mean Squared Error (RMSE), Mean Squared Error (MSE) And Mean Absolute Error (MAE) With 2-Split Time series Cross-Validation. CNN-LSTM Outperformed The Other Models With A MAE Of 2.0957 For A 45 Time steps Forecast. For The Classification Task, We Evaluated CNN, MLP, And CNN-MLP Using The Metrics Accuracy, Precision, Recall, And F1-Score With 5 And 10-Fold Cross-Validations. CNN-MLP Outperformed The Others With Accuracy Of 96.11% For Bias, 99.33% For Drift, 98.61% For Random And 98.81% For Poly-Drift. The Average Accuracy Across The 4 Faults Is 98.21%, Which Is A 0.3% Increase From The Baseline Work 97.91%. By Adopting A Proactive Approach To Sensor Fault Prediction And Classification, Our Research Aims To Enhance The Reliability And Efficiency Of Iot Systems, Allowing For Preventive Measures To Be Taken Before Faults Have A Detrimental Impact.