Government Engineering College Majali,

Karwar

Technical Seminar on

MACHINE LEARNING IN PREDICTIVE

MAINTENANCE
Submitted by Coordinator
Nishanth s Dr.Srinivasamurthy M
2GP21ME007 Head of the deept
Dept of Mechanical Engineering
Contents
• Introduction
• Key Components of ML-Based Predictive Maintenance
• Literature Survey
• Role of ML in Predictive Maintenance
• Types of ML Models in PDM
• Anomaly Detection & Failure Prediction
• Root Cause Analysis
• Implementing ML-Based Predictive Maintenance
• Data Collection and Preprocessing
• Conclusion
• References
Introduction

• Predictive maintenance leverages machine

learning to foresee equipment failures
before they occur.
• It shifts maintenance strategies from
reactive or scheduled maintenance to
condition-based monitoring.
• Machine learning enhances accuracy by
analyzing sensor data and identifying
patterns in equipment degradation
Key Components of ML-Based Predictive
Maintenance
• Data Collection: Gathering real-time sensor data, historical
maintenance logs, operational metrics.
• Data Pre-processing: Handling missing data, noise removal, feature
extraction.
• Machine Learning Models: Supervised, unsupervised, deep learning
techniques for failure prediction.
• Model Deployment: Real-time monitoring and alert systems for
proactive maintenance.
OBJECTIVES
1. Introduce Predictive Maintenance (PdM):
Define predictive maintenance and explain its importance in modern
industries.
2. Understand the Role of Machine Learning in PdM:
Explore how machine learning techniques are applied to predict
equipment failures before they occur.
3. Highlight Key Machine Learning Models Used:
Discuss common algorithms such as regression, classification, anomaly
detection, and deep learning in the context of maintenance.
4. Explore Real-World Use Cases:
Review case studies from manufacturing, transportation, energy, etc.,
where ML-based PdM has improved efficiency and reduced costs.
 LITERATURE SURVEY
YEAR OF
AUTOUR/REFERANCE JOURNAL/CONFERANCE
PUBLICACATION
"A Survey of Deep Learning- 2024
Driven Architecture for Predictive
Lima et al. Maintenance" This study reviews
deep learning architectures applied
to predictive maintenance
"Advancements in Machine 2024
Levine Learning for Predictive
Maintenance in Industrial
Systems"
. – "Machine Learning Applications 2023
Miraje et al. for Predictive Maintenance in
Mechanical Systems: Case Studies,
Algorithms, and Performance
Evaluation"
 Role of ML in Predictive Maintenance

• Early Fault Detection – ML algorithms analyze sensor data to detect

anomalies and predict potential failures before they occur, minimizing
downtime and repair costs.
• Optimized Maintenance Scheduling – Predictive models help
industries shift from reactive or scheduled maintenance to data-driven
maintenance, reducing unnecessary inspections and interventions.
• Real-Time Monitoring and Alerts – ML-powered predictive
maintenance systems continuously process real-time data from IoT
sensors, providing instant alerts for potential issues.
Types of ML Models in PDM
• Supervised Learning Models – These models use labeled historical data to
predict failures. Examples include Decision Trees, Support Vector Machines
(SVM), and Random Forest.
• Unsupervised Learning Models – These models identify patterns and anomalies
in unlabelled data. Techniques like k-Means Clustering, Auto encoders.
• Semi-Supervised Learning Models – Combining elements of both supervised
and unsupervised learning, these models use a small amount of labeled data along
with large unlabeled datasets to improve failure detection and anomaly
identification.
• Deep Learning Models – Advanced neural network-based models such as
Convolutional Neural Networks and Long Short-Term Memory networks analyze
complex time-series sensor data for accurate failure prediction.
Anomaly Detection & Failure Prediction
• Statistical Methods – Identifies deviations using techniques like Z-
score, moving averages, and standard deviation thresholds.
• Machine Learning Models – Uses supervised and unsupervised
learning methods (e.g., Isolation Forest, Autoencoders, k-Means
Clustering) to detect anomalies in sensor data
• Sensor Fusion – Combines multiple sensor for a more accurate
anomaly detection approach.
Data Collection and Pre-processing
• Data collection and preprocessing are critical steps in implementing
ML-based predictive maintenance, ensuring accurate and reliable
model performance. Industrial machinery generates vast amounts of
data through IoT sensors, SCADA systems, and Enterprise Asset
Management (EAM) platforms. This data includes time-series sensor
readings (e.g., temperature, vibration, and pressure), historical
maintenance logs, operational conditions, and environmental factors.
Root Cause Analysis
• Root Cause Analysis (RCA) is a systematic
approach used to identify the underlying
causes of failures in industrial systems. By
pinpointing the root cause, organizations
can implement effective corrective actions
to prevent recurring issues and optimize
maintenance strategies.
Fishbone Diagram (Ishikawa Diagram:- Visualizes potential failure
causes categorized into factors like environment, process, material, and
human errors.
Implementing ML-Based Predictive
Maintenance
Implementing ML-based predictive maintenance involves several key
steps to ensure efficient fault detection and failure prevention. First, data
collection is crucial, as IoT sensors and industrial systems generate vast
amounts of real-time operational data, including vibration, temperature,
and pressure readings. This data is preprocessed to remove noise, handle
missing values, and extract relevant features.
Adavantages in ml in pdm
• Reduced Downtime
• Cost Savings
• Optimized Maintenance Scheduling
• Early Fault Detection
• Data-Driven Decision-Making
• Faster Decision-Making
Disadvantages in ml in pdm
• High Initial Investment
• Complex Implementation
• Data Quality Issues
• Model Drift
• Complex Model Training
• High Computational Requirements
Conclusion
Machine learning has revolutionized predictive maintenance by
enabling early fault detection, optimizing maintenance schedules, and
reducing downtime. By leveraging data-driven insights, organizations
can improve asset reliability, extend equipment lifespan, and lower
maintenance costs. However, challenges such as data quality, integration
complexity, and cybersecurity risks must be addressed for effective
implementation
References
• Lee, J., Wu, F., Zhao, W., Ghaffari, M., Liao, L., & Siegel, D. (2014).
Prognostics and health management design for rotary machinery
systems—Reviews, methodology and applications. Mechanical
Systems and Signal Processing, 42(1), 314-334.
• Zhang, W., Yang, D., & Wang, H. (2019). Data-driven methods for
predictive maintenance of industrial equipment: A survey. IEEE
Systems Journal, 13(3), 2213-2227.
• Si, X., Wang, W., Hu, C., & Zhou, D. (2011). Remaining useful life
estimation – A review on the statistical data-driven approaches.
European Journal of Operational Research, 213(1), 1-14.
THANK YOU