leveraging machine learning for ML tools and approaches to enhance decision-making, efficiency, or performance in a specific domain.

Senior Design Project
Review-1 Presentation
[Leveraging Machine Learning for Predictive
Maintenance in Industrial Equipment ]
Supervised By: Prof. Swagat Kumar Jena
Group No.: B5
Name of the Student(s) with Regd. No.:
Praveen Sharma - 2141019272
Prince Kumar - 2141011213
Rahul Kumar Mahto - 2141013045
Lohit Mohanty - 2141013174
Department of Computer Sc. and
Engineering
Faculty of Engineering &Technology (ITER)
Siksha ‘O’ Anusandhan (Deemed to be) University
Bhubaneswar, Odisha
1

Presentation Outline
 Introduction
 Project Overview
 Problem Statement & Motivations
 Objectives & Expected Impact
 Background & Related Work/ Literature Review
 Existing Solutions/Related Work & Their Limitations
 Improvements Over Existing Solutions
 Proposed Solution & Architecture
 System Architecture /Workflow Diagram /Model Diagram /Block Diagram/Schematic Layout
 Key Components/Features & Modules
 Algorithms and Methods Used/Technologies, Frameworks, and Tools Used
 Progress in Implementation Plan & Methodology
 Summary
 Bibliography
2

Introduction
Project Overview
This project aims to create a predictive maintenance system using
machine learning to forecast equipment failures with high accuracy.
It focuses on optimizing computational efficiency while maintaining
strong predictive performance, reducing downtime, lowering
maintenance costs, and improving industrial operations through
proactive, data-driven decision-making using IoT sensor data and
advanced algorithms.
3

Introduction
Problem Statement & Motivations
Problem Statement
Design a predictive maintenance system that ensures a balance of performance
and computational efficiency by leveraging IoT sensor data and evaluating
multiple machine learning models.
Motivations
•Downtime due to equipment failures is costly and disruptive.
•Traditional maintenance strategies are either reactive (run-to-failure) or overly
resource-intensive (preventive).
•Machine learning and IoT present an opportunity for proactive, cost-effective, and
highly accurate maintenance systems. 4

Introduction
Objectives
•Compare multiple machine learning models for fault prediction.
•Identify the most reliable and cost-effective model.
•Use IoT sensor data to provide actionable maintenance insights.
Expected Impact
•Reduction in downtime and maintenance costs.
•More efficient and reliable industrial equipment operations.
•Data-driven decision-making for maintenance planning.
5

Background & Related Work
Related Work & Their Limitations
Literature Survey
•Paper 1: LightGBM achieves high accuracy but is computationally intensive.
•Paper 2: Random Forest is reliable but costly in large-scale applications.
•Paper 3: CatBoost is efficient for imbalanced data but needs further evaluation.
•Paper 4: MCC is underexplored in predictive maintenance systems.
Identified Gaps
•Existing models do not balance computational efficiency and predictive
performance well.
•MCC is an underutilized metric for evaluating predictive maintenance models.
•Few studies compare models based on Accuracy, F1-Score, and MCC.
6

Existing Solutions & Their Limitations
•Traditional maintenance strategies: Run-to-failure and preventive
maintenance.
•Machine learning models like LightGBM, Random Forest, and CatBoost
have strengths and weaknesses.
•Computational efficiency and model reliability are not always balanced.
•Existing predictive maintenance models often struggle with handling
imbalanced datasets, leading to inaccurate fault detection.
•Many current solutions lack real-time integration with IoT sensors,
limiting their ability to provide dynamic, on-the-spot maintenance
insights.
7

Proposed Solution & Architecture
1.Data Collection: Sensor data on air temperature, process
temperature, rotational speed, torque, and tool wear.
2.Data Preprocessing: Cleaning, checking for imbalance, and feature
extraction.
3.Model Evaluation: Comparing Random Forest, XGBoost, LightGBM,
and CatBoost.
4.Actionable Insights: Visualizing results to select the best model.
8

System Architecture/ Workflow Diagram
9

Key Components/ Features & Modules
Key Components
1.Sensor Data Collection – Gathers real-time data on air temperature, process
temperature, rotational speed, torque, and tool wear.
2.Data Preprocessing – Cleans data, handles missing values, and balances
class distribution.
3.Model Training & Evaluation – Compares machine learning models (Random
Forest, XGBoost, LightGBM, CatBoost) using Accuracy, F1-Score, and MCC.
Features & Modules
•Fault Prediction Module – Uses trained ML models to predict potential failures.
•Performance Monitoring Module – Tracks model accuracy and reliability over
time.
•Visualization Dashboard – Displays real-time sensor data and model
predictions for easy decision-making.
10

Algorithms and Methods Used/ Technologies,
Frameworks, and Tools Used
Algorithms and Methods Used
•Machine Learning Models: Random Forest, XGBoost, LightGBM,
CatBoost.
•Evaluation Metrics: Accuracy, F1-Score, MCC.
Technologies, Frameworks, and Tools Used
•Python, Scikit-Learn, TensorFlow, IoT Sensors, Jupyter Notebook.
11

Progress in Implementation Plan
12

Methodology
Each model is trained independently on
historical sensor data from industrial
equipment, combines multiple decision
trees to reduce errors from individual
models, leading to more accurate
predictions for complex equipment
data.
13

Summary
This project develops a machine learning-based predictive maintenance
system to reduce industrial equipment failures and downtime. Using IoT
sensor data, models like Random Forest, XGBoost, LightGBM, and
CatBoost are evaluated for reliability and cost-effectiveness. LightGBM
excels in fault detection, Random Forest ensures consistency, and
CatBoost is the most efficient. The system enhances proactive
maintenance, reducing costs and improving efficiency. Future work
includes real-time IoT integration and model optimization.
14

Bibliography
1. Ni, F., Zang, H. & Qiao, Y. (2024). Smartfix: Leveraging
Machine Learning for Proactive Equipment Maintenance in
Industry 4.0.
2. Vago, N.O.P., Forbicini, F. & Fraternali, P. (2024). Predicting
Machine Failures from Multivariate Time Series: An Industrial
Case Study.
3. Chandu, H.S. (2024). Enhancing Manufacturing Efficiency:
Predictive Maintenance Models Utilizing IoT Sensor Data.
4. Arunkumar, G. (2024). AI-Based Predictive Maintenance
Strategies for Electrical Equipment and Power Networks.

leveraging machine learning for ML tools and approaches to enhance decision-making, efficiency, or performance in a specific domain.

More Related Content

Similar to leveraging machine learning for ML tools and approaches to enhance decision-making, efficiency, or performance in a specific domain. (20)

Recently uploaded (20)

leveraging machine learning for ML tools and approaches to enhance decision-making, efficiency, or performance in a specific domain.