iNTERIOR_MAIN1

iNTERIOR_MAIN2
iNTERIOR_MAIN3
MOTOR_ENGINE_1
MOTOR_ENGINE_2
MOTOR_ENGINE_3

SOP -> EOP

keepware

MQTT ENGINE : Adds the functionality to the Ignition SCADA system platform to bidirectional
communicate with MQTT enabled edge-of-the-network devices securely via an MQTT Server.
CYBERARK
collection des donner et la visualtion aa travers SCADA

Project Title: "Real-time Monitoring and Optimization of Cable Manufacturing Processes using
IoT and Data Analytics"

Project Description:

Objective: The objective of this project is a comprehensive system for real-time monitoring and
optimization of cable manufacturing processes at Aptiv using IoT devices, SCADA systems,
MQTT engine, and data analytics techniques.

Components of the Project:

IoT Integration: Implement IoT sensors and devices at key points in the cable manufacturing
process to gather real-time data on parameters such as temperature, pressure, speed, and quality
metrics.

SCADA System Enhancement: Enhance the existing SCADA system to integrate with the IoT
devices and collect data from them. This will enable real-time visualization and monitoring of the
manufacturing processes.

MQTT Engine Integration: Utilize MQTT protocol to efficiently transport data between IoT
devices, SCADA system, and other components of the infrastructure. Implement an MQTT
engine for reliable communication and data exchange.

Data Analytics Pipeline: Develop a data analytics pipeline to process the raw data collected from
IoT devices and SCADA system. Apply machine learning algorithms and statistical analysis
techniques to identify patterns, anomalies, and opportunities for optimization.

Dashboard Development: Create interactive dashboards and visualization tools to present the
analyzed data in a user-friendly manner. These dashboards should provide real-time insights into
the manufacturing processes, highlighting areas of concern and opportunities for improvement.

Predictive Maintenance: Implement predictive maintenance algorithms to forecast equipment

failures and prevent downtime. Utilize historical data and machine learning models to predict
maintenance needs and schedule proactive interventions.

Continuous Improvement: Establish a feedback loop for continuous improvement by integrating

feedback mechanisms into the system. Enable operators and engineers to provide feedback on
process performance and suggest improvements based on real-world observations.

Benefits of the Project:

Improved visibility and transparency into the cable manufacturing processes.

Real-time monitoring and early detection of quality issues or equipment failures.
Increased efficiency and productivity through data-driven optimization.
Enhanced predictive capabilities for maintenance planning and resource allocation.
Facilitation of continuous improvement initiatives based on actionable insights and feedback.
By working on this project, you'll have the opportunity to gain hands-on experience with cutting-
edge technologies in the field of industrial IoT, data analytics, and digitalization. Moreover, you'll
contribute to enhancing the efficiency and competitiveness of Aptiv's manufacturing operations
while fulfilling the requirements for your master's degree internship.

Certainly! Predictive maintenance using machine learning involves building models that can
predict when equipment is likely to fail or require maintenance based on historical data and
various features related to equipment health and performance. Here's a more detailed explanation
of how machine learning can be used in predictive maintenance:

Data Collection and Preprocessing:

Gather historical data from sensors, SCADA systems, maintenance logs, and other sources. This
data typically includes variables such as temperature, pressure, vibration, run-time hours, and
maintenance records.
Preprocess the data to handle missing values, outliers, and inconsistencies. This may involve
techniques such as data cleaning, normalization, and feature scaling.
Feature Engineering:

Identify relevant features (variables) that may impact equipment health and performance. This
could include raw sensor readings, derived features (e.g., rolling averages, trends), and contextual
information (e.g., time of day, production schedules).
Engineer additional features if necessary to capture complex relationships and patterns in the
data.
Labeling Data:

Define the target variable or label for predictive maintenance. This could be a binary variable
indicating whether a failure occurred within a certain time window (e.g., next week, next month).
Model Selection:
Choose appropriate machine learning algorithms for predictive maintenance. Commonly used
algorithms include:
Logistic Regression: Suitable for binary classification tasks.
Decision Trees and Random Forests: Capable of capturing complex interactions and non-linear
relationships.
Support Vector Machines (SVM): Effective for small to medium-sized datasets with high-
dimensional feature spaces.
Gradient Boosting Machines (GBM) or XGBoost: Ensemble methods that often perform well in
predictive maintenance scenarios.
Deep Learning: Neural networks can be used for more complex data patterns but may require
larger amounts of data and computational resources.
Training the Model:

Split the dataset into training and testing sets to evaluate the performance of the model.
Train the selected machine learning model using the training data. Adjust hyperparameters as
needed to optimize model performance.
Evaluate the trained model using the testing data to assess its accuracy, precision, recall, and
other relevant metrics.
Deployment and Integration:

Deploy the trained predictive maintenance model into the production environment.
Integrate the model with the existing SCADA system, IoT infrastructure, and data analytics
pipeline for real-time monitoring and decision-making.
Alerting and Actionable Insights:

Set up alerting mechanisms to notify maintenance personnel or operators when the predictive
model detects potential equipment failures.
Provide actionable insights and recommendations based on the model predictions to facilitate
proactive maintenance planning and scheduling.
Continuous Monitoring and Updating:
Monitor the performance of the predictive maintenance model in real-time.
Periodically retrain the model using new data to ensure its accuracy and effectiveness over time.
Feedback Loop and Improvement:

Establish a feedback loop to collect data on the outcomes of maintenance actions taken based on
the model predictions.
Use this feedback to refine the model, improve its accuracy, and enhance predictive capabilities
over time.
By leveraging machine learning techniques, predictive maintenance can help Aptiv optimize
maintenance activities, reduce downtime, and improve overall equipment reliability and
performance in the cable manufacturing processes.

-----------------------------------------------------

Data Collection and Preprocessing:

Gather historical data from sensors, SCADA systems, maintenance logs, and other sources. This
data typically includes variables such as temperature, pressure, vibration, run-time hours, and
maintenance records.
Preprocess the data to handle missing values, outliers, and inconsistencies. This may involve
techniques such as data cleaning, normalization, and feature scaling.
Feature Engineering:

Identify relevant features (variables) that may impact equipment health and performance. This
could include raw sensor readings, derived features (e.g., rolling averages, trends), and contextual
information (e.g., time of day, production schedules).
Engineer additional features if necessary to capture complex relationships and patterns in the
data.
Labeling Data:

Define the target variable or label for predictive maintenance. This could be a binary variable
indicating whether a failure occurred within a certain time window (e.g., next week, next month).
Model Selection:

Choose appropriate machine learning algorithms for predictive maintenance. Commonly used
algorithms include:
Logistic Regression: Suitable for binary classification tasks.
Decision Trees and Random Forests: Capable of capturing complex interactions and non-linear
relationships.
Support Vector Machines (SVM): Effective for small to medium-sized datasets with high-
dimensional feature spaces.
Gradient Boosting Machines (GBM) or XGBoost: Ensemble methods that often perform well in
predictive maintenance scenarios.
Deep Learning: Neural networks can be used for more complex data patterns but may require
larger amounts of data and computational resources.
Training the Model:

Split the dataset into training and testing sets to evaluate the performance of the model.
Train the selected machine learning model using the training data. Adjust hyperparameters as
needed to optimize model performance.
Evaluate the trained model using the testing data to assess its accuracy, precision, recall, and
other relevant metrics.
Deployment and Integration:

Deploy the trained predictive maintenance model into the production environment.
Integrate the model with the existing SCADA system, IoT infrastructure, and data analytics
pipeline for real-time monitoring and decision-making.
Alerting and Actionable Insights:

Set up alerting mechanisms to notify maintenance personnel or operators when the predictive
model detects potential equipment failures.
Provide actionable insights and recommendations based on the model predictions to facilitate
proactive maintenance planning and scheduling.
Continuous Monitoring and Updating:

Monitor the performance of the predictive maintenance model in real-time.

Periodically retrain the model using new data to ensure its accuracy and effectiveness over time.
Feedback Loop and Improvement:

Establish a feedback loop to collect data on the outcomes of maintenance actions taken based on
the model predictions.
Use this feedback to refine the model, improve its accuracy, and enhance predictive capabilities
over time.
By leveraging machine learning techniques, predictive maintenance can help Aptiv optimize
maintenance activities, reduce downtime, and improve overall equipment reliability and
performance in the cable manufacturing processes.

------------------------------------------------------