B2405 - PREDICTIVE MAINTENANCE

Learning outcomes

The course aims to provide the conceptual, methodological, and practical bases for the Predictive Maintenance of industrial equipment. At the end of the course, the student is able to apply the most advanced techniques based on Machine Learning and Neural Networks, characterizing Industry 4.0, of Condition Monitoring and Predictive Maintenance. Finally, significant cases related to industrial machines will be analyzed.

Course contents

MACHINE LEARNING INDUSTRIAL AUTOMATION

Fundamentals of Machine Learning

• Bagging, Boosting and Blended learning.
• Main algorithms such as Support Vector Machine, Random Forest, Naive Bayes and their application to learning procedure
  

Application of Machine Learning to Industrial Automation

• Condition Monitoring (CM) and Predictive Maintenance (MP) within the Smart Factory
  

• Sensorization to obtain Big Data of systems for the CM and MP: real-time measurement of mechanical, electronic, and electrical data, data on wear, overheating and consumption

• Fault Detection and Isolation, Remaining Useful Life (RUL) prediction, Fault-tolerant Control: mitigation of fault effects

• Matlab/Simulink Predictive Maintenance package illustration: programming, examples and synergistic use of the Matlab Machine Learning Toolbox

• Application of Machine Learning techniques to the diagnosis of failures in ball bearings through the use of accelerometers (real data): vibration analysis, spectrum and envelope spectrum of vibrations increase of the signal-to-noise ratio by Kurtogram, Support Vector Machine based classification of the fault (inner and/or outer race fault). Programming in Matlab/Simulink

• Application of Machine Learning Techniques to the diagnosis of faults and to the prediction of the remaining useful life of a Hydraulic Pump by means of pressure measurements (data from digital twin). Programming in Matlab/Simulink
  
  

DEEP LEARNING

• Convolutional Neural Network
• Transfer Learning: use of Google Net and Squeezenet
• Condition Monitoring by Deep Learning
  
• Application to Rolling Element Bearing Fault Diagnosis
  

NEURAL NETWORK INTELLIGENT CONTROL

Neuro-Adaptive Control

1. Fundamentals of Neural Networks: Radial Basis Function Neural Networks
2. Fundamentals of Error Learning Control Feedback
3. Fundamentals of Sliding Mode Against

Machine Vision Applications to Industrial Systems

1. "Machine Vision" and "Imaging Transformations"
2. "Multi Cameras-Based Visual Servoing for Industrial Robots"
3. "Distributed Filtering for Sensorless Control"

Readings/Bibliography

• Pedro Larrañaga et al, Industrial Applications of Machine Learning. Editore: Chapman & Hall/CRC Data Mining and Knowledge Series

Teaching methods

Frontal Lessons, laboratory, industry didactic visits.

Assessment methods

Oral Colloquium. Optional project on a topic agreed with the student.

The oral colloquiom could be taken also on-line by TEAMS or ZOOM platform.

Teaching tools

Computer, laboratory, didactic visit to local industry

Office hours

See the website of Paolo Castaldi