CME388 INDUSTRIAL SAFETY
Dr. J. Allwyn Kingsly Gladston
Department of Mechanical Engineering
UNIT – IV
FAULT TRACING
COURSE OBJECTIVES
4 To study the faults in various tools, equipments and machines.
OUTCOMES:
At the end of the course the students would be able to
4. Evaluate faults in various tools, equipments and machines
UNIT – IV
FAULT TRACING
FAULT TRACING 9
Fault tracing-concept and importance, decision tree concept,
need and applications, sequence of fault-finding activities, show
as decision tree, draw decision tree for problems in machine
tools, hydraulic, pneumatic, automotive, thermal and electrical
equipment’s like, i. Any one machine tool, ii. Pump iii. Air
compressor, iv. Internal combustion engine, v. Boiler, vi.
Electrical motors, Types of faults in machine tools and their
general causes.
DECISION TREE FOR A FAULT IN MILLING
MACHINE
1. Healthy and Faulty Face Milling Tool Conditions (Acceleration
Signals)
What happens:
Data is collected from the milling machine under both normal
(healthy) and various faulty conditions (such as chipping,
breakage, and flank wear).
Data type:
Acceleration signals, typically captured using vibration
sensors, are used as the primary input. These signals reflect
the dynamic behavior of the tool and can indicate anomalies.
2. Data Acquisition System (NI DAQ-9234)
What happens:
The acceleration signals are fed into a data acquisition system
(specifically, the NI DAQ-9234 module), which digitizes and
DECISION TREE FOR A FAULT IN MILLING
MACHINE
3. Statistical Features Extraction
What happens:
The raw acceleration signals are processed to extract
statistical features.
Examples of features:
Mean
Standard deviation
Skewness
Kurtosis
Root mean square (RMS)
These features help summarize the signal characteristics and
are used as inputs for the next steps.
DECISION TREE FOR A FAULT IN MILLING
MACHINE
4. Feature Selection Using Decision Tree (J48 Algorithm)
What happens:
The extracted statistical features are evaluated using a
decision tree algorithm (specifically, the J48 algorithm, which
is an implementation of the C4.5 decision tree).
Purpose:
The decision tree selects the most relevant features that are
effective in distinguishing between healthy and faulty tool
conditions. This reduces dimensionality and improves
classification accuracy.
DECISION TREE FOR A FAULT IN MILLING
MACHINE
5. Splitting Data: Training and Testing Data Sets
What happens:
The dataset is divided into two parts:
Training set: Used to train the classification model.
Testing set: Used to evaluate the model’s performance on unseen
data.
6. Training of Naïve Bayes
What happens:
The selected features from the training set are used to train a
Naïve Bayes classifier.
Naïve Bayes:
A probabilistic machine learning model that predicts the class
of a sample based on the likelihood of its features.
DECISION TREE FOR A FAULT IN MILLING
MACHINE
7. Model Training Check
What happens:
The system checks if the Naïve Bayes model is trained
successfully.
If No, it loops back and continues training.
If Yes, it proceeds to the next step.
DECISION TREE FOR A FAULT IN MILLING
MACHINE
8. Diagnosis of Face Milling Tool Condition
What happens:
The trained Naïve Bayes model is used to diagnose the
condition of the face milling tool on new, unseen data (from
the testing set).
Possible diagnoses:
Healthy
Chipping
Breakage
Flank wear
The classifier provides an automatic and reliable diagnosis
based on the statistical features of the vibration signals.
DECISION TREE FOR A
FAULT IN
MILLING MACHINE
