Scribd
Upload
2 views

Previous AI Projects - 10 Sample Projects

The document outlines various AI and machine learning projects in manufacturing, including surface quality inspection, automated quality control, predictive maintenance, and more. Each project details the methodologies, software used, and datasets involved, showcasing applications of CNNs, IoT, and sensor data analysis. The projects aim to enhance quality assurance, maintenance prediction, and defect detection in manufacturing processes.

Uploaded by

me22d032
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
2 views

Previous AI Projects - 10 Sample Projects

The document outlines various AI and machine learning projects in manufacturing, including surface quality inspection, automated quality control, predictive maintenance, and more. Each project details the methodologies, software used, and datasets involved, showcasing applications of CNNs, IoT, and sensor data analysis. The projects aim to enhance quality assurance, maintenance prediction, and defect detection in manufacturing processes.

Uploaded by

me22d032
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

AI in Manufacturing

ME6324
Previous Projects
1. Surface quality inspection using Convolutional neural networks
Built a deep convolutional neural network based on the popular VGG net or Resnet.
Retrained this network on the dataset of surface finish images of different machines products.
o The network takes in images of machined product surface as shown in figure and
classifies this surface finish as either good or bad accordingly.

 Software Used
o tensorflow, Python
o openCV libraries

 Dataset had been taken


from Kaggle.
2. Automated Quality control using AI and IoT

Developed a system to detect the key quality indicators for different products.
Database of goods had been created with the details of their compliances.
Database had created with the product code and quality rating mapped inside a central system.

Software and techniques used:


o Line Scan cameras and backlit conveyors are
used for imaging and OpenCV for reading images
and tensorflow for creating the CNN.
o Anomaly detection techniques to identify the defects
in the products. (Kalman filter Anomaly detection
Temporal Anomaly Detection )
3. Multi sensor fusion method for surface finish
prediction in grinding process
Multi sensor fusion based approach has been used to see the physical relationship between the
operational parameters and the output i.e. Surface quality and improve surface quality
assurance.
Time Domain, Frequency Domain and Time- Frequency Domain Features had extracted.

Subsequently a Data Driven Model had been developed to predict Ra (Surface Roughness)
using the features of the extracted signals.
The Accuracy has been judged under various test conditions using value.

Software Used:
o For Extracting the Features – MATLAB 2018.
o For making the Data Driven Prediction Model – Python 3.6
4. Predictive Maintenance Models using Machine Learning
 Predicted the time in which aircraft engines fail based on sensor data, so that maintenance activity can be
carried out at the optimal time.

 4 data sets has been prepared corresponding to different operational conditions and fault modes (HPC
Degradation, Fan Degradation). The data is contaminated with sensor noise.

 Software Used:
o Python, Numpy, Scipy, Pandas, Scikit, Pytroch, MATLAB

 Algorithm Used
o Decision Tree
o Regression
o Clustering Algorithm
o Time Series algorithm
5. CNC tool wear prediction ML methods
Pattern recognition had been used to predict both nature of wear and inadequate
clamping.

Relevant parameters, to used for statistical learning, can include but weren’t limited to:
velocity, acceleration, position, power supplied, feed rate, clamp pressure, material etc.
Identified patterns in CNC tool data. Dataset will be obtained from open source
manufacturing repositories.

Software used:
ML: Python( NumPy, sklearn, Pandas)
Deep learning : Tensorflow, Keras, pytorch
Time series Analysis : R(ASTSA, Forecast)
6. Age and Gender estimation using facial Recognition
Approach
o Dataset
o Assignation of facial features to an age group eg: wrinkles
o Image pre-processing
o Algorithm to be used CNN’s such as mobile net or resnet

Libraries and software used:


o Keras library
o Tensorflow
o OpenCV
o Python
7. Adaptive positioning CoBots
Packages used
o Tensor flow python
o Keras, NumPy, Pandas
o MATLAB
o Arduino

Data Set:
o Collected the data set comprising of pictures of human hands from various sources online and had also
added their own pictures to it and using data augmentation.

Yolo algorithm had been used as the state of art to localisation and classification.
Trained the algorithm on their augmented dataset comprising of human hands.
8. LICENSE PLATE/NUMBER PLATE RECOGNISATION
Process Involved:
1)IMAGE ACQUISATION:
• Getting the car/motorcycle image using a mobile camera(input)
2) LICENSE PLATE EXTRACTION:
• Using different boundary and edge conditions/colour and characters constituting it.
(adaboost classifier )
3) LICENSE PLATE SEGMENTATION:
• Done by using character contours(neural network assistance)
4)CHARACTER RECOGNISATION:
• Using neural network for character recognition.

Software used :
o PYTHON/MATLAB(neural network toolbox)
9. Optical character recognition
Tools used
o OpenCV
o Google API
o MATLAB Image Processing
o Python
The proposed system was OCR ON A GRID INFRASTRUCTURE which has a
character recognition system that supports recognition of the characters including
alphabets & numerals.

This feature is what call as grid infrastructure which eliminates the problem of
heterogeneous character recognition. In this context, Grid infrastructure means the
infrastructure that supports group of specific set of characters.
10. Estimation of PM (Particulate matter) concentration
using Optical sensor data at boundaries
Optical monitor PM sensors had been used which uses light source and photo diode positioned at
an angle.
An Arduino based board – with low cost PM sensor, Temperature sensor and SD (Secure Digital)
card module has programmed to collect Temperature, PM2.5 and PM10 concentration data into
a CSV file.

MATLAB toolbox had been used for training


the model, with input and output variables as
data (temperature and PM2.5 and PM10
concentration) from respective
sensors.
11. Detecting malaria cells using CNN
Visual detection of malaria infected cells is used for detection and report the infected person.

Process Involved:
o Data Pre-processing
o Basic ML algorithm
o CNN

Software used
o Python
o Sklearn, keras, pytroch, tensorflow
o Numpy, scipy

Data Source - Kaggle repository

You might also like