Week 8

Software Team:

Python implementation of storing data in MongoDB

Finish working with the Linear SVM

• Determine X and y vectors to be used (training and target vectors)
• Begin testing on data, determine what class labels are returned and compare
• Further look into model persistence

Discover Image Processing Python Libraries: Keras, OpenCV, TensorFlow

Physical use cases and company examples that are similar to our IoT framework:

• To understand how we can match our project with ‘next level’ and to see if we are on the right path

Next Steps:

Learn more about pyMongo and associated packages

Write a python script to parse several data files into one file and push data into mongoDB

Finish testing the Linear SVM to ensure it’s accuracy

• Use more data sets

-Determine whether or not it is best to save the model after it is trained so it does not need to be retrained later

• As a file - joblib (specific to scikit learn)
• As a string – pickle

Week 7

Software Team

Attempt 3 damage state model implementation (Benchmark Material Laws)

Straight Edge Fatigue Data

Get initial Probabilities

Get Transition Matrix

GaussianHMM 3 components

Final Deliverable:

RUL Predictions

Research and implement Support Vector Machines : Supervised Learning Model

Model Persistence: Save a model in scikit-learn

Make a file reader that takes the data inputted by the system and constantly updates the matrix.

Make it so the matrix is converted to a pandas dataframe after data collection is done.

How process large sets of data efficiently? [store and access] And does our project fall in-line with any corporate applications/products?

Next Steps:

Finish implementing Support Vector Machines

Understanding current Machine Learning architecture in SHM

Cornell Cup: Autodesk Fusion updates

AIRBUS proposal due Friday

Hardware Team:

To implement the three raspberry pi network successfully

Be able to setup a NAS to allow for data sharing

Next Steps:

To troubleshoot the raspberry pi’s and see what the issues are

Finish up network of raspberry pi’s and NAS setup

IoT Applications:

Smart Grid

• Implemented smart grid technology and has reduced outage times by over 50 percent, saving over $1.4 million in operations costs during a single storm.
• The implementation of a smart grid, coupled with high-speed internet infrastructure, has provided a significant economic boost to the city.
• Allowing residents to more easily participate in battery storage, contributing to greener, cleaner cities.
• Promotes integration of the vast amount of renewable energy that is currently being mandated
• Uses sensors to change brightness of lightening based on activity in certain areas

Smart Parking:

How does it work?

IoT enabled smart parking makes use of low-cost sensors, a cloud and mobile applications to give real time information about availability and location of parking space and even provide reservation or payment features.

Beneficiaries

• Drivers – Flexibility to reserve, choose and modify parking spots
• Law enforcement – Prioritize high priority enforcement and space management
• Business owner - Space management and opportunity to lure more customers.   

Week 6

Week 6 11/5/2018

Software Team

Consolidate Codes
Create Repository
.txt file to data array conversion using panda
Github created: https://github.com/Drexel-VIP-IOT
Upload Rest of Data from Dropbox to GitHub
Benchmark FileReader.py with Rakeen’s previous python code

Next Steps

Input Clustering Code into Cloud
Train Linear SVM
Input Data Projections into a Computer on the FOG
Input Linear SVM and Data Processing Code to Raspberry Pi

Hardware Team

Finished full Cloud Set up

Note: Use Raspberry Pi chain for beowulf clusters to increase computing power of EDGE

Start Raspberry Pi Mapping

Read latest Cornell Cup Proposal to better undertstand the scope of the project

Research about methods to share data over a network using Raspberry Pi and External Drive 

Next Steps

To complete raspberry pi network mapping
Setup remote connectivity to the cloud
MongoDB can now be remotely accessed outside the lab
Ran into errors for raspberry pi mapping
Looking into more documentation
Equipment has arrived
Integrate and setup the multi-raspberry pi environment
Use same process to map all pi's
Complete network attached storage between raspberry pis

Week 1-4 Review of Previous Work and Timeline

Previous work began with a Data Mining Approach to NDE Data:

Fatigue Damage Assessment Leveraging Nondestructive Evaluation Data

Initial Frameworks:

Timeline:

Week 5

HARDWARE
-Map DIC and XP PC document folders to Raspberry Pi
-Various document sharing methods
-Raspbian OS to Samba to Network Map
-Adding external disk to Raspberry Pi and sharing it over Network
-continue connection of raspberry pi with computers
Creat NAS
-Revamp of IoT framework

-MongoDB Server is live on the Cloud

Next Steps
-Prove different data stream types beside .txt
-Finish Mapping Raspberry Pi
-Test Mongo DB server

SOFTWARE
Matthew
-successfully produced PCC for correlation comparison

-Principal Component Analysis sci-kit and Matlab use normalization factors differently

-Same eigenvectors are used

-Successfully produced Gaussian Mixture Model for AE Clustering

-Data Trend Analysis for Monotonic Testing


Next Steps
-interpolative data fusion
- SVM development
-Data Quality
-combine machine learning algorithm