# Present Smart Horsepower Per Tonne (SHPT) has following restrictions:
- Lead can never be Isolated - Fuel Imbalance
- Need eMU (Ethernet Over Multiple Units) Hardware
- Remote Consist cannot be Isolated
- Loco Health Not considered - Sometimes could lead to mission failure
- Locomotive Operation Efficiency Not Considered Locomotive Control System is already at high CPU Utilization value. The aim to solve all above problem with just one platform was inspiration for this topic.
What it does
The Brain of SHPT used to reside on Lead Locomotive but with the proposed architecture it will be moved to cloud platform where it can have integration with weather website, locomotive health from offboard, estimation model of locomotive ton-mile/gallon efficiency & status data from locomotive to decide which locomotive in consist can be ISOLATED or DB_ONLY.
How we built it
We used Raspberry Pi to simulate data from Locomotive consist, UI on Web to simulate data coming from Weather Website & Loco Health from Offboard. Using TimeSeries data from various sources are send to Predix Cloud. On Predix Cloud Adhesion Estimation, Locomotive Tonn-Mile/Gallon & RHP/FBR efficiency are calculated using Linear Regression Model. Based on all these inputs decision to ISOLATE/DB_ONLY/RUN data to send to respective locomotives.
Challenges we ran into
- Using TimeSeries to send data into Predix Cloud
- Binding Python Analytics App on Cloud
- Sending/Receiving data in TimeSeries using combination of Python, JAVA & Qt.
Accomplishments that we're proud of
- We have implemented end to end system where Raspberry Pi send data as Locomotive Consist & we are receiving commands in real time from Predix. All this done in 24 hours without prior knowledge of Predix
What we learned
- We learned to send data from different data sources to Predix Cloud.
- Running Analytics App using Python on Predix
- Team effort & bonding yield better results.
What's next for Cloud Controlled-Smart Horse Power Per Tonne
The next plan is to develop more advanced analytics algorithm and train it with real data to quantify the benefits of using proposed architecture vs present smart horse power per tonne.