Team_Neurons

Inspiration

Industrial Metal Furnaces are used to hold the metal until it liquefies (Metal Bath). The controlled environment inside the furnace heats up the metal to its melting temperature where it starts to melt and reaches where there is no solid metal in it(Flat bath). It isn't easy in determining the temperature of the metal(Metal Temperature) based on the Temperature inside the furnace(Roof Temperature). Opening the Door of the furnace to measure the temperature of the metal before the cycle completion, imbalances the air pressure causing heat loss that built up, dross formation, and increases the oxidation in the chamber. If the metal is not achieved flat bath – requires heating the furnace again which again causes energy loss and time loss. If there is a ** method to predict the Metal Temperature from Roof Temperature, this loss can be avoided by opening the furnace on time.**

What it does

Create a Machine Learning Model which predicts the Metal Temperature from the Roof Temperature to determine whether the metal achieved the flat bath or not.

How we built it

The Data Consist of 33280 rows and 6 columns. Preprocessing reduces the data to 3 columns and 1470 rows(relevant data). Relevant Columns: Minutes(transformed DateTime column), Roof Temperature and Metal Temperature Assumptions:

The door status is always closed
The Burner Temperature and Roof Temperature are always constant for the cycle.

Challenges we ran into

Insufficient Data in each cycle to train and test the model.
The Target value has a lot of missing data points corresponding to the Roof Temperature
It is difficult to consider the data as time series(Each cycle starts at a different time and no information about the date)
No Data on the nature of metal as the rate of transfer of heat depends on the parameters like Area, Thickness, Thermal conductivity and so on.

Accomplishments that we're proud of

With the given model, Metal Temperature can be predicted and Energy Loss with respect to the Metallurgy cycle can be reduced with this prediction. This brings the industry saves a lot of time and money