In this end to end Machine Learning Project, I have decided to take on predicting cement fineness for a cement plant operating a ball mill type of system . The data used for the study was acquired from the control system through a central database system. For security reasons, the data was exported as an excel file that entails both process and quality parameters.
To give an overview, fineness is an important quality in cement as it significantly affects various properties and performance characteristics of the final product. One of the primary impacts of cement fineness is on the rate of hydration. Finer cement particles have a larger surface area in contact with water, which accelerates the chemical reactions involved in the hydration process. This leads to faster development of strength in the cement, making it possible to achieve higher early strength in concrete structures.
Additionally, cement fineness plays a crucial role in the adhesion and cohesion properties of the cement mortar. Finer particles enhance the bonding between the cement paste and the aggregates, resulting in a more uniform and compact structure. This improved cohesion not only contributes to the overall strength of the concrete but also enhances its durability and resistance to environmental factors.
Moreover, the fineness of cement can influence other properties such as workability, setting time, and shrinkage. For instance, finer cement can improve the workability of the mix, allowing for easier handling and placement. However, it can also lead to increased water demand, which must be carefully managed to avoid compromising the strength and durability of the concrete.
Overall, understanding and controlling cement fineness is essential for optimizing the performance and quality of cement-based materials. By ensuring the right level of fineness, manufacturers can produce cement that meets the specific requirements of various construction applications, ultimately contributing to the longevity and safety of the built environment.
The workflow in this project constitutes of the following :
DATA COLLECTION
Two dataset were gathered that comprised data for both process and quality. The data were then merged accordingly based on the timestamp where fineness were measured in a laboratory environment so that it will also matched the process parameters data during the period of sampling
DATA PREPARATION
Based from domain knowledge , some unimportant and highly interacting features were dropped . Some values that were also processed accordingly to correct measurement inaccuracy from the sensors . Those instances which does have values for the target feature were also dropped. Outlier data were also handled as it indicates abnormality in operation