Process optimization cannot be done without continuous process measure and increasing plant margin cannot be done without continuous process optimization. Normally, products are kept within the specifications by using online monitoring or some time using laboratory off-line sampling. The second method practice has a huge time delay between process monitoring and control which add more challenges for process optimization. This work aims to come up with cost-effective solution to measure and predict the physical properties of distillate column in Ethylene Oxide (EO) plant, continuously, without having any additional on-line stream analyzers. Here, we apply soft sensors model to provide process optimization solution. The model output can be integrated with a Distributed Control System (DCS) or Model Predictive Control (MPC) system to minimize product giveaway and energy consumption. In this work, we will collect actual plant dataset and conduct the required training to come with the right model to be used for future predication to monitor plant process without any additional hardware and we will show the process of applying soft sensor. One of Artificial Intelligent solution is applying a neural network. Various structures of neural networks with different numbers of neurons in each hidden layer were created and assessed for their performance on the estimation of product composition in the distillate stream. Other solution of Artificial Intelligent is using regression model (Ordinary Least Squares OLS, Weighted Least Squares - WLS and Partial Least Squares PLS) based on soft sensor for online measurement which were developed in this work. The performance for each developed method was tested on the estimation of Ethylene Oxide composition in the distillate stream. The simulation results showed that the neural network has most accurate prediction. So, it will be the best choice to use for EO composition prediction, however, due to company recourses and restriction we hav...