An object oriented process simulation environment was developed using object oriented programming. It consists of a sequential modular simulator, an equation-based simulator, a physical quantity system, a physical property system and a databank. The modified Powell's dogleg method was evaluated and found to be a reliable nonlinear equation solver for both dense and sparse equations. Two approaches of keeping unknowns in the feasible region were evaluated. They performed remarkably well. For the update of the sparse Jacobian matrix, Schubert's update and Bogle's update were both effective. We extended the root solving technique of Topliss and it performed very well in the computation of physical properties in process simulation. Strategies proposed by Mathias et al. were used in the extension. Working equations for computing densities in severe conditions where real densities do not exist were presented. Physical properties computed were found to be continuous and they also maintained the original trends of the real physical properties. Gundersen's algorithm and Li's algorithm for tearing were studied and they performed poorly in the thirteen flowsheets studied. We proposed a new tearing algorithm that does not use the cycle matrix for tearing and it efficiently produced the optimal tear set for all problems studied in this work. The convergence behavior of sequential modular simulators for constrained simulation was affected by numerical methods used in converging tear streams. Equation-based simulators can not outperform the popular sequential modular simulators without a very robust numerical method. The current best numerical method in an equation-based simulator still needs user input in guiding the solution process. This makes the simulator difficult to use successfully.