MSF design taken into account availability

In this paper a methodology for MSF process optimization taken into account process availability will be presented. The process optimal operative conditions, and the optimal equipment allocation in stand-by units when necessary will be performed, given a process model and cost data.

In this paper a methodology for a MSF-M process optimization taken into account process availability was introduced. The process optimal operative conditions, and the optimal equipment allocation in stand-by units is achieved, given a process model and cost data. It is shown that maintenance and reliability play an important role in process operability and productivity. The problem is modeled as a MINLP problem. The stand-by units allocated in each pumping stage introduce integer variables as exponents. The problem is transformed in a MINLP program introducing binary variables and suitable restrictions which captures the more important variables of the process, and it relationships. Both, the process and the availability models are integrated and solved easily according to the proposed strategy. The model is flexible and robust. We solved the minimization of total cost, or the maximization of the process availability. Also, in both problems we can introduce the natural restrictions on a maximum available capital, or minimal availability, achieving solutions between these extreme operating conditions. Several points remains for future works. On one hand, a better, more rigorous process model must be used; not only for MSF-M, but for conventional MSF processes also. On the other hand, better models for availability must be introduced, in particular, the relationships among reparation times and corrective maintenance costs. In addition, preventive maintenance can be introduced. Also, more rigorous equations for calculation of stand-by equipment availability must be contemplated. Of course, if different process structures (a superstructure) is considered for the process design (synthesis), each option explored during the search must interact with the availability block in order to calculate the total process cost. So, a link among both models must be introduced. This also will surely introduce a much more complicated MINLP program to be solved. However, this first step proved to be adequate both, in model representation and programming efficiency to solve the problem. In future works, all of these open problems will be incorporated to the existing software step by step.