PARFES, a Parallel Finite Element System Contributions January 1987 Dimitris Zois In this paper we describe a Parallel Finite Element System (PARFES) for the exploitation of Parallel Processing Techniques on FEM analysis for structural applications. Imperial College, Aeronautics Dept. Prince Consort Road London SW7 2BY, U.K.  Supercomputer 17, January 1987 The Finite Element Method (FEM) is a well known method for the solution of a wide range of engineering and scientific problems. Furthermore it is well known that FEM evolved concurrently with the digital computer hardware and software because the method is almost useless for practical purposes without computers. The amount of data that FEM handles is large, as well as the number of operations involved, even in the case of the Static Analysis of structures by the Displacement Method, which is the simplest problem in FEM. As new computers appear, having the resources to handle larger and more complex problems, there is an attitude to use these extended facilities as well as to explore new areas of application of FEM. Because of the physical limitations that cannot be exceeded in the construction of computers, the computing power that is required for future uses of FEM, as well as for other applications of computers, will not be provided by single processor machines. Instead of that, computers that will have more processors working in parallel on one problem are considered to provide the required power to attend these computational demanding problems. The Parallel Finite Element System (PARFES) is a software system that implements the FEM for structural analysis and runs on a MIMD type parallel computer. Furthermore PARFES follows the Communicating Sequential Processes (CSP) model of parallel computation [1]. According to CSP the processors in a parallel computer, each having its own private memory, are interconnected by a network of communication channels. Co-ordination between the processors for the solution of a problem is obtained through message-passing and not through a shared memory common to all processors.