Model based hazard identification: Process time accelerated by GPU redesigning approach

摘要

Process safety and risk assessment are major requirements in the industrial context and hazard identification is essential for ensuring safe design and operation of a process. Numerous automated software approaches to risk assessment have resulted in many improvements in human brainstorming techniques of conventional risk assessment. In terms of computing time as an important aspect of these automated tools, mathematical simulation of physical and chemical states of the process is most time consuming in comparison to results collection and evaluation. In this context, GPU parallel computing has many advantages which meet the demanding requirements on complex and precise process hazard analysis without the involvement of massive processing architectures. This paper presents an efficient low-cost way of significant acceleration of targeted prediction of incident consequences by dynamic simulation of process fault deviations in the context of safety analysis. The GPU based simulation computing algorithm acceleration has been demonstrated on a hazard and operability analysis of propylene glycol production carried out in a closed loop CSTR (continuous stirred tank reactor). In this application, HAZOP multi-parameter process deviations have been simulated 100 times faster than when using its CPU code version.