An ISL-8 and ISL-15 study of the physiological simulation benchmark experiment

Abstract

PHYSBE is an acroynm for PHYsiological Simulation Benchmark Experiment, and was introduced1 as a suggested standard for evaluating tools and tech niques available for physiological simulation. The mathematical model relates pressure, volume, flow, and resistance in a simplified cardiovascular system. PHYSBE is an excellent benchmark problem because it is not easy to obtain accurate results with an all-analog or all-digital simulation. The large differences in time-dependent constants (1.63 to 90.0) allow a good check of digital integration routines. Roundoff accumulation errors can be devastating to digital solutions. On the other hand the compliance generators (the driving func tions) are difficult to program with general- purpose diode function generators found on analog computers. Previous PHYSBE articles2,3,4 have discussed the amount of time it has taken to program with FORTRAN (on a CDC 3600 computer), MIMIC and DSL/90 (on an IBM 7090 computer), CSMP (on an IBM 1130 computer), and DARE I (on a PDP-9 computer). Very little attention has been focused on solution accuracy. Examination of published results for the all digital simulations of PHYSBE show serious accuracy problems. There appears to be a 5% to 10% difference in pressure and volume values that are published. There is good reason for the FORTRAN, MIMIC, and DSL/90 results to agree, but they do not. Also one would expect the CSMP and DARE I results to agree, but they do not. This ISL-8 simulation shows that good agreement was obtained with an IBM 1130 CSMP solution after an error was corrected in the CSMP program. Detailed accuracy comparisons with the other solu tions are complicated because not all of the same variables are reported in the same manner. There fore the documentation format suggested by McLeod2 is used here so that ISL-8 and CSMP solutions can be compared with future work with PHYSBE.