Computer Simulation

5AT Performance Predictions based on Computer Simulation

Having (manually) completing the full 17 sets of Fundamental Design Calculations, Dave Wardale undertook an independent verification of the 5AT’s performance using a computer-based steam loco performance simulation software package created by the late Professor Bill Hall

Bill Hall wrote two simulation programs (see downloads page):

  • Perform” – in which simplified valve events are assumed, and
  • Perwal” – a more sophisticated package in which actual valve events are predicted.

Dr David Pawson, a retired industrial chemist, undertook a substantial amount of work (in association with Bill Hall) in calibrating the software packages by comparing their outputs with records from the BR stationary testing plant at Rugby and the GWR testing facility at Swindon, and from documented reports on various locomotive road tests in the 1950s and early ’60s. Since Bill Hall’s death, David Pawson has taken on the role of advocate for his software and can justifiably claim to be its leading exponent.

In the Introduction to his final set of Fundamental Design Calculations, FDC 18 “Performance Predictions”, Dave Wardale describes both the overall methodology and the Perwal software package that he used to verify the 5AT’s performance, while noting his reservations about the applicability of the software for simulating SGS (Second Generation Steam) locomotives. He writes as follows:

  • The accurate prediction of (locomotive) performance requires an indicator diagram for each chosen condition to be estimated: this has been done in Figs. 1.3.1. and 1.3.1.F. for the estimated conditions obtaining at maximum drawbar power. The production of such indicator diagrams is laborious and not justified other than for particular conditions of interest, such as maximum power. Computer programs are however now available that predict the indicated power – speed – evaporation – cut-off relationships from relevant data about the locomotive. Two such programs, developed by the late Prof. W. Hall, are ‘Perform’ and ‘Perwal’ (see Software Downloads page). The former makes an approximation to the locomotive’s valve events, whereas the latter links ‘Perform’ to the program used for determining the valve motion given by Walschaerts valve gear, i.e. Walschaerts, Ref. [14] of FDC.5, and is thus more accurate as it is based on a precise determination of the locomotive’s valve events. For this reason the present calculations are made using the Perwal program.
  • As there is no transparency in the computer programs, their validity cannot be directly verified. They require the use of data, such as discharge coefficients, that has to be estimated, and however good the theory behind the programs, the results are clearly only as accurate as the estimation of this data. Good correlation between performance figures obtained from these programs and test results for B.R. locomotives is said to exist, depending on the values used for the various coefficients (i.e. the coefficients have been chosen such that there is agreement between calculation and test). This would point to an indirect verification of the programs’ validity when applied to First Generation Steam (FGS). However when using them to predict performance where there are no test results available for comparison, as in the present case, there is no way of confirming that any estimated input data is correct.
  • Although, with the reservation given above, the programs appear applicable to FGS, this does not necessarily guarantee applicability to SGS such as the 5AT, where engine design shows a number of advances over that of FGS designed to improve steam flow and reduce heat transfer, i.e. to make the power generation process approach closer to the isentropic ideal. Such advances are, however, at least partly accounted for in Perwal by the input data (in respect of such items as discharge coefficients, expansion and compression indices, valve motion and cylinder cover temperature). Steam flow and heat transfer processes in an engine are complex, and there must be doubt that any analytical approach is better than an approximation. However as the calculations concerned are not design calculations, which of necessity must be accurate, a degree of uncertainty can be accepted, i.e. the present work can be taken as giving at least a reasonable guide to the expected 5AT performance. It also follows that any error will tend towards under-estimating performance of SGS, i.e. the 5AT should perform as well as or better than Perwal predicts.

Wardale’s conclusions about the results of the simulation are as follows (with minor abbreviations):

“Good agreement is seen for all results except for cut-off and indicated thermal efficiency. The cut-off discrepancy is not too great and is of no consequence for the FDC’s: it is because of differences in the estimated indicator diagrams, Fig. 1.3.1.F. and that given by Perwal (which can be compared), the former being better in that it predicts a lower cut-off for the required mean effective pressure. The reason for the discrepancy in indicated thermal efficiency is discussed below. Agreement on the important parameter of cylinder indicated power is within 0,4%, and as the locomotive’s rolling resistance is the same for either method, a similar agreement applies to the drawbar power, i.e. Perwal confirms the cylinder and drawbar powers used as the basis for the FDC’s. However due to the different shape of the two curves of maximum continuous indicated power vs. speed, Perwal predicts a slightly higher maximum drawbar power than [1.3.F.(1)] at slightly lower speed.

The difference in maximum drawbar power predicted by the two methods is within the accuracy of the calculations. Although Perwal’s prediction of speed at maximum horsepower is is somewhat less than that estimated in FDC 1.3, the drawbar power vs. speed curve derived from Perwal is very flat in this vicinity, varying only from 1,920 kW up to 1,925kW then down to 1 900 kW over the range 90 to 110 km/h. Only a slight variation in the indicated power vs. speed curve would therefore be needed for the speed at maximum drawbar power to deviate significantly from Perwal’s prediction.

The probable reason for the discrepancy in indicated thermal efficiency is because the denominator in the expression (cylinder work ÷ heat in steam to cylinders) is different depending on whether ‘heat in steam to cylinders’ is taken as the total heat transferred to the tender water or the heat transferred in the boiler only (i.e. that heat which comes from the fuel), which is less than the total when exhaust steam feedwater heating is used. Taking the former as the basis, the 5AT indicated thermal efficiency corresponding to FDC 18 [127] would be ([1.3.F.(82)] ÷ ([1.3.F.(77)] – [1.3.F.(117)])) = 16.3%, which is only 2.5% higher than the Perwal figure of 15,9%, i.e. there is acceptable agreement if the same basis for efficiency is used.”

Wardale’s Fig 18.2 (referred to above) is copied below.  The methodology he used to generate the diagram is described on the “Using Perwal and Perform” page.

Note: Copies of the programs Perform and Perwal are available from the Software Downloads page of his website.