|
|
|
ERRORS AND MISTAKES
Last updated 9 August 2002
CONTENTS:
INTRODUCTION |
|
|
|
|
|
|
ERRORS |
||
|
|
|
INTRODUCTION
It can't be avoided, the Gunners have sometimes been called 'drop shorts' because the shells didn't land where they were expected. Sometimes, despite all the checks, this really is the result of a mistake. More usually it’s the result of natural variability in the system - errors. These are a natural consequence of the variability of outcomes for repeated 'identical' events. It's a branch of statistics called probability theory, although this may not give much comfort to anyone unfortunate enough to be in the way.
It appears that the British (in collaboration with the Canadians) were the only army to undertake detailed studies into errors in the 'wash-up' after WW2. The task was undertaken by the Field Artillery Working Group of the Standing Committee on the Accuracy of Artillery Fire (FAWG of SCARF). The reason for this work, undertaken by several sub-committees, was concern that predicted fire had been inaccurate, perhaps as few as 7% of predicted engagements being accurate. The committee used many sources including reports by 1st (Italy) and 2nd (21st Army Group) Operational Research Sections, data collected in various trials and tests in UK and Canada, including specially commissioned ones, those undertaken by field army formations notably 1st Canadian Army, and studies and trials undertaken by the Ordnance Board and published in their Proceedings.
It's highly unlikely that all the various errors and mistakes were a uniquely British phenomena, and many applied to the artillery of all nations, although mistakes are likely to increase in a poorly developed artillery system.
Key definitions are:
Errors are in one of two categories - accuracy or consistency. Accuracy means the closeness of the mean outcome to where it should be. For example the difference between the mean point of impact (MPI) of fall of shot and their aim point. Consistency means the size of the spread of outcomes (eg shells) around their MPI. In artillery this variability is called 'consistency' or 'dispersion' (the optimistic or pessimistic views of the world - do you call a glass half-full or half-empty?)
Figure 1 illustrates the concept in terms of the fall of shot around an aimpoint. Of course the apparent irregularity really reflects the small size of the sample.
Figure 1 - Accuracy & Consistency
Both accuracy and consistency are measured in terms of 'Probable Errors' (PE) and exist throughout the artillery system. For example the MPI of the fall of shot from a single gun may or may not be at its aimpoint. The distance MPI to aimpoint is accuracy, the size of the fall of shot's spread around its MPI is its consistency. Similarly the MPI of the MPIs of each gun in a battery will be the battery MPI, and the MPI of the MPIs of the batteries in a regiment will give a regimental MPI (assuming they were all aimed at the same point). The measure of how far these MPIs are from their common aimpoint is their accuracy and defined in terms of PEs.
Conventional statistics defines variability in terms of 'standard deviations' from the mean with a 'normal distribution' (a 'bell curve'), such that all outcomes for a particular type of event will fall within about three standard deviations from the mean of them. Artillery does not use standard deviations, but 'probable errors' (PEs). In essence its just a matter of slicing up the spread of outcomes in a different way. Instead of them all falling with 6 standard deviations of the mean (3 either side) they fall within 8 PEs, 4 either side of the mean, the total distance is the same. Figure 2 shows the normal proportional distribution of outcomes for an event. The 2 PEs (one either side of the MPI) are called the 50% zone, because 50% of the shells fall there, the central 4 PEs give the 82% zone, and so on. The size of a PE will vary but their distribution does not, Figure 2 shows the distribution of PEs as a histogram, the horizontal axis represents the size of a PE.
Figure 2 - Standard Distribution of Probable Errors
PEs occur throughout the artillery system. For example the spread of the fall of shot from an individual gun firing repeated shells at the same data, the size of theses PEs are documented in Range Tables. For a 25-pr firing charge 3 at 8000 yds range the PEr was 30 yds (PE range - along the line of fire) and PEl was 4 yds (PE line - across the line of fire). Figure 3 illustrates the distribution of 100 shells when combining their range and line PE, (rounded numbers). However, most properly conducted actions in the artillery system have a normal distribution of errors.
Figure 3 - Range and Line PEs Combined
It must, of course be remembered that the PEl is much smaller than the PEr, this means that in distance terms 50% of the rounds will fall within one PEr of the target.
For some variables the PE could be circular around the mean (PEc). PE for different things can be added together using the root mean square rule (square the PEs, add them together and find the square root).
MISTAKES AND ERRORS IN WORLD WAR 2
In September 1945 FAWG of SCARF produced the 'Final Report of the Committee Set Up to Investigate the Accuracy of Predicted Fire'. The committee was established by the joint direction of the Director Royal Artillery and the MGRA Training. It's noteworthy that the DRA had also been the originator of Operational Research in the British Army. The work was undertaken by several sub-committees with representatives from a wide range of military organisations.
The main sources of mistakes were identified as:
People are prone to make mistakes, particularly when under stress or severe environmental conditions. The system of drills and double checks was designed to eliminate mistakes. For example the data was set on the sights by the layer and the charge was prepared by the loader, with both being checked by the Number 1. However, while the official reduced detachment for a 25-pr was 4 men, in the field 3 men were often used because this gave 'half on, half off' with the 6 man detachment. The result was that the Number 1 acted as loader while doing his other Number 1's duties, but there was nobody to double check his preparation of the charge, and he may have been a bit coy about reporting any 'bad ram'.
Mistakes could also lead to errors. If shells were loaded with dirt on their driving bands then after a few hundred such rounds there was measurable ovality in the barrel, which caused increased error.
Of course mistakes also happened outside artillery control. Most notably infantry were in the wrong place at the wrong time. For example they failed to stop at a phase line denoting the safe distance to the opening rounds of the next phase of a fireplan.
ERRORS
Table 1 shows the size of the PEs that affect accuracy in range for predicted fire, line errors, Table 2, were generally small. The FAWG of SCARF report goes into considerable detail to explain all the factors, including when they could be larger or smaller. It may be noted that Range Tables give PEs of 35 yds and 55 yds for 25-pdr and 5.5-in respectively at the ranges used. The selected ranges and charges have an elevation of about 25 degrees.
TABLE 1
Errors which affect the range to the mpi in predicted fire
TABLE 2
Errors which affect accuracy in line
Note that as a horizontal angle 1 min subtends about 1 ft at 1000 yds, therefore at 9000 yds range 2 min = 6 yds, and that in these tables 'v' signifies square root.
The size of the errors in serials 8 to 22 in range (Table 1) and 6 to 11 in line (Table 2) increase with the range between gun and target, in some cases it is a function of range in others it is more closely related to time of flight. In Table 1 serials 3 and 4 originate outside the gun position and are sometimes called the 'target location error'. Putting the range and line errors together helps identify the main problem areas.
The most significant were errors in correction of the moment, the main source was that the meteorological data could be used several hours after being measured. Wind is the main variable because this can change quickly and significantly. Another problem was the method of applying large corrections. Varying propellant temperature from various ways of storing ammunition on gun positions was also significant (TV news films show that this remains a problem in many armies!). Finally the interpretation of air burst ranging results was also a source of error.
Next were those associated with target location, interesting because they show height as a bigger problem than location. The reason for this was that predicted targets tended to be acquired from locating systems such as sound ranging and flash spotting or gridded air photos that were reasonable accurate. However, the locations then had to be plotted on a map to find the target height. The problem then was the inaccuracy of the maps, particularly in France where a well marked feature such as a road junction could be mis-positioned by up to 100 metres. This meant that height was often wrong because it had been taken from the wrong place on the ground. Obviously these problems with maps meant that predicted targets selected by ground observers were also inaccurate, exacerbated by the much greater PEs for target locations by FOOs (some trials revealed a PE of over 100 yds by FOOs, even with good maps).
Plotting was based on the use of 1:25,000 scale plots used on artillery boards in command posts, however, long range guns were plotted at 1:50,000 and this doubled the PE. The underlying causes were the inability to plot to greater than 10 yd precision on a 1:25,000 grid and bowing in the steel range arms from rough handling.
Variation in propellent from type to type and lot to lot. Different propellant types were available for most types of gun, eg FNH, NQ, WM, etc, these types were matched in cartridges for new gun MVs. However, worn guns behaved differently in terms of MV equivalence. Lots were manufacturing batches, with manufacture in factories around the world. Just as light bulbs have different length lives due to manufacturing variability so is there variability in the manufacture of propellant lots, and British lots were relatively small so there were many different ones in the supply chain and reaching gun positions.
Calibration results, particularly applying calibration corrections for one charge to others, this saved a lot of time in calibration firing, but at the price of MV accuracy. Furthermore the comparative calibration method was less accurate than full calibration of individual guns, a more time consuming business. This included the updates to MV calculated from wear measurements, slow wearing guns such as 25-pr weren't too much of a problem but fast wearing ones such as 5.5-in, 4.5-in and 155-mm M1 were. Research in 2000 using new precision instruments (micron measuring) revealed that it is wear to the grooves not the lands that causes loss of MV. Worn guns had a knock on effect that was picked up in full calibration but not otherwise. Shells fired from worn guns (particularly when there is ovality towards the muzzle) are less stable in flight and this affects their range, and exacerbates the effects of reduced MV. This was revealed by the divergence of MVs measured by calibration firing and by camera techniques as guns wore. There seems little doubt that worn barrels were not being replaced early enough.
The last significant problem was day to day variation in gun performance, and again 155-mm M1 was particularly notable in this respect. The causes of this variation are not well understood to this day.
From the above tables for predicted fire some key points emerge, assuming no mistakes. For 25-pr at a range of 9,500 yds:
Of course there will be a tendency for the MPIs to be skewed away from and not 'around' the target because some errors, such as target location and meteorological data, affect all fire units in much the same way.
Air-burst shells, with clockwork or powder burning fuzes were a further problem. It was virtually impossible to get the optimal height of burst (HOB) for HE (about 10 yds) by prediction, although a 'good enough' result for higher bursting shells such as smoke was possible. Even with ranging the HOB for air-burst fuzes was normally distributed so that from a battery there would be some very low 'daisy cutters' or ground-burst and some at 20 yds or more, and ranging HOB for regimental or higher targets usually meant ranging each battery's HOB in turn. The great advantage of VT fuzes when they were released for field use was their accurate and consistent HOB.
The next table gives data for errors that change the consistency of the fall of shot from a single gun. It applies to both predicted and observed fire. The values are for a multiplier to the Range Table 50% zone (2 × PE), 70 yds for 25-pr.
TABLE 3
Errors which affect dispersion in predicted and observed fire
This table clearly shows that the main cause of increased dispersion was worn guns, and that some data was unavailable (serials 5, 9 and 10). Serial 6 relates to an engagement lasting about 1 hr and the met conditions changing during that period. Data for dispersion in line shows this to be small, totaling about 3 mins. An underlying cause of range tables giving smaller PEs than found in the field is that the data for range tables was collected during range and accuracy firings. These were undertaken with all possible measures to eliminate variability between rounds.
Applying these multipliers to a 25-pr 50% range zone of 70 yds, gives 84 yds for new guns and single lots to 126 yds for old guns and mixed lots.
DISPERSION COMPENSATING FOR INACCURACY
Dispersion can compensate for inaccuracy. Figure 4 shows a illustrative relationship between an actual target on the ground, its location as ordered by the originator (the aimpoint, offset from the target by the target location error) and the fall of shot from a regiment of three batteries each with their own MPI, only one shell from each gun is represented. There is some effect on the target, and there would be more when the dispersion for the fall of shot of each gun is added (see Figure 3), furthermore the spread of fragments from each bursting shell is laterally from the line of fire and forward (see 'Effects and Weight of Fire'). Of course each troop has a different line of fire, how different depends on how the regiment is spread out.
Figure 4 - Illustrative relationship between target, aimpoint and fire unit MPIs
Note that the guns' individual aimpoints are not necessarily the battery's. A British battery was divided into two troops, usually a few hundred yards apart. Each aimed their pivot gun at the battery's aimpoint with the other guns parallel to them and firing at the same range. This gave a natural spread to the battery's fall of shot relative to the battery's aimpoint.
The FAWG of SCARF report investigated the issues of accuracy and dispersion for different numbers of fire units. In essence the question was "will employing 18 troops instead of one give 18 times the density of fire over the same area, or the same density over 18 times the area, or something in between". The following Table 4 refers back to Table 1 above and refers to 5.5-in guns firing 100-lb shells with charge 4 at 13,000 yds. Of course there was almost certainly never an AGRA of 4 regiments all with 5.5-in!
Table 4 - Errors which affect accuracy and dispersion in range in a concentration
The data assumes an AGRA was deployed on a 6,000 yd frontage, so for 9,000 yds range the angles between the extreme lines of fire were 9 degrees for a regiment and 26 degrees for the AGRA. This affects the error in wind as well as the overall pattern of the fall of shot.
Table 5 - Errors which affect accuracy and dispersion in line in a concentration
The results from Tables 4 and 5 are combined in Table 6. Table 6 shows that as the number of guns increase the accuracy of the concentration improves and the dispersion of MPIs from individual guns and their fall of shot around the overall MPI increases.
Table 6 - Summary of PEs for range and line in concentration of fire
Table 7 provides multipliers that can be used with Range Table data to give a reasonable approximation of the area covered by different numbers of guns and the density of fire. The area covered by fire is approximately proportional to the square root of the number of guns firing. The density of fire is proportional to the ratio of the square root of the number of gun and the area covered by fire.
Table 7 - Estimate of increase in area covered and density of fire arising from increase in number of guns
|
|
|
Copyright © 2001, 2002 Nigel F Evans. All Rights Reserved.