Late in June, the Australian Defence Ministry released a report on the status of the Royal Australian Navy's Collins-class submarines, the largest current Australian defense procurement project. It was damning.
Completed three years ago, HMAS Collins, the lead ship, cannot yet be considered operational, according to the report, and could be sent into combat only in a national emergency. The last of the six-submarine class, HMAS Sheean, is about 95% complete. The situation was so bad that the planned decommissioning of the last Australian Oberon-class submarine, HMAS Otama, planned for December 1998, had to be delayed; there is a serious possibility of a project to extend her life.
The new report is particularly embarrassing because leaks of earlier reports were vehemently denied, and even attributed to companies competing with the Australian Submarine Corporation for orders from some Far Eastern navies, such as Korea. Now it turns out that many of the rumors were, if anything, understated.
The Australian submarines were designed by Kockums of Sweden, which had previously built many somewhat smaller craft for the Royal Swedish Navy. They differ from Swedish submarines, however, in two crucial aspects: First, they are much larger and their engines have 50% more power (provided by 18 rather than 12 cylinders); second, they have entirely different sonars and combat direction systems.
Moreover, the two submarines have different operational profiles. Swedish submarines operate fairly close to home and reserve their maximum speeds for escape after an attack. Australian submarines, in contrast, are likely to operate at very long range (hence their large size, about 3,300 tons), and they may be required to transit for extended periods at high speed to their operating areas.
At the outset, the Royal Australian Navy must have thought that the program would be relatively risk-free. All elements were direct developments of existing successful systems. The hull and machinery combination came from Kockums, which had considerable experience. The French Scylla sonar suite was a developed version of the widely sold Eledone system. The Rockwell combat direction system was a developed version of the Mark I Singer-Librascope submarine fire-control system used on board Australian and Canadian Oberon-class submarines. In the navy's view, the real question was whether the new Australian Submarine Corporation could match Swedish welding standards in assembling the submarines. Indeed, the program was unusual in that major parts of each submarine were brought into the new yard from all over the country. As it turned out, this was no problem. The welding was excellent, and the modules fit together perfectly. All the problems were inside the submarines.
In the end, however, the larger size and new technology caused difficulties that must be addressed. The 50% upgrade to the diesels has resulted in problems; of such magnitude that ultimately the diesels may have to be replaced altogether—which may not be too serious a problem. In the Collins class, as in other modern diesel submarines, the engines are generators, connected only by wiring to the motors that actually drive the propellers (in older submarines, diesels were connected directly to the propeller shafts). The seals around the propeller shafts leaked badly, but that problem was soon solved; the Australians also found that their solution could be exported to other submarine-operating navies. There also were flow-noise problems, caused both by the shape of the hull and the material used for the propellers.
The main noise problems are cavitation and blade rate. Propellers cavitate at shallow depths (how shallow depends on the propeller and on the speed). Air bubbles form and then collapse, generating noise and, incidentally, eroding the propeller. In contrast, blade rate occurs at all depths. It is caused by irregularities in the flow of water around the hull and particularly around the fins at the stern of the submarine. Because the flow is not completely uniform, propeller blades vibrate when they hit irregularities. The vibration travels up the propeller shaft into the submarine, which often radiates sound from her bow. This effect is particularly marked in modern single-screw submarines. Cures include the specially shaped propellers whose automated manufacturing technique was compromised to the Soviets in the 1980s. Apparently, such propellers may have poor cavitation properties. That would be quite acceptable in a nuclear submarine designed to spend most of her time at depth, but it is clearly a problem for a snorkeling submarine. In the Australian case, the designers seem to have sought a compromise; propellers were shaped largely to limit cavitation. They were made of an exceptionally rigid material to minimize vibration caused by flow irregularities. As it turned out, rigidity meant brittleness, and the Australians have had propeller cracking problems. Such problems are presumably universal, and they do not really reflect on the quality of the submarine design.
The masts also have problems. The periscopes apparently vibrate badly when extended at depth, and the search periscopes suffer from a gray band that forms across their field of vision (the "double-dove" effect). Both search and attack periscopes lose focus when changing from one power to another.
The issue of the combat system, which apparently is poorly matched to the sonar, is far more serious. Almost from the beginning, accounts of the Collins-class combat system stressed the sheer quantity of data the sonar was likely to produce, perhaps more than an order of magnitude greater than in earlier types of submarines. Sonar ranges and bearings flow to the combat direction system, which is intended to construct a tactical picture on their basis. The tactical picture in turn is used to conduct attacks and other operations. In a modern submarine, most of the information flowing into the combat direction system is passive: sonar bearings, not ranges. The computer(s) in the system use the bearings to deduce the positions (range as well as bearing), courses, and speeds of underwater objects around the submarine. Ideally this target motion analysis (TMA) process is applied automatically to every object picked up and tracked by the ship's sonar systems, the accuracy of the fixes improving as the submarine maneuvers. In a sonar system with multiple arrays, it also is necessary to match up targets detected with one array with targets detected (at different frequencies) by others.
The Rockwell (now Boeing) system for the Collins class is credited, by its manufacturer, with being able to detect 1,000 objects automatically, initiate tracks on 200 of them (i.e., classify the others as of no interest), and automatically localize (conduct automatic TMA on) 25. This contrasts with the earlier Oberon system, which could track only 8 to 12 targets. The Collins-class sonar operates over a much wider frequency range, producing about ten times more data than that of an Oberon; the system uses Motorola 68000-series microprocessors, the same type as MacIntosh computers. It was designed in the late 1980s, and reportedly there were difficulties in writing sufficient amounts of code. Recent Australian press accounts suggest that the system has a poor man-machine interface, using large numbers of commands where a newer one might use touchscreens and windows. Possibly the problem is that some of the software has survived from predecessor systems; at first, the Royal Australian Navy may have considered it better to keep using the interfaces developed for the Oberons. As of 1999, the requirements levied on the combat system (target classification, track management/TMA, and weapons control) have been drastically reduced. Although all sonar data are being processed, only 25% of sonar analysis functions are being performed; some of the missing functions are considered essential for a submarine on operational patrol.
It is a distributed system, in which all sensors provide information to a local bus, which distributes it as needed to the consoles. As in other distributed systems, each console can take over the load on others should the others fail. To make that possible, there must be a considerable layer of software between the operations the console actually performs and the data bus. Several navies have discovered that this extra layer severely slows the operation of the system; also, bus capacity often is overestimated. In the Australian case, according to the report, information transfer has slowed to such a degree that the multipurpose consoles are inadequate and too cumbersome for the operators. Similarly, the automatic processors in the consoles are proving so inefficient and that they actually increase the operator load.
This is hardly a new problem. At one time, personal computers ran only one program at a time and even slow central processors could make very quick calculations. Then systems like Windows appeared. They interposed thick layers of software between the data and the applications. Often the layer was very useful. For example, it made for relatively easy systematic transfer of data between programs running more or less simultaneously. Users found that, to run their applications at much the same speed, they needed much faster processors. That may be true of the Collins class. In their case, the situation may be aggravated by the fact that the 68000-series processors no longer are being improved. The lesson may simply be that system developers should take more account of the need to provide an upgrade path. The end of 68000-series development may merely be Boeing's (and Australia's) very bad luck, since software written for such devices will not run on the alternative 80xxx-Pentium series. The Australians were aware of the problem, and they demanded that all software be written in Ada, a portable language. That may merely have compounded the problem, since Ada itself reportedly is extremely inefficient.
For the time being, surplus U.S. Navy sets of equipment are being installed in the first two ships. They are auxiliary sonar processors, interposed between the sonar and the combat direction system. From the official description, they are probably part of the BSY-1 system, which had problems handling sonar data. Delivery is scheduled for August 1999. The key BSY-1 problem, which apparently affects the Australian system also, was that raw sonar data was distributed to the combat direction system consoles to enable operators to match data from different sensors and make better assessments. But raw data mean a flood of data—which the busses sometimes cannot accommodate. The solution is to process the raw data into much simpler form closer to the sonar arrays, so that the outputs sent to the consoles are processed data (ranges and bearings) used directly to form a tactical picture. The Australian report describes this as a less-integrated approach.
The official Australian assessment, however, is that the Boeing system in place cannot be salvaged—it must simply be replaced altogether. This drastic solution is not unique. The British designed their Type 23 frigates to use a system called CACS, which was related to earlier British combat direction systems. By the time the ships were being built the official view was that CACS could not possibly handle their requirements. The system was canceled, and eight ships went to sea with no computer combat systems at all; they were useless as warships, even worse than the Australians say the Collins class is today. Meanwhile, the British held a competition to develop a wholly new system. It was developed quite rapidly, and it was quickly installed using the data buses already in place. Reportedly it was too slow at first (as in the Collins class), but that apparently was remedied when Pentium II processors replaced the original Pentiums, offering much higher speed but running exactly the same software. This replacement is the upgrade path mentioned above.
The damning comment in the report is that the combat direction problem was first noticed as early as 1993, but there was no attempt to change either the basic configuration of the system or replace the contractor, Rockwell. Instead, both the Australian Submarine Corporation and Rockwell, its combat system contractor, were released from key responsibilities; in the view of the report that reinforced a widespread view that the original submarine contract could not be modified. These problems probably were the result of a decision to buy all six submarines at a fixed cost, even though many parts (such as the combat system) did not exist at the time of contracting.
Both the United States and the United Kingdom have experienced considerable problems with submarine combat systems, particularly because commercial technology has accelerated far beyond what can be achieved with equipment built to military specifications. The British found themselves canceling several systems altogether in favor of their current commercial based distributed system. The U.S. BSY-1 system, in turn, had to be modified drastically, because what had begun as an integrated combination of sonar and combat direction had to be abandoned.
For the present, the hope is that the Boeing system can be modified by replacing its software and some of its hardware. Bids for new systems probably will be requested; Boeing may well offer a system of its own. Alternative suppliers include Thomson (French) and Sonartech (allied to Krupp Atlas, the German sonar producer). Both have strong contacts with the Royal Australian Navy; Thomson supplied the Collins-class sonar (which has apparently been quite successful), while Sonartech supplied the current mine-hunting sonar.
In the end, how bad is the situation? Ironically, in a modern warship much of the money goes to the combat system, but that is the easiest element of the ship to rip out and replace (the system bus may be a different proposition). Despite all of its criticisms, the report says quite clearly that the hull of the Collins class is quite sound. The problems can all be fixed, given time and some investment. The diesel engines can be fixed or replaced. The combat direction system can be ripped out and its consoles and software replaced altogether, probably without opening the hulls at all. It may even be that inserting the U.S. equipment and modifying the software will solve most of the problem. The comforting point here is that software often dominates the way a combat system works, and that the software is easiest to replace, from a physical point of view (writing software is a very different proposition). Perhaps the gamble the Australians took, in undertaking to build a new class of submarines despite limited experience, will turn out all right in the end. Perhaps, too, they will take the opportunity to seek some commonality in software among their submarines, their surface fleet, and other defense systems.
An alternative lesson may be that the Australians' error was to try to integrate two major electronic systems, the sonar and the combat system, which were not entirely compatible. The Collins-class project probably was the most complex the Australians had yet attempted, although it may be rivaled by some of the proposed upgrades of Oliver Hazard Perry (FFG-7)-class frigates. It may have been disastrous mainly because the Australians never realized the extent to which they were not buying anything quite off the shelf. The French normally sell their sonars in combination with French submarine combat systems. Presumably the French systems did not meet Australian requirements. It is probably also true, however, that the way the French divide up functions between sonar and combat direction differed from Rockwell's practice. It is quite possible that the difference was overlooked in hopes that the combination could indeed be assembled at the desired fixed price. That sort of disaster has happened before, and surely it will happen again.
In the United States, it is accepted that software may well be incompatible, and that the only cure is to run it in a shore test facility. The Australian report mentions a recently developed "virtual submarine," which may be used to test possible alternative replacement combat direction systems. It appears to be a land-based test site; that it has only recently been developed suggests that the absence of such a site was a major flaw in the combat system development process.