1. Introduction
In highly complex industries, such as modern shipbuilding, CAD (computer aided design) systems play a fundamental[ds_preview] role in achieving and maintaining a competitive advantage. Central design systems use an assortment of simulation software to assess design alternatives. Fach (2006) illustrates the variety of simulation software used in the modern ship design process. This includes advanced flow simulation software grouped together under the term CFD (computational fluid dynamics).
The procedures described in this paper are neither original nor are they limited to the selection of software packages for the shipbuilding industry. Many of the recommendations are adapted from previous consulting experience and from guidelines relating to software selection in the German automotive and mechanical industry, VDI (1990). However, the selected topics and examples focus on software used widely in the shipbuilding industry. The term CAD will be used in a wide sense, including the broad spectrum of all simulation software used in the design process.
2. In-house or outsource
A key decision, to be made early on in a project, is whether the engineering task should be performed in-house or outsourced: i.e. should the work be carried out by one’s own company (requiring the development or licensing of specialised software) or should the work be outsourced to specialists external to the organisation. If the engineering task is outsourced, the software selection decision is shifted to the consultant providing the service.
Since the focus of this paper lies on the software selection process, we will only briefly describe the reasoning behind outsourcing engineering tasks, in particular CFD simulations. See Bertram (1993) for a more detailed discussion; despite the progress in CFD systems, the fundamental economic considerations have not changed.
If CFD calculations are undertaken infrequently, there can be large savings to be made by outsourcing. This is due to the high fixed costs of hardware, software and especially training (which is often not considered at all). However, if several computations are performed annually, considerable economies of scale can be achieved: if ten computations are performed per year, the unit cost will be reduced by 80 % compared with the unit cost if only a single computation is performed in the same period. This makes CFD computations for occasional, or first-time, users far more expensive than buying the service from third parties who themselves can profit from these considerable economies of scale due to frequent usage.
For the rest of this paper we shall examine how to assess the cost of in-house analysis.
3. The software selection process
The software selection task can be greatly facilitated by breaking down the analysis into small, self-contained areas containing a limited number of staff / departments. In each area of concern, »preliminary analysis« and »assessment and decision« phases will consider organizational, technical, and economic aspects of software implementation. These will be treated in more detail later in this paper.
The two main phases of this process, which will be described in some length, are:
• Primary phase: Analysis
• Secondary phase: Assessment and Decision
3.1. Primary phase: analysis
The selection of a software system should be based on economic and strategic aspects. It is advisable to structure the decision process by the assessment of related aspects:
• External aspects:
Market situation, software product survey
• Internal aspects:
Business strategy, product spectrum
While we evaluate software largely in terms of features and associated costs, it is worth contemplating the psychological effects of introducing new software. The early involvement of the eventual end users of the software will help to increase its acceptance and avoid extended disputes (in the extreme »guerrilla warfare«) during the software implementation phase. A labour force trained in the use of advanced software may be entitled to higher wages and such an effect should not be overlooked in later cost-benefit analyses.
Effective and well-designed user interfaces generally facilitate acceptance and reduce transition costs in introducing new software. Bruce (2009) aptly describes the problem, albeit for a different application:
»A bigger obstacle has been the interface to the user. All too frequently, GUI’s (graphical user interfaces) reflect the needs of the software developer, rather than the needs of the user. They cause user confusion and distrust which hamper adoption. The art of GUI design is maturing to a point where genuinely friendly interfaces are available.«
As a general rule, modern commercial systems, with a large user base, have better designed and more user-friendly interfaces than academic systems. During software development, the user interface can take over 80 % of the effort to design and implement; academic codes, which are typically used by a small number of expert users, often lack user-friendly interfaces. This is because the focus of the software development is on the implementation of the engineering analysis and not on the user interface.
3.1.1. Defining »Where are we now?«
An analysis of one’s own company serves to prepare a profile for the desired CAD or simulation software. At this point a SWOT (Strengths, Weaknesses, Opportunities, and Threats) analysis might also prove to be enlightening. For this investigation, both internal and external factors should be considered:
• Competitive position of the business
– Current and anticipated future market share of (current and planned) products
– State of software usage by major competitors
– Customer pressure to introduce software systems
• Value delivery chain
– Identify processes that have great potential for increased productivity (drawing, analyses)
– Forward integration (interfacing/integration to structural design, piping, production)
• Work force profile
– Activity distribution of work force (difficult to assess as staff are likely to hide true figures)
– Staff qualification profile, training requirements, motivation, morale
• Product and service spectrum (current and planned)
– e.g. requirement for analysis of: propeller flow; seakeeping; aerodynamics; fire; etc.
– Standardisation of documents or services
• Current IT environment
3.1.2. Defining »Where do we want to go?«
The requirements for the CAD or simulation software system should be grouped in categories, using, for example the MoSCoW system:
M MUST have.
S SHOULD have if at all possible.
C COULD have if it does not affect anything else.
W WON’T have this time but WOULD like in the future.
Any system not fulfilling the »Must have« requirements can be automatically removed from the list; the remaining alternatives can be classified by how many of the other requirements they meet.
The requirements themselves can address different aspects:
• Functional requirements
– Ability to handle industry specific geometries (free form shapes for ships)
– Supported technical features such as analysis techniques
– Supported import / export data formats
– Potential for future extension (modularised software; not hardware limited; open system that allows custom integration of in-house or third-party software)
• System requirements:
– System architecture: If the system supports parallel processing, then powerful multi-node computers can be built at quite low cost and this can provide a very effective way of increasing the processing capabilities of the system (by adding more nodes, rather than replacing the hardware).
– Scalability
– Operating system (this is likely to depend on the system architecture)
– Additional software (such as pre- and post-processing tools).
• Quantitative requirements:
– These requirements concern the amount of data to be handled by the system: processing power, short-term storage, long-term storage and working memory should all be assessed.
– Likely requirement due to anticipated future expansion of the system should be assessed. This can be very challenging since it can be very hard to estimate how software and hardware will develop except in the quite short-term.
• Strategic requirements: CAD investments often bind a company to their selected product for more than one decade! Disregard of strategic aspects is one of the most frequent reasons for fatal decisions in IT investments. It is crucial to contemplate the following points:
– Development capacity of CAD vendor
– Economic situation (market position) of CAD vendor
– Number (present and development) of installations
– Training and after-sales service provided by the CAD vendor
• Ergonomic requirements
– Software (user interface, help system, training, manuals/documentation, user macros, response time, system stability)
– Hardware (screen, keyboard, desk, chair, lighting, etc.)
3.1.3. Market analysis
A broad review of the software market should be made. A first impression of the market: the available products and their relative capabilities may be gained by visits to other users, technical brochures and websites, exhibitions, conferences, and external consultants. Obvious starting points are COMPIT (www.compit.info) for marine CAD systems and Numerical towing Tank Symposium (NuTTS) for marine CFD applications.
A review of shipbuilding-specific software is relatively easy due to the fairly small number of available software systems, see for example: Couser (2007, 2006) and Bertram and Couser (2007). A list of available systems should be drawn up and their capabilities compared with the list of requirements. Any systems which fail to provide all of the »Must have« requirements can be eliminated at this early stage. For the remaining systems, the other requirements can, at this stage, be given a simple »yes« / »no« answer without trying to quantify the extent to which a requirement is fulfilled. This enables a coarse ranking of the alternative software systems to be made, highlighting which systems are viable and where more information is required.
3.1.4. Assessment of economic benefits
For a first estimate of the expected economic benefits due to the implementation of a new CAD system, a simple cost-benefit comparison may suffice. For this simple analysis, costs can be considered as global sums and the benefits considered just as reduced man hours. To be beneficial, the reduced man-hours must at least compensate for the additional costs. This conservative estimate considers only the direct costs and benefits due to the implementation of the CAD system, but neglects possible indirect benefits and savings due to reduced errors etc. For simulation software, like CFD, this simple approach does not work and the expected additional income through the ability to offer a wider scope of services must be also considered.
3.1.4.1. Economic benefits of CAD
VDI (1990) gives a comparison of CAD versus manual work. Today, »manual« should, in most cases, be replaced by »using the existing CAD system«. However, the generic approach is still applicable: for a fixed amount of annual drawings and constant personnel costs per hour, L, for CAD or manual procedures, a minimum factor for increasing productivity from a cost point of view, Cp,min, may be derived:
(1)
B denotes the system operating costs per hour with the index c indicating CAD work (with the new CAD system) and the index m indicting manual work (or work with the old CAD system). M is the additional fixed costs per hour for the new system. These fixed costs include: yearly licence fees (or depreciation of initial cost over the expected period of use or before the purchase of an upgrade is required); installation, training and maintenance costs; and interest on loans (or opportunity cost on bound capital).
During formal training, the costs for each user include: the appropriate fraction of the cost of the trainer; the operating costs of the CAD system used for training; opportunity costs since the trainees cannot work productively during the training period. Where users are self-taught, the costs arise due to low user productivity during the training period. Training time and costs are frequently underestimated. Amounts suggested by vendors and other users are generally too low. In many cases several man-months are needed for staff to become fully productive using the new software. The disparity in the training costs of different systems (due to, often considerable, differences in user-friendliness, technical support and quality of training) may significantly outweigh the differences in the licence fees alone. Cheap software frequently turns out to be expensive if all costs (including training and opportunity costs) are considered.
3.1.4.2. Economic benefits of CFD
The value of computer technologies can be classified according to time, quality and cost aspects. The main benefits of CFD in these respects are, according to Bertram (1993):
• Problems solved more quickly than when using conventional approaches, due to better (direct) insight into design aspects. CFD analyses of ship hulls and appendages before final model testing are now standard practice in ship design. CFD can also help with much faster trouble shooting in cases where problems are found. For example: hydrodynamically excited vibrations as described by Menzel et al. (2008).
• Improved designs through detailed analysis and / or formal optimisation, e.g. Hochkirch and Bertram (2009).
• The speed of CFD now allows its applications during preliminary design. The use of CFD early in the design phase reduces the potential risk associated with the development of new ships. This is especially important when exploring niche markets for unconventional ships where the design cannot be based on previous experience, e.g. Ziegler et al. (2006).
• CFD does not significantly reduce the cost of the actual design process, but it improves quality and helps with the early detection of design flaws and this can lead to significant cost savings. Within the first weeks of design, 40 % to 60 % of the total ship production cost is determined, Johnson (1990). The costs of design modifications increase by orders of magnitudes the further into the project they are made; ideally no fundamental modifications should be made after the conceptual design phase. Achieving this goal can be greatly facilitated by the use of CFD. If CFD is employed consistently to determine the final hull form at an earlier stage, numerous decisions that influence the production costs can be made earlier in the design process, thus reducing the risk of expensive design modifications being necessary later in the project. This is especially important in the context of modern workflow methodologies (e.g. concurrent engineering and lean production).
When assessing the economic benefits of CFD, merely considering cost aspects will therefore lead to incorrect strategic decisions. We can assess the benefits of CAD by using an analogy with another computer technology: CIM (Computer-Integrated Manufacturing), aptly put by Dietrich (1988):
»In other words, when taking CIM decisions, the question of ›What will we save in the short term?‹ is not the right one to ask. The real issue is rather ›How will our operational situation develop if we do not introduce CIM?‹.«
3.2. Secondary phase: assessment and decision
Areas suitable for the application of CAD (and / or CFD) may be selected considering:
• Departments / groups suited to CAD work; and
• Applications (products, processes).
An economic assessment should compare costs and benefits (benefits being savings and increased turnover). Naturally, costs are easier to quantify than benefits. However, it is necessary (and possible) to estimate most costs and savings to provide a sound basis for deciding on a CAD investment. Splitting the total costs into individual items and estimating these items separately reduces the risk of gross errors when estimating the total costs. Once costs and benefits have been (approximately) quantified, a traditional assessment of the economic aspects (profitability) based on a dynamic method (discounted cash flow) of the investment is trivial using a spreadsheet.
3.2.1. Costs Analysis
In introducing CAD systems, one should contemplate the following initial and running costs as detailed below.
Initial costs:
• Purchase of hardware and software. (Leased software contributes to running costs.)
• Physical site installation (partitioning, furniture, lighting, air-conditioning, etc.)
• Hardware installation: cables, networks
• Software selection process (staff costs, travelling costs)
• Implementation
• Initial training, including opportunity costs due to loss of productive time
• Inferior performance (until the expected productivity gain is realized)
• Installation and integration of the software system (customised set-up, macros, interfaces)
• Data input: creation of fundamental databases and libraries
Running costs:
• Consumables, energy
• Maintenance and repair and updates to hardware that may be necessary to run the updated software
• Software maintenance / updates
• Insurance
• Interest on bound capital
• Rental costs (rooms, data lines, hardware, software)
• Ongoing training, including opportunity costs due to lost productivity during training
• Staff cost (user and IT support staff)
• Data security and backup
Some of these costs may be negligible in which case they may then be omitted; this exhaustive list is intended to prevent any cost items being overlooked.
3.2.2. Benefits Analysis
Now the more difficult task of estimating the benefits must be undertaken. Information on benefits that might realistically be expected should not be based solely on the promises of software vendors, although these may sometimes supply valuable information based on their experience with other customers. The experience of other users within the same industry should be assessed and can be found in publications, personal interviews and demonstrations. Software vendors may be able to help with introductions to such users. Benchmark tests are useful to test individual systems under real conditions. These tests should be selected by users to exercise the software’s capabilities using representative examples of actual products and processes.
Users should quantify (and verify with the use of benchmark tests) the expected benefit that will be provided by the software based on the following criteria:
• Modelling (for CAD: solid models, surface representation, hidden line removal, multiple view techniques, maximum number of data structure elements, etc.; for CFD: grid techniques, grid generation, turbulence modelling, cavitation modelling, free-surface technique, etc.)
• Programming (capability and user-friend-liness of command language)
• Ergonomics (user-friendliness, GUI, help system)
• Additional applications for CAD systems (geometric and physical analyses, integrated NC (numerical control) module, finite element analysis (FEA) pre- and post-processing, etc.)
• Display options (export of graphics, virtual reality models, videos, etc.)
• Interfaces for data exchange (STEP, IGES, VDA-FS, etc.; compatibility with common CFD formats, e.g.: CCM+, Fluent, etc.)
While improvements in productivity due to implementation of a CAD system (in design, structural design, production scheduling, assembly, etc.) may be estimated quite reliably, improvements in quality and flexibility of workflow and capability are difficult to quantify. The following list gives some items to contemplate in this context:
• Reduced time and costs for provision of existing products / services
• Potential for increased turnover due to ability to provide new products / services (due to the extra capabilities of the software)
• Increased quality of products / services; reduced cost of quality control
• Improved product development (design)
• Greater level of detail available at concept design stage; e.g.: when bidding for tenders
There are various methods to quantify improvements in productivity for individual activities (e.g. drawing or standard CFD analyses) that allow estimation of a PIF (productivity increase factor). While the estimates for individual activities may have considerable error (or uncertainty), the average PIF, obtained as the sum of all estimates, is usually accurate enough for the estimation purposes required at this stage. The individual activities are weighted according to their frequency (time share) in the usual operation of the company.
Further points in assessing the effect of CAD on profitability are listed below:
1. CAD improves product quality for constant capacity.
2. CAD improves productivity, i.e. unless staffing levels are reduced, CAD results in extended capacity; if there is no corresponding increase in demand and staffing levels cannot or must not be reduced, CAD investments will not pay off!
3. Flexible work hours increase CAD profitability.
For a better understanding of the cost of CFD use, it is useful to take a closer look at the workflow pattern for CFD analyses:
4. Pre-processing (grid generation; specification of boundary and initial conditions; specification of control parameters for the computation)
5. Computation
6. Post-processing (results interpretation; graphical output; report preparation)
These individual steps sometimes have to be performed repeatedly, in an iterative cycle. Cost aspects will be discussed separately for each step:
8. Pre-processing:
Pre-processing requires staff that are familiar with specialist software for grid generation especially on vessel hulls. This requires at least a basic understanding of the subsequent CFD computations that will be performed. Grid generation is best performed on workstations or fast personal computers. Time and labour costs of this step are mainly determined by the user’s experience and the user-friendliness of the grid generation software. Pre-processing represents the major part (40 % to 80 %) of the duration of most CFD projects. Staff training and software (in-house development or licences) are the main fixed costs.
9. Computation:
The computation involves almost no man-hours except for computer supervision. State-of-the-art computations employ parallel computers. The fixed costs for these hardware architectures are relatively small: for frequent (professional) users, typically less than 5 % of total costs.
10. Post-processing:
Post-processing involves production and interpretation of numerical, graphical and video output from the CFD computations. The software required for the majority of post-processing tasks is relatively inexpensive and often comes as part of an integrated CFD package. High-quality, photo-realistic or interactive virtual-reality post-processing requires dedicated software and user training, but their use remains, so far, the exception in maritime applications. Post-processing also involves interpretation of results by experts and their documentation for clients. User-friendliness of post-processing software and the use of standard graphical output and report templates contribute to keeping time and cost requirements to a minimum.
4. The CFD Software Market from a Strategic Management Point of View
The CFD market is in a transitional phase from a young market to a mature market. The problems in the pioneering days of CFD software can be attributed to the special features of highly-competitive, immature markets as discussed, in a general way, by Porter (1980). These features can be summarised as follows:
• Customer confusion due to contradicting statements from different providers
• Product reliability and quality problems
• Hesitant customers waiting for the next (cheaper and better) release
• Lack of standards and compatibility
• High fixed costs and low turnover for the providers, leading to high investment costs and insufficient development power
Fortunately these problems have been largely overcome. A consolidation process has reduced the number of serious suppliers to a point where stable conditions, benefiting the whole industry, appear feasible.
However, there is always the opportunity for »new players« to arrive on the scene. This is particularly true of universities who can often, relatively cheaply, develop new products (software modules) that reflect the latest technologies and are often superior to existing software in terms of functionality.
However, if companies follow the guidelines suggested in this paper, especially if they consider strategic and economic aspects, they will, almost always, decide in favour of established companies and products. The high complexity of today’s software products and customer service requirements simply do not favour »cowboy« software development companies. We may thus see a trend similar to the aviation industry where, in the end, only a handful of companies will survive. In the long run, this should be rather beneficial for the industry as a whole: fewer suppliers mean more customers per supplier, i.e. more development power and the burden of development costs shared by more shoulders. It also means that agreements on standards and data transfer between different products are easier to achieve.
Recently, the open-source CFD package OpenFOAM, has attracted a lot of attention in the maritime CFD community, Schmode and Bertram (2009). In the hands of well-trained users, OpenFOAM yields good results for a variety of complex maritime flows (seakeeping with six degrees of freedom; cavitating flows around propellers; sloshing with breaking waves; etc.), see e.g. the 2009 Numerical Towing Tank Symposium, www.uni-due.de/imperia/md/content/ist/nutts_12_2009_cortona.pdf. Zero licensing fees, wide scope of applicability and access to source code for personal modification and research (that may lead to doctoral degrees) make OpenFOAM very attractive, particularly in the academic environment. Industry users, on the other hand, should take a more prosaic perspective: savings in license fees may be far less than the added costs of training. Experienced CFD users have reported (in personal communication) that it took them several months to two years to come to terms with OpenFOAM. Nevertheless, Germanischer Lloyd has decided to employ both commercial CFD software (CCM+ developed by CD-adapco) and OpenFOAM in parallel. The investment of considerable resources in a »free« software package was motivated by co-operative projects with several academic partners. However, it is fair to say that the bulk of professional consulting work continues to be based on the use of commercial software products.
5. Conclusions
A systematic approach to software selection is recommended. The outlined procedure is intended to prevent important items from being overlooked. Specifically labour cost associated with training and lack of productivity are frequently underestimated. Companies should select software based on how long training will take and how much man-time will be required for a trained user to complete a typical project. This in turn depends on user-friendliness, software user community networks, vendor support, etc. For key software, strategic considerations are important to ensure long-term support by vendors.
For professional CFD applications, labour costs are decisive. Therefore vendors should focus on faster (i.e. more automated) grid generation and post-processing.
Prof. Dr.-Ing. Volker Bertram
Germanischer Lloyd, Hamburg/Germany
volker.bertram@GL-group.com
Patrick Couser
Formation Design Systems
Bagnères de Bigorre/France
patc@formsys.com
Volker Bertram, Patrick Couser
