In an aeronautical sector company, where engineering design process steps and activities are developed, a decision aiding methodology was required to support the processes. We have proposed the integrated use of linear programming and multiple criteria methods, which can be used to orient the conceptual design of alternative functional and physical solutions and to cope with complex design problems. Linear programming has been used in relation to some case studies, to generate design alternatives that satisfy the set of the initial requirements, while multiple criteria methods have been proposed to interact transparently with the client, in relation to some aspects that a linear programming model cannot include, to evaluate, compare and select alternatives in order to identify and formalize new expectations that the first solutions have not be able to resolve. The iterative use of the two approaches, in a cyclic procedure of mutual learning, allows the requirements to be defined more clearly and a final satisfying solution to be reached.
Engineering design is an iterative decision-making
process which is developed to devise a component,
product, process or system that meets the customer’s
needs
An aircraft is a complex system, but it can also be seen as a component of an even more complex structure, a System of Systems (SoS), in which different systems communicate and work together to achieve specific targets. In an SoS, integration and synergic work may vary from a simple collaboration, in which the single components work alone, to a situation in which the single components are not able to work in an autonomous way, when extrapolated by the SoS.
The MBDD supports product development by managing the relationship with the client in the initial engineering design process phase. The client’s needs have to be identified, in order to decide whether and how a specific legacy aircraft has to be updated in relation to these needs, or to understand what kind of aircraft has to be designed or (at least partially) re-designed, in order to guarantee its integration in the new SoS that the client perceives as possible or essential for the future.
The client's involvement in the initial phase of the
design process is analyzed in the literature in "front end"
models of the product development process
In aeronautics, a partial and apparently limited redesign requires years of work (five years on average) and therefore uncertainty concerning the evolving nature of the client’s requirements is normally present, with an evident impact on the engineering design process. The MBDD asked our research group for suggestions and methods, in order to improve the interaction with the client (who has to understand every step of the process and freely propose his point of view), to reduce time and guarantee quality of the results, which can be solutions and/or a better definition of the needs, objectives, priorities and future scenarios of aircraft use.
We proposed the integrated use of two kinds of
Operation Research methods, in relation to some case
studies, and the cyclic use of the method application
outputs, as new inputs for the other method application
The first section of the paper focuses on the iterative nature of the engineering design process and offers a synthetic overview of the methods, theories and tools that are used by designers.
In the second section, the problem, as perceived by the MBDD, is presented and, in the third section, a set covering model is proposed for the generation of design alternatives.
In the fourth section, some multiple criteria approaches are described, in relation to the evaluation of design alternatives, and the integrated use of two methods is proposed to support communication with the clients, in order to better define their needs and expectations. The possible development of the procedure, in relation to more complex projects and decision contexts, is analysed in the conclusions.
Several theories and various tools are proposed in engineering design to aid designers in different ways: to understand stakeholders’ needs, improve quality, address variability and uncertainty in the design process or generate alternatives for designers.
The engineering design process, as described by
Once at least the structural components of the design have been identified, above all with inputs from testing, manufacturing and marketing teams, the design team generates alternative conceptual solutions that are oriented in different ways to achieve predefined goals (i.e. requirements that have to be satisfied).
Considering costs, quality and risk, as the main
selection criteria, the most promising alternatives are
selected for a further analysis
In the conceptual phase of the design process, it may be necessary to go back to a previous step at any point in the process. The chosen solution may prove to be unworkable for different reasons and may require specification redefinition, new solution generation, the collection of more information or, in the worst situation, redefining the problem. This is a continuous and iterative process.
Several tools are commonly used to aid designers.
Methodologies and theories that have been proposed in
the literature, usually offer a more analytically rigorous
support for engineering designers. Concurrent
engineering may be the most practical methodology to
improve the design process. The approaches that are
most frequently suggested to obtain input from
stakeholders in the design process are the Pugh Method
The problem definition step, in the engineering design process, is critical when the client has to face an evolving situation and cannot clearly communicate needs that are no well defined. This criticality is often present in aeronautics, where many years are required to create a new aircraft, but also to innovate some elements of a legacy system.
The problem definition step is developed in the MBDD
by anticipating some activities that pertain to the
successful steps of the process
Clearly understanding the point of view of the client, at a functional level but also in organization terms, is essential to identify and structure the requirements that orient the design. The MBDD arrives at a complete problem definition through a procedure that involves the organization of a client in a comparative analysis of some promising draft solutions. These solutions are elaborated in the MBDD, in relation to general technical requirements, and then the strength and weakness elements of the solutions are discussed with the client or, more precisely, with some organization-client key actors (for example, a pilot or whoever is in charge of maintenance).
Even if the innovation is related to a single aircraft component, the future use of the aircraft in an integrated SoS has to be analyzed. Various types of aircraft, but also satellites and maritime or ground systems can be involved, in order to achieve an assigned target in missions of various kinds (i.e. military, civil or a combination of the two situations). Innovation is often required in order to specifically facilitate coordinated work and communication in the SoS.
The MBDD procedure includes two subsequent stages:
in the first one, some “functionally acceptable” solutions
are identified or elaborated, in relation to the
functionalities that are required. In the second stage, the
client’s attention is focused on these solutions, in order to
evaluate the associated costs (which are not only
monetary), their economic sustainability and specific
benefits and risks, as proposed in
In the last few years, some clients have required the use of an Operations Research tool, in order to facilitate comparisons of the solutions in a multiple criteria analysis. Having found the tool very interesting, the MBDD asked to our research group for a method to help the generation of “interesting” and acceptable solutions, in order to reduce time and guarantee the completeness of the acceptable solution set. We analyzed their use of the tool and the weak and strong points of their applications. We then proposed the integrated use of LP and MCDA models in a procedure that fits the MBDD approach to the problem, but also improves the interaction with the client, who can propose his point of view (in terms of limits of the solutions and opportunities that have to be stressed), in a simple but formal language, and who can almost immediately analyze all the new solutions that are consistent with the new vision.
The request of a client can be very generic and the real needs are not always easy to understand. In order to reduce uncertainty, a request can be expressed in terms of mission types that the new system (or asset) has to face, as a component of an SoS.
From a technical point of view, an asset is a system that guarantees specific functionalities. The assets may be component parts of a single aircraft that have to be integrated to complete a mission, or to be integrated with other assets in other aircraft or in systems that operate on the ground. The assets may also be specific kinds of aircraft (or other resources) that have to be activated together in a specific mission. In all these situations, the integrated assets can be seen as an SoS and both the performances of the assets and the relationships between them allow the missions to be performed.
The purpose of the analysis is to define a mathematical model in which the variables are the different assets that can be activated to accomplish a mission. The functionalities that have to be guaranteed (or guaranteed at a required level), in relation to the “nature” of the mission, can become the constraints of the model.
The different objectives, in relation to a specific decisional problem, can be: minimize the costs, maximize the effectiveness, minimize the risks of a mission and so on. A combination of assets that is acceptable because it guarantees the Required Level of Performance (RLP) for each functionality (i.e. for each constraint of the model) becomes an admissible solution, which is called architecture (of the SoS). The optimal solution is an admissible solution that minimizes (or maximizes) the objective. If a single admissible solution does not exist, the need for a technological innovation (i.e. at least a partially new asset) is underlined. A new product, or an improvement in a legacy system, satisfies the client’s needs if all the missions that the client had proposed to describe his needs can be faced with a minimum cost.
The problem can be represented by a linear
programming model, if all the constraints and the
objectives are linear functions. If there is only one
objective, the oldest and most famous method of
Operations Research, the Simplex method
At the start of the model setting, the assets that have to be included in the model and a list of functionalities, i.e. the constraints of the model, are defined in relation to the (generic or specific) request of the client and above all using the Universal Join Task List (UJTL) Report2.
A complete list of about 720 functionalities, in terms of ability to perform a task, is proposed in the UJTL Report, in relation to the strategic, operational and tactical level of mission in a military context. The UJTL was developed for the U.S. Armed Forces, but it has been used by several other countries and international military organizations, such as NATO. The MBDD has structured and adapted the Report to facilitate its use with the clients. The MBDD synthesizes all the coordination, monitoring and controlling functionalities for military missions in the Mission Management macro functionality and Find-Fix-Track is the code that is used to indicate the set of functionalities which, at different levels of detail, allow the area of interest to be patrolled, in order to indentify and trace the target. Using this framework, xx main functionalities, that have to be guaranteed in a military mission, are always present as model constraints. When the mission requires a specific and not usual functionality or for non military missions, the UJTL Report is used directly as a check list.
If the adopted objective is to minimize the number of assets that have to be involved in the proposed missions, the mathematical problem can be re-formulated in terms of a set covering problem, which consists in finding the
minimum number of service centers (in our model, the
assets) so that the request for each service (the guarantee
of a required level of a specific functionality) is covered
In this mathematical model, the performance pij of the jasset for the i-functionality is compared with Si, the RLP that has to be guaranteed for the i-functionality, in order to define the covering matrix [tij], in which the elements tij are equal to 1, if pij≥ Si, or equal to 0 otherwise.
The set covering problem can be formulated in the following way:
Min ∑ xj ∑tijxj ≥ qi j = 1,…, m i = 1,..., n xj={0,1} where xj has a value of 1 when the asset is included in the solution (which in this case, is an SoS architecture), otherwise it is equal to 0.
The value of the redundancy, for each functionality with redundancy (i.e. a critical functionality that requires more than one asset that is able to satisfy this task, in an SoS architecture), is equal to qi. For the others, qi is equal to 1.
We used Xpress-MP, version 2007 (Mosel 2.0.0, IVE 1.18.01, Optimizer 18.00.01), produced by DASH Optimization, to treat models with a single mission or multi scheduled missions that are included in the model. For a multi missions model with 18 variables and 210 constraints, the application has provided six admissible and three optimal solutions in 0.15 seconds.
The model structure and the linear programming application to the problem were tested in relation to some previous military cases, where the solutions and their characteristics were well known for the MBDD. We spent a great deal of time defining and modifying the constraints, in order to have a better fit of some specific requirements, but the immediate calculation of the solutions facilitated convergence towards a good model. The same procedure was then applied to a new case, in relation to the surveillance of a critical sea canal. The model development and PL application were accepted by the MBDD as effective steps of a procedure that can support communication with the client.
At this point attention was focused on the tool that should be used to understand why a solution is not adequate enough.
The U.S. Air Force Center of Expertise for Analyses of
Alternatives
The MBDD adopted a different approach when a client suggested the use of a multiple criteria method for the comparison of the alternatives. The first application was not totally satisfactory. Some other experiments allowed the MBDD to realize that the correct approach has to involve structuring the evaluation model in macro aspects that can be analysed separately by the organisation actors in charge of each specific aspect. At the same time, the MBDD elaborated a way of translating each personal judgment into an analytical function.
When we analysed the procedure they were using, we noticed that the results were very interesting, in relation to the first aim (improve communication with the client in order to understand his point of view and adequately model his requirements), but very poor as far as the second (transparently arrive at the decision and analytically document the decision motivations) was concerned. In fact, their analytical functions did not result to be consistent with either the original judgements or with the logical structure of the method. At the same time, their need to artificially construct evaluation functions, after the application of the method and in order to explicitly document the process, was analysed together with them, to show them that the wrong method had been adopted.
There are many multiple criteria methods that can be
used to aid decision making
We proposed the analysis of two methods, Analytic
Hierarchy Process (AHP)
ELECTRE III starts by comparing each solution with each of the other solutions. A fuzzy outranking relation, based on the two principles of concordance and discordance, is modelled in phase I of the method through the computation of a concordance index, a discordance index and an outranking degree. The method uses the latter result in the second fuzzy relation exploitation phase, in order to construct two complete pre-orders through a descending and an ascending distillation procedure. Outranking relation modelling offers some interesting advantages, in comparison to other multiple criteria methods: each criterion can use a different ordinal or cardinal scale, since a unique specific scale (such as the cost-benefit analysis monetary scale or the 0-1 utility scale of the multi attribute utility theory) is not necessary and the outranking relation is not compensatory (or partially not compensatory).
A weak point of ELECTRE III is its software package, which does not pay any attention to dialogue with the decision maker, which is essential in model structuring and parameter definition and when the results require a collective analysis. A new product, which is more suitable and includes several multiple criteria methods, is currently being developed in the Decision Deck project3. This weak point is related to the original nature of the method which was invented to be used when a problem was well structured, i.e. when:
· a set of solutions is identified, or elaborated, and tested in terms of completeness, admissibility and comparability, and
· a family of evaluation functions (i.e. criteria)
which has been created to represent all the different
aspects of the problem at hand contains a sufficiently
small number of criteria to be a basis for discussion
(legibility condition) and to be considered by all the
actors as a sound basis for the continuation of the
decision aid study; its coherence (exhaustiveness,
cohesiveness and redundancy) has to be verified by
operational tests
For this reason, ELECTRE III is not normally used until the problem (and/or the model) is structured and only when these conditions are satisfied does it become a powerful method to transparently compare solutions, in relation to all the different criteria, and to rigorously synthesize evaluations that are associated to the consequences of each decision.
Therefore, our proposal was: the AHP would be used in the problem definition step, when pertinent information has to be identified together with the client, by means analysis and selection of conceptual solutions, while the ELECTRE III would be used at the end, in
relation to the defined problem, when a decision has to activate the subsequent design process phases.
In the examined cases, the aircraft and the other systems were under production, or at least in the final phases of the production process, and the nature of the missions was clear, since the MBDD knows the decision context very well. Therefore, the principal aspects of the evaluation problem were easily identified (SoS performance in relation to the operational scenarios of the missions, technical effectiveness in relation to the operational management process and life cycle costs) and their disaggregation into organizational and functionaleconomic components was visualized through the SW Expert Choice and its multilevel decision tree.
In order to support interaction with the client, three models were elaborated during some simulation sessions in the MBDD: an AHP-Expert Choice model that is sufficiently general to be used in different decision situations, with a decision tree that is articulated in five levels and twenty-six elementary components for the comparison of the solutions, and two models for ELECTRE III, with twelve criteria for the first case study and fifteen criteria for the second one. Different decisional scenarios were hypothesized, in order to analyze which impact could have on the result the importance that the criteria assumed.
The AHP-Expert Choice model and the results of some applications were then analyzed to understand their potentiality to facilitate communication between the MBDD and the different clients. The ELECTRE III results were examined in terms of robustness and reliability, and the models in terms of formal validity and consistency with the internal procedures of the company.
A client's involvement in the initial phase of an engineering design process is always important and has to be carefully managed. The temporal horizon to produce an innovation in the aeronautic sector always involves a difficult definition of the client’s needs and some risks in translating the needs into formal requirements. The analysis and comparison of some draft solutions is an effective approach to understand the client’s point of view and the general structure of his/her preference system. However, this approach requires time to elaborate understandable technical solutions, analyse them with the client and elaborate new solutions for a new collective analysis, in a learning cycle.
Complexity and uncertainty elements can have a negative impact on the problem definition in some decision situations, above all when different, and sometimes conflicting, points of view require the involvement of some specific competences, from the client’s organization, as a not easy, but almost obligatory course of action.
A structured procedure can support the acquisition of the different points of view and their translation into mathematical models and then into product requirements, and can prevent, or at least control, ambiguous specifications by an activity that has the aim of verifying the overall consistency of the models.
The opportunity to produce conceptual solutions in a short time (a solution requires only few seconds of calculation time), with the guarantee of technical acceptability and specific performance levels in relation to an objective, makes communication possible and effective in the engineering design process.
Mathematical models that use an intelligible language introduce a positive psychological effect, in terms of clear thinking structure and perception of the logical progress. At the same time they facilitate the traceability of the process steps and results.
The integrated use of linear programming and multiple criteria methods can make the active collaboration phase with the client more rigorous (no acceptable solutions are lost and the evaluations can be documented and used consistently) and efficient, because all the structured and partially structured indications can be introduced into the models and transformed, by means of the methods, into information for the decision process.
The MBDD is planning to test the new approach with its clients and our group will be involved in analysing the criticalities and opportunities.
CEUR Proceedings 4th Workshop HCP Human Centered Processes, February 10-11, 2011 16