Exploring design spaces in the search for embodiment design solutions and decision support Dominique SCARAVETTI, Patrick SEBASTIAN, Jérôme PAILHES, Jean-Pierre NADEAU Laboratoire inter-établissements TREFLE, UMR CNRS 8508 Esplanade des Arts et Métiers, F-33405 Talence cedex, France Phone: +33 5 56 84 54 22, Fax: +33 5 56 84 54 36, E-mail:
[email protected] Abstract – The design of an air-conditioning system for a commercial aircraft, which functions in various life cycle stages, proves difficult because atmospheric conditions differ greatly according to the flight phase. This paper proposes the use of constraint programming in the exploration of design spaces and the search for feasible embodiments. This approach produces the best design compromises between various design objectives, whatever the life cycle stage. The entire design space is explored for potential solutions and their performance assessed, thus allowing for decision support at the end of the embodiment design phase. Keywords: embodiment design, life cycle stage, CSP, decision support, aircraft air conditioning
I. INTRODUCTION The purpose of the embodiment design phase of a mechanical system is to define the system’s main characteristics and evaluate the global performances of the concepts selected that will satisfy the design requirements. It is an intermediate phase between concept research and detailed design [PAH, 96], a feasibility analysis combined with an estimate of the performances of a product concept. In the context of industry today this early phase in the life cycle of a product is of vital importance as it can to a large extent determine the successful development of the product (Hicks and Culley 2004). So, in the embodiment design phase, choices are made regarding working structure [PAH, 96], standard components and their characteristics, and structuring dimensions are determined. At the end of this design process a design configuration (product architecture) is produced. Some of the difficulties that are often encountered [SCA, 04] are as follows: (i) in a sequential design problem-solving mode, a priori choices are necessary, and these may obscure a large part of the potential design solution space; (ii) design solutions can only be validated when all the architectural choices have been made; this generates much repetition. Moreover, with a trial and error mode, the designer converges towards a solution which is not necessarily the optimum one. Design problems in embodiment design can be naturally expressed as mixed Constraint Satisfaction Problems [SCA, 04] [THO, 96] which cannot be solved using the traditional simulation tools used in mechanics. In this
field, designer-mechanics do not have the tools to help them make decisions with regard to the choice of concepts whose performances they have to assess [O'SU, 01]. The problem that we deal with here concerns the design of an aircraft air conditioning system, where we optimise performances through the choice of exchange surfaces. With the results obtained we were able to validate an improvement with regards to the complexity of system design problems processed with mixed CSP solvers based on interval analysis. II. AIR-CONDITIONING SYSTEM DESIGN FOR A COMMERCIAL AIRCRAFT The air-conditioning system studied (fig. 1) consists mainly of: (i) two cross-flow plate-type heat exchangers (main and pre-cooling exchangers), (ii) one turbine and one compressor coupled together, (iii) one filtration system to remove water from air, (iv) diffusers, nozzles, valves, and pipes to introduce outside air and pressurized air from the turbojet into the system. Fig. 1 shows the internal structure and the air flows across the system. These air flows are from the aircraft’s turboreactor (“main air”) and a scoop that draws in air from outside (“ram air”). The ram air flow crosses a diffuser, two heat exchangers (Main heat exchanger and Pre-cooling heat exchanger) which cool the main air flow
Fig. 1. Air-conditioning system for a commercial aircraft.
which is then returned to the atmosphere via a nozzle. The main air flow in turn follows a Joule-Brayton thermodynamic cycle. The pressure and temperature of the air from a turboreactor are high as it enters the system. The air undergoes pre-cooling before being sent into a compressor, a heat exchanger and a turbine which rapidly brings down the pressure, ensuring that the air is cooled before being sent into the cabin. The energy produced by the turbine is used to operate the compressor by means of a coupling shaft. This simplified system contains no ancillary components, more specifically, no elements to regulate the temperature or humidity of the air in the air conditioning system. The design problem that interests us here consists of optimising the internal structures of the heat exchangers while satisfying the functioning constraints imposed by the system environment. The exchangers consist of plain or finned plates stacked one on top of the other and through which main air and ram air flow in alternate layers. Fig. 2 shows the structure of the exchanger. Four life cycle stages corresponding to four dominant flight phases are identified by the aircraft manufacturer: Takeoff, cruising flight at low altitude, cruising flight at high altitude, landing. However, it is difficult for the designer to properly assess their performance which is very dependent on the context in which they are used. The physical functional effects that ensure heat transfer between the air flows in the exchanger can also produce harmful effects which may slow the flow of air through the exchanger (loss of load) and pressure may deteriorate. As the system has to work at several aircraft flight points and because of its complexity, it is a difficult component to design, particularly as regards the heat exchangers. For each air circuit (main air and ram air), the finned exchange surfaces are chosen from among six types (fig. 3) commonly used in aeronautics [KAY, 84]. These surfaces have a decisive influence on the performance of the air conditioning system as they ensure heat transfer to the elements that evacuate the heat outside the system. Since there are two exchangers, and two air circuits per exchanger, there are 1,296 (64) possible arrangements. In addition, materials may vary according to the temperatures encountered. Solving the problem manually would involve pre-setting values for some variables, in particular choosing the type of heat-transferring surface for each exchanger. Many such choices are made according to the experience of the designer or the company, and he proceeds by making many assumptions during the early phases [CHA, 92]. If a complete and detailed description of the problem is not available, the designer relies on existing concepts validated in earlier projects [FRE, 92].
Fig. 2. Structure of the cross-flow exchangers.
Moreover, system dimensioning is carried out for the most critical life cycle stage, the design risk or design hazard. Then, by successive applications, the designer checks that the required specifications are met. Finally, it is difficult to optimize design risk assessment for the most critical life cycle stage while also taking into account the constraints of the other life cycle situations. III. CSP SOLVING APPLIED TO ENGINEERING DESIGN In order to avoid the process of trial and error and not dismiss a solution because of pre-determined choices, we propose to take all the constraints into account simultaneously, without any hierarchy or preliminary choices, by modelling the problem as a Constraint Solving Problem (CSP). Using CSP we can evaluate various working structures, and examine all the design configurations that satisfy the system working conditions. Performance assessment and the reduction of the solutions space then facilitate decision-making. A. CSP applied to engineering design The constraints traditionally used to represent the designer’s input can be divided into three categories: - equal constraints (type “X=Y”) usually represent laws of physics or definitions of performance criteria, - unequal constraints (type “X473 K, or aluminium if T
