Research

8 pages
10 views

A Knowledge-Based Functional Reasoning Strategy to the Conceptual Design of Mechanical Products

of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
A Knowledge-Based Functional Reasoning Strategy to the Conceptual Design of Mechanical Products
Transcript
   INTERNATIONAL CONFERENCE ON ENGINEERING DESIGNICED 01 GLASGOW, AUGUST 21-23, 2001 A KNOWLEDGE-BASED FUNCTIONAL REASONING STRATEGY TO THECONCEPTUAL DESIGN OF MECHANICAL PRODUCTS W. Y. Zhang, S. B. Tor and G. A. Britton  Keywords: Design process, design strategy, expert systems, mechanical product design. 1   Introduction Computer-Aided-Design (CAD) is a mature technology now and provides industry with advancedgeometric modeling and capturing techniques. It is well suited for the downstream stage of design. However, CAD technology is still not well developed for the upstream stage of design;that is, the initial and most abstract stage of the design process, starting with a desiredspecification and resulting in concept variants. This is because it doesn’t have the built-inintelligence to perform reasoning, and it lacks the knowledge to make decisions.Knowledge-based expert systems for the conceptual design of engineering systems have been avery active area of research for the past two decades. For example, Zhang et al  [1] have used anexhaustive search method to implement a knowledge-based conceptual synthesis strategy, whichcan derive all the feasible concept variants from a given design specification.In this paper the authors propose a new knowledge-based functional reasoning strategy, which is based on a heuristic search method, to automatically reason out physical behaviors from thedesired functions.   The reasoning process produces a chain of interconnected behaviors. Two behaviors are connected when there is compatibility between the functional output of one and thecorresponding functional requirement (i.e., driving input) of the other. The proposed heuristicsearch method can lead to a near optimum design solution more efficiently than the exhaustivesearch method used by Zhang et al  [1].This research project was developed using an expert system shell CLIPS (C Language IntegratedProduction System) [2]. A design example of a terminal insertion unit, which is part of anautomatic assembly system is presented to demonstrate the practicality of the proposed approach. 2   Functional reasoning in design Functional reasoning is the technology that adds functional concepts into model-based reasoningtechnology. Function is a crucial design characteristic in the study of functional reasoning.However, there is no uniform definition of function, with different researchers [e.g., 3, 4, 5]attributing different meaning either to indicate the purpose or the action of a design object.Designers generally agree that function is what  a design is going to do, while behavior is how a    design will do it, and there is tight coupling between function and behavior. A behavioralreasoning model linking function and structure has been presented by Gero et al  [3]. Thereforewe encapsulate structure (behavior actor), driving input (functional requirement) and functionaloutput (intended purpose) in a unified behavior object [1].Umeda et al  [4] developed a Function-Behavior-State (FBS) modeler to reason about function.The FBS modeler regards functional decomposition as of two types: causal decomposition andtask decomposition. Deng et al  [5] devised a Function-Environment-Behavior-Structure (FEBS)model, in which the causal decomposition of function [5] has been extended by incorporating theworking environment of the system-being-designed. The research reported in this paper alsoincludes the working environment as part of the functional model.The proposed knowledge-based functional reasoning strategy in this paper is based on the abovecausal decomposition and task decomposition, but differs in the following ways:1)   Given a desired function, the system first scans the object-oriented behavior base to search for a behavior with a functional output that matches the desired function. Only if no matching behavior can be found, then will the desired function be automatically decomposed into sub-functions using a domain specific function decomposition rule. This search strategy reducesthe possibility of combinatorial explosion that can occur during function decomposition.2)   The functional reasoning strategy is based on a heuristic search method, which can lead to anear optimum design solution more efficiently than the exhaustive search method. 3   Knowledge-based functional reasoning strategy In this section we present a knowledge-based functional reasoning strategy, which is based on aheuristic search method, to automate the conceptual design of mechanical products. Thereasoning strategy is implemented by causal behavioral reasoning process and functiondecomposition process, as is discussed as follows. 3.1   Knowledge-based causal behavioral reasoning process It is common in mechanical design that a single physical mechanical device implements severalfunctional requirements simultaneously (i.e., function integration)[6]. Thus it is preferable tocreate a behavioral configuration of mechanical devices which maximize function integration.This usually results in a more compact and less costly concept variant.The system always begins by searching the object-oriented behavior base for the matching device behaviors with functional outputs that match some or all of the functional requirements. Thesearch criteria include function integration (if applicable). If there are several sets of device behaviors which achieve all functional requirements, they will be retrieved to develop into agroup of new design alternatives with their driving inputs becoming new functional requirements.The system will then check whether some or all of functional requirements of these new designalternatives are satisfied in environment (matched by the environmental outputs), and match thecorresponding ones to update these new design alternatives. After adding these new design    alternatives, the system will rank all the unexplored design alternatives by increasing order of thesummation of the numbers of current functional requirements and current retrieved behaviors.The design alternative which ranks first is selected. If the selected design alternative consists of only physical specifications (i.e., a set of retrieved device behaviors) without any unsatisfiedfunctional requirement, it is adopted as the design solution and the process terminates.Otherwise, starting from current functional requirements of the selected design alternative, thesystem repeats the above causal behavioral reasoning process to seek all matching device behaviors again.Let’s illustrate the concept of basic operation of the causal behavioral reasoning process with a portion of a design example, which will be illustrated completely in Section 4. Suppose that thecurrent design goal is to achieve four functional requirements  F  11 :  Locate housing  ,  F  12 : Clamphousing  ,  F  13 :  Hold terminal  and  F  14 :  Insert terminal  (Figure 1a).Before the analysis of the causal behavioral reasoning process, we explain the notations of designalternative  A i :{(PHY:  B  j , …) | (FUNC:  F  k  , …,  B l   D m …, …), or both}. It means that the i th designalternative  A i may consist of physical specifications PHY, functional specifications FUNC, or  both of them. PHY may be composed of a set of retrieved device behaviors (e.g., the  j th device behavior   B  j ). FUNC may be composed of a set of direct functional requirements (e.g., the k  thfunctional requirement  F  k  ) and driving inputs of retrieved device behaviors (e.g., the m th drivinginput  B l   D m of the device behavior   B l  ).Typically, a design alternative  A i : {FUNC only} is the starting state (i.e., overall functionalspecifications). A design alternative  A i : {PHY only} is the completed state (i.e., design solution).A design alternative  A i : {PHY, FUNC} is the intermediate state (i.e., composed of a set of retrieved device behaviors and a set of current functional requirements). In this example, the   selected design alternative node  A 1 consists of four functional requirements  F  11 :  Locate housing  ,  F  12 : Clamp housing  ,  F  13 :  Hold terminal  and  F  14 :  Insert terminal  , i.e.,  A 1 : {FUNC:  F  11 ,  F  12 ,  F  13 ,  F  14 }.For all functional requirements  F  11 ,  F  12 ,  F  13 and  F  14 of design alternative  A 1 , the system scans the behavior base to search for the matching device behaviors with functional outputs that matchsome or all of functional requirements. It is found that the device behavior   B 1 :  Housing insert-locating device , B 2 :  Housing slide-clamping device ,  B 5 : Terminal holding device and  B 6  : Terminal inserting device can respectively achieve  F  11 ,  F  12 ,  F  13 and  F  14 . The device behavior   B 3 :  Housing locating and clamping device can achieve both  F  11 and  F  12 simultaneously. The device behavior   B 4 : Terminal holding and inserting device can achieve both  F  13 and  F  14 simultaneously.Because  B 1 ,  B 2 and B 4 can mutually achieve all functional requirements, they are retrieved todevelop into a new design alternative  A 1.1 : {(PHY:  B 1 ,  B 2 ,  B 4 ), (FUNC:  B 1  D 1 ,  B 2  D 1 ,  B 4  D 1 )} with  B 1  D 1 :  Provide translational motion (first driving input of   B 1 ),  B 2  D 1 :  Provide translational motion  (first driving input of   B 2 ), and  B 4  D 1 :  Provide translational motion (first driving input of   B 4 ) becoming new functional requirements. Similarly  B 3 and  B 4 are retrieved to develop into other new design alternative  A 1.2 : {(PHY:  B 3 ,  B 4 ), (FUNC:  B 3  D 1 ,  B 4  D 1 )} with  B 3  D 1 :  Fix the device and  B 4  D 1 :  Provide translational motion becoming new functional requirements. Suppose theenvironment  E  1   can satisfy  B 3  D 1 , design alternative  A 1.2 is subsequently updated to be  A 1.2 :{(PHY:  B 3 ,  B 4 ), (FUNC:  B 4  D 1 )}. Similarly, two more design alternatives  A 1.3 : {(PHY:  B 3 ,  B 5 ,  B 6  ), (FUNC:  B 5  D 1 ,  B 6   D 1 )} and  A 1.4 : {(PHY:  B 1 ,  B 2 ,  B 5 ,  B 6  ), (FUNC:  B 1  D 1 ,  B 2  D 1 ,  B 5  D 1 ,  B 6   D 1 )}    are developed from  A 1 . Where,  B 5  D 1 :  Provide translational motion ;  B 6   D 1 :  Provide translational motion .Since there are now four unexpanded design alternatives  A 1.1 ,  A 1.2 ,  A 1.3 and  A 1.4 , the system willrank them by increasing order of their heuristic function $ h (i.e., summation of the numbers of current functional requirements and current retrieved behaviors). Because $ . h 12 = 1 + 2 = 3 < $ . h 13   = 2 + 3 = 5 < $ . h 11 = 3 + 3 = 6 < $ . h 14   = 4 + 4 = 8,  A 1.2 is selected as the best one for further causal behavioral reasoning. B 4 D 1 F 11  F 12  F 13  F 14 B 3  E 1  A 1.2  B 4 B 5 D 1 B 6 D 1  F 11 F 12 F 13 F 14  B 3 E 1 A 1.3  B 5 B 6  B 1 B 1 D 1 B 2 B 2 D 1 B 5 D 1 B 6 D 1 F 11 F 12 F 13  F 14  A 1.4  B 5 B 6  B 1  B 1 D 1  B 2  B 2 D 1  F 11  F 12  F 13  F 14  A 1.1  B 4 D 1  B 4  F 11 F 12 F 13 F 14 A 1.1   Figure 1a Illustration of causal behavioralreasoning process. F 11 F 12  F 13  F 14 A 1  F 1  A 0  F 1   Figure 1b Illustration of functiondecomposition process.Figure 1 Heuristic search tree in knowledge-based functional reasoning. Causal behavioral reasoning process or function decomposition processDesign alternativeHeuristic search path A B Conjunction (A AND B)Functional requirementDevice behavior Environmental node Legend:    3.2   Knowledge-based function decomposition process Complex functional requirement which can not be matched directly by any device behavior isfurther decomposed into less complex sub-functions by means of relevant domain-specificfunction decomposition rules. A new design alternative will be developed with the sub-functions becoming new functional requirements. Then the system will check whether some or all of functional requirements of this new design alternative are satisfied in environment, and match thecorresponding ones to update this new design alternative. After adding this new designalternative, the system will rank all the unexplored design alternatives, and select the best one for further causal behavioral reasoning unless the selected design alternative is already a designsolution. If the previous complex functional requirement can’t be decomposed by anycorresponding function decomposition rule, this design alternative will be discarded. Then thenext best design alternative will be selected for further causal behavioral reasoning unless it isalready a design solution.Figure 1b shows a portion of a design example, which will be illustrated completely in Section 4.Suppose a currently selected design alternative is  A 0 : {FUNC:  F  1 }. Where,  F  1 :  Insert terminal into housing  . The system begins by searching the object-oriented behavior base for the matchingdevice behaviors whose functional output matches  F  1 . Assume no matching is found after scanning the whole behavior base, one domain-specific function decomposition rule is fired todecompose function  F  1 into less complex sub-functions  F  11 :  Locate housing  ,  F  12 : Clamp housing  ,  F  13 :  Hold terminal  and  F  14 :  Insert terminal  . Now design alternative  A 0 is expanded to a newdesign alternative  A 1 : {FUNC:  F  11 ,  F  12 ,  F  13 ,  F  14 }. After adding this new design alternative, thesystem will rank all the unexplored design alternatives, and select the best one for further causal behavioral reasoning.   4   A design example This section discusses a complete design example to demonstrate the applicability of the proposed functional reasoning strategy. The product-to-be-designed is a terminal insertion unit,which is part of an automatic assembly system for manufacturing electronic connectors. 4.1   Problem description and user input Assume the following design specifications are given:1)   Design task is to design a terminal insertion unit with an overall functional requirement of   Insert terminal into housing  .2)   The environment can provide the following environmental outputs:  E  1 ’s environmental output: Fix the device .  E  2 ’s environmental output:  Provide pneumatic air  . 4.2   Knowledge-based functional reasoning strategy The proposed knowledge-based functional reasoning strategy is illustrated by a heuristic searchtree (Figure 2). Recall from Section 3 that two portions of this complete design example arequoted here. The following are its logical steps.
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x