INSTITUTE OF PHYSICS PUBLISHING J. Micromech. Microeng.

13 (2003) 509522

JOURNAL OF MICROMECHANICS AND MICROENGINEERING PII: S0960-1317(03)36842-1

Manufacturing process and material selection in concurrent collaborative design of MEMS devices
Xuan F Zha1,3 and H Du2
Singapore Institute of Manufacturing Technology, 71 Nanyang Drive, Singapore 638075 School of Mechanical and Production Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798
2 1

E-mail: zha@cme.nist.gov

Received 10 May 2002, in nal form 31 March 2003 Published 14 May 2003 Online at stacks.iop.org/JMM/13/509 Abstract In this paper we present knowledge of an intensive approach and system for selecting suitable manufacturing processes and materials for microelectromechanical systems (MEMS) devices in concurrent collaborative design environment. In the paper, fundamental issues on MEMS manufacturing process and material selection such as concurrent design framework, manufacturing process and material hierarchies, and selection strategy are rst addressed. Then, a fuzzy decision support scheme for a multi-criteria decision-making problem is proposed for estimating, ranking and selecting possible manufacturing processes, materials and their combinations. A Web-based prototype advisory system for the MEMS manufacturing process and material selection, WebMEMS-MASS, is developed based on the clientknowledge server architecture and framework to help the designer nd good processes and materials for MEMS devices. The system, as one of the important parts of an advanced simulation and modeling tool for MEMS design, is a concept level process and material selection tool, which can be used as a standalone application or a Java applet via the Web. The running sessions of the system are inter-linked with webpages of tutorials and reference pages to explain the facets, fabrication processes and material choices, and calculations and reasoning in selection are performed using process capability and material property data from a remote Web-based database and interactive knowledge base that can be maintained and updated via the Internet. The use of the developed system including operation scenario, use support, and integration with an MEMS collaborative design system is presented. Finally, an illustration example is provided. (Some gures in this article are in colour only in the electronic version)

1. Introduction
It is now widely accepted that the nal cost of a manufactured product is largely determined at the design stage. Designers
3

Present address: Manufacturing System Integration Division, National Institute of Standards and Technology, Mail Stop 8263, 100 Bureau Drive, Gaithersburg, MD 20899, USA.

will tend to conceive parts in terms of processes and materials with which they are familiar and may, as a consequence, not consider process and material combinations that may have proven more economic [1, 3, 13]. Microelectromechanical systems (MEMS) device design must be separated from the complexities of the fabrication sequence and packaging processes with consideration of different materials and processes [23, 28]. A MEMS device designer requires a high 509

0960-1317/03/050509+14$30.00

2003 IOP Publishing Ltd Printed in the UK

X F Zha and H Du

level of fabrication and packaging knowledge in order to embody a successful design. Furthermore, the development of even the most common MEMS device requires dedicated work to formulate a suitable fabrication sequence and packaging process. MEMS and microelectronic device performance analysis is rendered virtually useless if the material information used in the model is wrong. The error associated with the nal simulation results will correlate one-to-one with the error associated with material property estimation. However, most MEMS devices are currently modeled using weak analytical tools, resulting in a relatively inaccurate prediction of performance behaviors [23]. The MEMS design process is usually performed in a trial-and-error fashion, which requires several iterations before the performance requirements of a given device are nally satised. This non-ideal design methodology combined with the length of time and high costs associated with MEMS prototyping results in a very inefcient and ineffective scenario for commercial product development. With the development of MEMS, advanced simulation and modeling tools for MEMS design are urgently needed. The design and manufacturing (fabrication and packaging) of MEMS and microelectronic devices and systems need to improve considerably from their current primitive state [23]. The advanced simulation and modeling tools for MEMS design must provide an advisory service so as to help the designer select manufacturing processes and materials for MEMS devices [9, 29, 3237]. It is therefore signicant to develop an efcient method and system in a computer-aided concurrent collaborative environment for designers to use at the early stages of MEMS design. With the advent of wide-area networks and the Internetbased World Wide Web (WWW), it is believed that many of the largest beneciaries of MEMS technology will be rms that have no capability or core competency in microfabrication technology and access by these companies is critically important to the successful utilization of MEMS fabrication facilities. A mechanism or framework allowing these organizations to have responsive and affordable access to MEMS fabrication resources for prototyping and manufacturing is essential [23]. The advisory service system can be used to help communicate the abilities of new processes to designers using the Internet and Web, as advocated rst by Frost and Cutkosky [14]. The availability of suitable design tools combined with computer networks to provide access to high performance workstations and local or remote supercomputer capability can radically alter this situation. The Internet-based WWW enables developers to provide intelligent knowledge services [10, 3337]. Expert systems running on servers can support a large group of users who communicate with the system over the network, in which user interfaces based on Web protocols provide access to the knowledge servers for services. Expert system technology would lead to the development of many small and mediumsized advisory systems that could help many categories of novice users in performing expert-level tasks. Therefore, it provides an opportunity for making the manufacturing advisory expert system widely available via the Web. The aims of this paper are to develop a knowledge-based method and a Web-based advisory system for selecting suitable 510

processes and materials for MEMS devices in concurrent collaborative design environment: (1) to explore a new knowledge intensive intelligent methodology for estimating and ranking manufacturing processes/materials; (2) to develop a new clientknowledge server architecture and framework for manufacturing process/material selection; (3) to develop a prototype advisory system for manufacturing process/material selection using Java and CORBA over the Internet and WWW; (4) to integrate the developed manufacturing advisory system into a self-developed MEMS design system. The system is especially dedicated for generating process and material selection advice during the embodiment design of micro-machined components. The remaining parts of this paper are organized as follows. We begin with a review of the existing MEMS design systems for process and material selection. Then we discuss concurrent design framework, manufacturing process and material hierarchies for MEMS, selection strategy, method, and implementation of such a strategy and fuzzy decsion support method including the development of a Web-based prototype advisory service system and its use. Finally we summarize some conclusions and observations.

2. Current status of research

In this section, the previous and current work related to manufacturing process and material selection is briey reviewed. The focus is on an overview of the functionality of the current MEMS design systems related to manufacturing process and material selection for MEMS devices. Generally, the selection of a suitable processmaterial pair to manufacture a component or device is not a straightforward matter [5, 6, 12, 30]. There are many factors which need to be considered at the design stage, for example, the size of component, the material to be processed and tolerance on dimensions. While all processes have slightly different capabilities, there is also a large overlap; for many components there are a large number of processes which can be used. Methods and software tools for process selection stem from the more widespread use of computer tools to assist with material selection. Some popular commercial tools are documented for material selection [1]. Process selection tools are more rare than the material selection tools such as Computer-aided Material and Process Selection (CAMPS, [2]), Design Advisor (DA, [38]), Material and Manufacturing Process Selection (MAMPS, [15, 16]), Cambridge Process Selector (CPS, [12]), Process Sequencing Expert Shell (PSES, [13]), and Manufacturing Advisory System (MAS 1.0 & 2.0, [4, 7, 30]). These systems are mainly dedicated to common (macro level) product manufacturing process and material selection. Details can be found in [30, 34]. For a micro level product (i.e. MEMS) manufacturing process and material selection, no such system is yet available although there are several commercial MEMS design systems such as MEMCADTM (now COVENTORWARETM) [8], IntelliSuiteTM [18], MEMSCAPTM [22], etc, which support integrated modeling and process. These systems have some

Manufacturing process and material selection in concurrent collaborative design of MEMS devices

functionality for MEMS manufacturing process and material selection or contain some embedded modules for this purpose. In what follows, we give an overview of the functionality of the above three MEMS design systems in manufacturing process and material selection for MEMS devices. DESIGNERTM is Coventors powerful front-end design tool for creating layouts and three-dimensional (3D) models of manufacturable MEMS devices. From DESIGNERTM users/designers can create models for further analysis or export les for mask making and fabrication. Manufacturing processes and material properties can be added to create 3D models. The process emulator in it can emulate fabrication steps for MEMS devices and use standard processes of deposition and etching with control of bulk and thin-lm geometries and materials. The material properties database is used for proper identication of each process layer in the MEMS device. The layout geometry and fabrication process information are furnished for automatic 3D model generation and visualization. IntelliSuiteTM takes a process-oriented approach to MEMS design and analysis, which starts the design process not from device geometry but from fabrication machine settings. Incorporating process templates, material data, mask layout, and device analysis, it provides a platform for the entire design team to develop manufacturable devices. It is the only CAD for MEMSTM tool to address process parameters linked with thinlm material properties. Processes can be custom-designed one step at a time, or designers/engineers can draw from a wide range of foundry-ready process templates. Users can also create their own process ows from a database of over 70 process steps or by including custom process steps. IntelliSuites process and materials databases have become a major MEMS resource. MEMaterialTM, as a most comprehensive thin-lm materials database with over 70 materials included, can provide a vital link between the process parameters and the device behaviors. It allows users to predict mechanical, electrical, thermal, physical, optical, and other material properties as a function of processing parameters. As the database is extensive, based on measured data rather than constitutive relationships, users can expand with their own proprietary data. MEMSCAP offers two primary software MEMS design tool suites: MEMS XplorerTM, for UNIX workstations and MEMS ProTM, for PC operating systems. The two CAD software tools provide a system-level approach, enabling designers to develop new MEMS designs, integrate existing designs (intellectual property or IP), and couple them with the system electronics that will drive them. The tool suites offer a comprehensive and customizable design environment for the development and testing of MEMS-based products. MEMSCAP CAD tools are open-platform products that support leading electronic design automation environments used for integrated circuit (IC) development. These tools allow data sharing between system designers, IC designers, process engineers and MEMS experts, permitting earlier and consistent design checks between multidisciplinary teams. MEMSCAP also provides modular subsets of these tools for those customers not requiring full capability. Each of the tool suites has easy-to-use graphical interfaces for rapid design. MUMPStartTM is an all-in-one MEMS design kit.

MEMS Pros built-in Technology Manager permits targeting of specic process technologies. From the above overview of the current MEMS design systems, their functionality for MEMS manufacturing process and material selection can eliminate a large amount of work during design, analysis, and simulation such as table lookup, data sharing, form lling, and process/material coding. However, they may still suffer one or more of the following drawbacks: a heavy dependence on the experience and knowledge of designers/users; no built-in ways for selection or decision-making knowledge representation; no mechanisms to utilize the given knowledge and guide the designer/user; difcult to modify, not easy to extend or update, and no mechanism to provide advisory services or explain the results and what-ifs. Also, they are generally specialized and standalone applications, and still lack effective and efcient methods for selecting manufacturing processes and materials and their combinations. It is very difcult for designers to use them for understanding and designing the integrated distributed performances of product systems when selecting different materials and processes. Thus, they are unable to support and coordinate highly distributed and decentralized collaborative MEMS design and modeling activities. To overcome the above drawbacks and improve the current method of selecting manufacturing processes and materials and their combinations, this research develops a knowledgeintensive decision support method and a Web-based advisory system to help designers/users collaborate and make rapid and more intelligent decisions in selecting manufacturing processes and materials for MEMS devices. The motivation and vision presented in this paper share some themes with the above macro-level process and material selection software system, especially MAS 2.0 [30] but focus on MEMS (microlevel), providing not only the choice but also advisory service. Details are discussed below.

3. MEMS design and manufacturing integration: concurrent design framework

MEMS is a hybrid of mechanical elements, sensors, actuators, and electronics on a common silicon substrate through the utilization of microfabrication technology. It promises to revolutionize nearly every product category by bringing together silicon-based microelectronics with micromachining technology, thereby making possible the realization of complete systems-on-a-chip [18]. In this section, we discuss issues of MEMS design and manufacturing and how to select materials and manufacturing processes. 3.1. Concurrent MEMS design framework As stated above, one of the challenges of MEMS technology is that MEMS devices or systems design must be separated from the complexities of the fabrication sequence and packaging processes with consideration of different materials and processes. In our previous work [3237], a knowledgeintensive methodology for design and simulation of MEMS devices was developed, and a Web-based knowledge intensive support framework was built up to support concurrent collaborative design of MEMS. In this concurrent 511

X F Zha and H Du

(a )

(c )

(b )

Figure 1. Hierarchies of MEMS manufacturing processes (a), (b) and their implementation (c).

collaborative design framework, an interface is designed to separate design from fabrication/packaging processes, which allows the designer to use process-independent design tools and methodologies [23, 28]. It will enable more manufacturable designs, correct the rst time or with fewer iterations, to become routine so that the amount of time and effort required to realize MEMS devices can be reduced. Since extensive knowledge of fabrication is no longer be a prerequisite before starting design activities, more designers are able to participate in design activities, and this will result in increased levels of innovation and creativity. Furthermore, the interface separating design from fabrication enables higher levels of integration without increasing development time or costs. To be of most utility, the interface allows designers to have ability and know the manufacturing implications of their designs at design time, and fabrication specialists can provide the needed functions to aid designers. 3.2. MEMS manufacturing process hierarchy While the electronics are fabricated using IC process sequences, the micromechanical components are fabricated using compatible micromachining processes that selectively 512

etch away parts of the silicon wafer or add new structural layers to form the mechanical and/or electromechanical devices. To divide the universe of all-possible semiconductor fabrication processes into increasingly specialized domains, a process hierarchy, as shown in gure 1, is established into which all-possible steps in the fabrication of MEMS devices, including metrology and testing, will t [23]. This kind of hierarchical process structure can be used to visualize and organize the process capability database containing information about the component processes, materials, and vendors, etc. It can bring two major benets: (1) to help familiarize newcomers to semiconductor fabrication with the technology and terminology; (2) to give experienced users a reference point to describe the scope of (potentially) available processes and the terminology for description. As shown in gure 1, the top level of the hierarchy includes deposition,etch,and pattern transfer,mask making, and other broad, high-level terms. Descending a layer,thedeposition sub-divides into chemical vapor deposition (CVD), epitaxy, andphysical deposition; the other high-level nodes in the hierarchy (or generic processes) divide similarly. As such, the designers or manufacturers allow expanding branches of the hierarchy and view a set of actual processes under

Manufacturing process and material selection in concurrent collaborative design of MEMS devices

All materials & processes

5 Parts

Copper

Days

Requirement Values

Screening apply limits (Eliminating, ranking, evaluating and selecting)

Fuzzy

Table Lookup

Integer

Rank Method

77

100

50

Subsets of processes or materials

Weighting Function

Requirement Rank

Supporting information & knowledge (Database & Knowledge base)

Merit Figure Option Rank

Figure 3. Rank methods.

Shortlist of candidates

Local conditions or requirements or constraints

and material performance in this stage. The kernel of the knowledge intensive method is a fuzzy ranking algorithm for multi-criteria decision-making problems. Details are discussed in the next section.

Final material and process choice

4. Knowledge support scheme for manufacturing process and material selection

In this section, we describe details about ranking methods and a knowledge decision support scheme for manufacturing process and material selection. The focus is on the fuzzy ranking algorithm for fuzzy multi-criteria decision-making problems. 4.1. Ranking methods and knowledge support scheme The method used in this research is based on ranking techniques [27, 20, 21] for evaluating and selecting possible process/material combinations for a particular component part in terms of total part costs. The basic ranking scheme is described as follows. The user enters design specications for one or more requirements (R). Each possible process/material can then be assigned a requirement rank, Rreq, based upon the requirements value, Vreq. To obtain the ranking for a process, each of its requirement ranks contributes to a weighting function. The output of this weighting function, the gure of merit, is used as the nal option rank, RO. The system repeats this for every option (O) each time a requirement value is changed. Figure 3 illustrates this process. Each requirement has a method for calculating the rank (R). The method parameters are process-dependent. A brief description of the methods used in obtaining the requirement ranks has been discussed in [30]: (1) Trapezoidal fuzzy numbers. This generates a rank for a requirement based upon the design specication and a trapezoidal membership function. This is actually a simplied mode from the fuzzy rank method (see section 4.2). (2) Boolean list membership. The option has a list of things it cando. Anything not on the list is impossiblefor the process to do. 513

Figure 2. The strategy for process and material selection.

each branch. Through the open Web database of process capabilities below, users/designers can search along different dimensions such as material deposited or etched. Thus, designers/users can also select processes from the database of process capabilities to construct a process sequence through a process sequence builder. 3.3. MEMS manufacturing process and material selection The selection of a suitable processmaterial pair to manufacture a MEMS component or device is not a straightforward matter. When selecting MEMS manufacturing processes and materials, a reasonable number of possible alternatives are available. The procedure of selecting manufacturing processes and materials for MEMS devices is to examine the alternatives against econo-technical criteria. It is actually a multi-criteria decision-making problem. Thus, the problem of MEMS manufacturing process and material selection can be dened as: given a set of alternatives, evaluate and select an alternative that can satisfy customer needs, meet design requirements and t the technical capabilities of a company. Figure 2 shows a general strategy for process and material selection. Accounting for the complexity of manufacturing process and selection, a variety of methods should be used to rank the appropriateness of an option with the value of a requirement. In this research, a comprehensive knowledge-intensive method is adopted, which is constructed using various decisionmaking methods to rank the options with requirement values and support knowledge. Due to the uncertainty and fuzziness of design specications and technical requirements in the early conceptual design stage, it is difcult to assess the process

X F Zha and H Du
Requirements & Constraints Secondary Requirements & Constraints

Eliminate Feasible Set Unacceptable Alternatives Base Set

Evaluate Candidates & Customization Ranks

Make Decision Final choice

Knowledge Decision Support System

Knowledge Source Required Differentiating Features Preferences & Importance (Weights) Trade- offs Utility Functions Heuristic Rules

Figure 4. Knowledge support scenario for process and material selection.

(3) Table lookup. For the more complex case of determining material/manufacturing process compatibility, a twodimensional array is used to look up a compatibility factor. (4) Integer programming. For qualitative requirements that have value ranges or orders of magnitude, such as production setup times hours or months. A single integer is used to represent the requirements value. The key elements of a process and material selection tool are composed of database and decision support systems/modules. The database support system communicates to the user with an extensive, up-to-date set of alternatives, while the decision support system concerns itself with evaluation, comparison and selection of alternatives. The decision support system consists of a multi-layered exploration and is knowledge-based. Figure 4 depicts a knowledge support scenario of process and material selection. The user may enter a bare minimum of data, just the batch size or the needed linear tolerance, and obtain initial feedback about the appropriateness of various manufacturing options [30]. However, should the designer wish to provide more information, they may ll in more requirements. Many of the requirements have an advanced mode to allow the users to more explicitly dene their requirements. The kernel of the knowledge-based decision support scheme is a fuzzy ranking algorithm for multi-criteria decision-making problems. Details about the fuzzy ranking methods are discussed below. 4.2. Fuzzy ranking method One of the well-known methods for multi-criteria decisionmaking is the procedure for calculating a weighted average 514

rating ri by use of the value analysis or cost-benet analysis introduced in [27]

n n

ri =
j =1

(wj rij )
j =1

wj

(1)

where rij denotes the merit of alternative ai according to the criterion Xj , and wj denotes the importance of criterion Xj in the evaluation of alternatives. But this procedure is not possible for the situations where uncertainty exists and the information available is incomplete. For example, the terms very important, good, or not good themselves are a fuzzy set. In what follows, the fuzzy ranking problem of a set of alternatives against a set of criteria is described [17]. Let a set of m alternatives A = {a1 , a2 , . . . , am } be a fuzzy set on a set of n criteria C = {C1 , C2 , . . . , Cn } to be evaluated. Suppose that the fuzzy rating rij to certain Xj of alternative ai is characterized by a membership function Rij (rij ), where rij R, and a set of weights W = {w1 , w2 , . . . , wn } are fuzzy linguistic variables characterized by Wj (wj ), wj R + . Consider the mapping function gi (zi ) : R 2n R dened by
n n

gi (zi ) =
j =1

(wj rij )
j =1

wj

(2)

where zi (w1 w2 . . . wn , ri1 ri2 . . . rin ). membership function (zi ) by

We dene the (3)

Zi (zi ) = n=1 Wj (wj ) o j k=1,...,n Rik (rik ).

Thus, through the mapping gi (zi ) : R 2n R, the fuzzy set Zi induces a fuzzy rating set Ri with membership function Ri (ri ) = sup
Zi g(zi )=ri

Zi (zi ),

ri R.

(4)

Manufacturing process and material selection in concurrent collaborative design of MEMS devices
MEMS Process and Material Database User/Designer MEMS Process Selection Java Java WebMEMS-MASS GUI (Netscape or IE) Jess/FuzzyJess Knowledge Base MEMS Material Selection Knowledge Base

TCP/IP Character Stream

Problem Solver

Cost Estimation Client Applet Knowledge Server

Figure 5. Clientknowledge server architecture for MEMS manufacturing advisory service system.

In this case, the nal fuzzy rating of design alternative ai can be characterized by this membership function. But it does not mean the alternative with the maximal R (ri )is the best one. The following procedures further evaluate the following two fuzzy sets: (1) a conditional fuzzy set is dened with the membership function; I /R (i|ri1 . . . rm ) = 1 if ri > rk , 0 otherwise; k (1, 2, . . . , m) (5)

5.1. System overview and architecture WebMEMS-MASS is a Web-based engineering reference tool for concept level MEMS manufacturing process and material selection. Based on various input parameters provided by the remote designer, WebMEMS-MASS determines which manufacturing processes are most relevant to the input part. The goal is to provide the designer with knowledge of the future production requirements of the part concurrently. The service provides advice that, rst, indicates which manufacturing process is most suited for the emerging MEMS design and, secondly, how the design could be best modied to satisfy the constraints of that particular MEMS process. WebMEMS-MASS can also be used as a library of manufacturing techniques since it contains detailed websites for more than 50 manufacturing processes. The main components of the proposed clientknowledge server architecture for MEMS manufacturing advisory system are shown in gure 5. The Java-based front-end (left) communicates with the knowledge server (right) through a TCP/IP stream. The knowledge server that is CLIPS-based uses a Java module for communication with the front-end. Each of these components interact with one another using a communication protocol (CORBA, Compliant Object Request Broker) so that it is not required to maintain the elements on a single machine. As a gateway for providing services, the graphics user interface (GUI) invokes the necessary actions to provide the requested services. To request a service, the system must have an interface pointer to the desired interface. 5.2. System implementation Initially, a simple demonstrative system was designed as a nal-year undergraduate project, implemented using intelligent website techniques such as form-based common gateway interface (CGI) and Javascript. The formal implementation of the system uses a Java applet as a frontend to the knowledge server, which provides a Web advisory system for process and material selection. The new formal implementation has several advantages over the original version. Two of them are that the system can run on standard hardware and can be available/accessible via the WWW. Moreover, because Java programs are portable, the system can run with Web browsers on multiple platforms. 515

(2) a fuzzy set is constructed with membership function R (r1 , r2 , . . . , rm ) = o i=1,...,m Ri (ri ). (6)

A combination of these two fuzzy sets induces a fuzzy set I which can determine a best design alternative with the highest nal rating, i.e. I (i) = sup I /R (i|ri1 . . . rm ) o R (ri1 . . . rm ).
ri1 ,...,m

(7)

Compared with equation (1), the fuzzy ranking for evaluation and selection is more exible and presents uncertainty better. Based on this method, the designer can use linguistic rating and weights such as good, fair, important, rather important, etc, for alternatives evaluation and selection. It therefore looks natural and attractive in practical use.

5. Web-based MEMS manufacturing advisory service system

An integrated expert system could be divided into a twocomponent architecture with a narrow communication channel [10]. The knowledgeserver approach separates the userinterface front end from the problem solver. In this research, a clientknowledge server framework is developed for the development of a Web-based MEMS manufacturing advisory service system (WebMEMS-MASS). Details are discussed in this section.

X F Zha and H Du

Figure 6. Startup of WebMEMS-MASS applet.

Figure 6 shows the startup window of the WebMEMSMASS applet. Java applets allow us to design user interfaces that are more interactive than CGI-based interfaces. The major purpose is to assist the designer selecting MEMS manufacturing processes and materials or to teach inexperienced designers or students basic skills or knowledge in MEMS manufacturing processes and materials and interactive rule-based programming, and even allow them to experiment with the knowledgeserver approach to implementing expert systems. When the designer/user has learned to select manufacturing processes and materials, he/she can continue with the design of a small knowledge base that performs the task automatically. Thus, the designer/user can augment the pre-existing system with appropriate database and knowledge base that rank and select processes and materials. A sample CLIPS rule that is used to select processes is shown as follows:
(defrule Rule X (goal (type identifyProcess) (value "yes")) => (printout t "Is the process cost high? | explanatory | Answer the question " "by selecting one of the choices and then clicking on 'proceed'. |yes |no |end") (assert (attribute (type hasHighcost) (value (readline)))) )

Figure 7 shows fuzzy facts and rules represented in FuzzyJess [25]. To test the selection rules, users can start several simulation rules. 5.3. Database and knowledge base for WebMEMS-MASS WebMEMS-MASS supports many widely used manufacturing processes [18, 19, 4], such as plastic injection molding, forging, sand casting, sheet metal forming, extrusion, micro milling, die casting, shell mold casting, investment casting, and electrodischarge machining (EDM). However, it should include the capabilities of new or less well-known processes at any time if necessary [30]. One of the goals of WebMEMSMASS is to educate a designer or student about new manufacturing methods. The assessment used for the development of WebMEMSMASS is to create a repository of manufacturing data separate from the code for the main program, which requires the remote process capability database containing information about the component processes, materials, and vendors. Microsoft Access is selected to develop the relational database. Opening the database on the server brings up a menu of tools which include the following components (gure 8) [30, 34]: (1) Vendor Editor, provides the account management for companies that have processes; (2) Material Editor, edit

516

Manufacturing process and material selection in concurrent collaborative design of MEMS devices
Very Low Low Medium High Very High

Surface Roughness

16

20

24

28

30

32

Microinches

(deftemplate surfRough 0 100 mi ( (very low (0 1) (4 1) (6 0)) (low (4 0) (6 1) ( 8 1)(16 0)) ... (very high (30 1) (>30 1)) ))

(defrule surfRough-rule (surfRough not low and not high) => (printout t "surfRough is acceptable" crlf) )

Figure 7. Fuzzy facts and rule in FuzzyJess for selection.

Figure 8. Database administrator tools menu.

Access Database

JDBC

Java Program

User

Java Application Server JDBC

SQL Calls

Vendor Applet JDBC

SQL

Capability W ebPage

Administration Client Applet Tools Complete Access

JDBC Bridge

Decision Rules
ODBC Source

WebPage DB Form
Process Capability Database

Client Applet Decision Rules WebPage

Process Characteristics

WWW Server HTML Files Simple Data Java Classes

Figure 9. Java database system scheme.

the properties of the generic raw materials; (3) File Exporter, generates data les and human readable reports;

(4) Process Editor, species the performance of all processes, compatibilities with materials, and locations of on-line 517

X F Zha and H Du

(a ) MEMS process database

(b ) Material database

Figure 10. MEMS process and material database Java applet launched.

resources. These tools can add a new process or material to WebMEMS-MASS databases without any changes to the compiled code. JDBC (Java Database Connection) is chosen as the method for implementing remote vendors that access to content in databases [3337]. A Java database system was developed by using a Microsoft Access database to store the detail information of processes and materials and a Java program to access the database through a JDBC connection. Figure 9 shows a pictorial view of the Java database system scheme. Two Java applets (gure 10) were developed as well to access the database from the internet browser. Figure 10(a) also shows a visualization of MEMS material properties retrieved from the database. The knowledge base is actually a rule base to choose manufacturing processes and materials, which is constructed using FuzzyJess. Figure 11 shows that the system is loading an external knowledge base. The fundamental information that forms the database of MEMS manufacturing processes and materials was obtained from the Internet resources [23, 18, 19, 24]. These data and information are used to demonstrate the features of the system and can be replaced by relevant versions when the system is customized for specic use. 5.4. Integration with WebMEMS-Designer system WebMEMS-MASS is developed mainly for intelligent selection of MEMS manufacturing processes and materials. 518

Figure 11. Loading of external knowledge base.

It has also been incorporated as a sub-system into a concurrent collaborative MEMS design system, WebMEMS-Designer [3337], which is being developed. WebMEMS-Designer has a unique combination of manufacturing and CAD, which allows the incorporation of true process data into fabrication simulation. MEMS Material, a comprehensive materials database available in [35], provides a vital link between the process parameters and the device behavior [18].

Manufacturing process and material selection in concurrent collaborative design of MEMS devices

(b ) Material Search

(a ) Process Search

Figure 12. Process and material search in WebMEMS-MASS.

6. Use of WebMEMS-MASS system

In this section, we present an operation scenario of WebMEMS-MASS system is presented, and we demonstrate how to use it for manufacturing process and material selection. The comparisons with the current MEMS design systems are given. 6.1. Operation scenario WebMEMS-MASS extends Jess/FuzzyJess, Java Expert System Shell [11, 25] with a GUI, which can be run as a standalone application or as an applet via the WWW. CLIPS or Jess rule bases have been modied slightly so as to work with WebMEMS-MASS. The running sessions are inter-linked with webpages of tutorials and reference pages explaining the facets, fabrication processes and material choices. WebMEMS-MASS performs calculations and reasoning using capability data from a remote Web-based database that can be maintained and updated via the Internet with collaborative support. A frame is provided for the rst-order cost estimation along with examples for selected processes, and the generation of process chains using secondary processes to rene certain features on a part. While running, WebMEMS-MASS generates a dialogue with the designer to inquire and acquire about batch size, typical

tolerances, size, overall shape, and cost requirements. After entering values for a set of facets, or attributes, for a conceptual part, the user is given real-time feedback regarding plausible fabrication methods. Once a process is selected, process chains or cost estimates can be explored [30]. At each step along the way, the user is presented with an updated, ranked list of manufacturing possibilities. A similar method is used to dene the attributes for material selection (yield strength, density, etc), and to generate material rankings. The nal result is a ranked list of viable combinations, obtained through a process-material pair optimization. 6.2. Support for WebMEMS-MASS To help users, a step-by-step tutorial provides instructions to users/learners who are not familiar with manufacturing terms. Samples are offered for an extensive on-line help manual (see gure 13). Descriptions and sample values are given for each of the process and material requirements, which allow users to compare their tolerance values with common products. Each of the processes included in WebMEMS-MASS has a set of descriptive webpages. The information includes production numbers, shape capabilities, design rules, sample parts, material usage notes, pros/cons, related processes, and links to equipment suppliers and fabrication sites. All of the documentation is linked through the applet itself. The 519

X F Zha and H Du

Figure 13. Final results with WebMEMS-MASS for microgripper.

designer may select any processes, materials, or requirements, and click the Get Website Info button to call up an informative webpage. To begin analysis of a design, the user should start the system. The instruction for the use of the system and four simple steps are shown in the startup screen in gure 6. 6.3. Illustration example To illustrate the use of WebMEMS-MASS, a design example of a prototype microgripper is explored to show the possibilities for making a full production run. The gripper is subject to real use in the laboratory test for MEMS experiment [31, 37]. Thus, it is necessary to use the process search, material search, and results survey mode. The specications are made for both the procedures of process search and material search as shown in gures 12(a) and (b), respectively. The process search is for the lowest possible cost over a long production cycle. At the end of the process search, EDM (rank 100%) is ahead of the only other possibility, etch (rank 96%). Similarly, after material search, the system generates six viable materials, with carbon steel ranking the highest at 100% and aluminum and alloys ranking at 98%. Furthermore, after process search and material search, the Results Survey button would be enabled and clicked to combine the results of both searches to nd the best material/process combination. The two boxes at the top of 520

gure 13 give a summary of the viable materials and processes. The nal box lists all of the feasible combinations, taking into account a compatibility factor between each process and material. EDM with carbon steel is the best choice, with etched stainless steel second. 6.4. Results and discussion This work differs from existing commercial MEMS design systems such as COVENTORWARETM (MEMCADTM), IntelliSuiteTM and MEMSCAPTM, which support integrated design and process modeling. WebMEMS-MASS is knowledge-based, and embodies an effective and efcient method and mechanism to select MEMS manufacturing processes and materials and their combinations. The system can provide an advisory service and explain the results and what-ifs. Specically, it is able to provide a common language at the concept level, allowing a designer to describe a part so that an expert advisory/consultant system can decide which manufacturing processes/materials can produce the desired part, in the desired time, with the desired quality. This means that WebMEMS-MASS is designed as a tool for nding a good fabrication method for a part while still at the conceptual level of design, and making a diverse catalog of processing capabilities available to designers/users so that they can

Manufacturing process and material selection in concurrent collaborative design of MEMS devices

experiment with different fabrication technologies. New processes that have not yet achieved widespread understanding in the engineering community can also make their debut through such service system. The widespread use of WebMEMS-MASS is likely to lead many material suppliers to put database searches on-line, allowing users to lter inventories based on user-entered material property ranges. Also, WebMEMS-MASS allows developers to provide intelligent knowledge services and an open environment to support and coordinate highly distributed and decentralized collaborative MEMS design and modeling activities for designers/users. The Web-based interface lets designers/users assemble process sequences and submit them for review by MEMS engineers and fabrication sites if necessary. Thus, WebMEMS-MASS provides remote users advice that: (1) indicates which manufacturing process is the most suited to the emerging design; (2) shows how the design could best be modied to satisfy the constraints of that particular process.

References
[1] AMPTIAC Newsletter 1997 MaterialEASE Insert, Advanced Materials and Processes Technologya DoD Information Analysis Center vol 1, no 3, 3rd Quarter [2] Bock L 1991 Material process selection methodology: design for manufacturing and cost using logic programming Cost Eng. 33 914 [3] Boothroyd G, Dewhurst P and Knight W 1994 Product Design for Manufacture and Assembly (New York: Dekker) [4] Brown S M and Wright P K 1998 A progress report on the manufacturing analysis service J. Manuf. Syst. 17 38998 [5] Calister W D 1991 Materials Science and EngineeringAn Introduction (New York: Wiley) [6] Chen Y and Wei C 1997 Computer-aided feature-based design for net shape manufacturing Comput. Integr. Manuf. Syst. 10 14764 [7] BMI (Berkeley Manufacturing Institute) 2002 Webpage http://kingkong.me.berkeley.edu (accessed on 20 March 2003) [8] COVENTOR, Inc. 2003 Webpage http://www.coventor.com or http://www.memcad.com (accessed on 20 March 2003) [9] Da Silva M G, Giasolli R, Cunningham S and DeRoo D 2002 MEMS design for manufacturability Sensor Expro and Conference (2002) [10] Eriksson H 1996 Expert systems as knowledge servers IEEE Expert 14 149 [11] Friedman-Hill E J 1999 Jess, the Java expert system shell Webpage http://herzberg.ca.sandia.gov/jess, Sandia National Laboratories, USA [12] Esawi A M K and Ashby M F 1998 The development and use of a software tool for selecting manufacturing processes at the early stages of design Proc. 3rd Biennial World Conference on Integrated Design and Process Technology (Berlin, Germany, 1 July) vol 3 pp 2107 [13] Farris J and Knight W A 1991 Selecting sequences of process and material combinations for part manufacture Proc. Int. Forum of Design for Manufacture and Assembly (Newport, RI, 1011 June) [14] Frost H R and Cutkosky M 1996 Design for manufacturability via agent interaction 1996 ASME Design for Manufacturing Conference (Irvine, CA, 1826 August) [15] Giachetti R E 1998 A decision support system for material and manufacturing process selection J. Intell. Manuf. 9 26576 [16] Giachetti R E and Jurrens K K 1997 Manufacturing evaluation of designs: a knowledge-based approach Proc. 3rd Joint Conference on Information Sciences (JCIS) (Research Triangle Park, NC, 15 March) vol 1 pp 1947 [17] Gui J K 1993 Methodology for modeling complete product assemblies PhD Dissertation Helsinki University of Technology, Finland [18] IntelliSuiteTM, IntelliSense, Inc. 2003 Webpage http://www.intellisense.com (accessed on 20 March 2003) [19] ISI 2003 Webpage http://mems.isi.edu/mems/materials (accessed on 20 March 2003) [20] Knosala R and Pedrycz W 1992 Evaluation of design alternatives in mechanical engineering Fuzzy Sets Syst. 47 26980 [21] Kickert W J M 1978 Fuzzy theories on decision making: a critical review, Martinus Nijhoff Social Sciences Division [22] MEMSCAPTM, MEMS CAP Inc., 2003, Webpage http://www.memscap.com [23] MEMS-Exchange, 2003, Challenges for future of MEMS technology, Webpage http://www.mems-exchange.org/ MEMS/challenges.html (accessed on 20 March 2003) [24] MEMSNET 2003 Webpage http://www.memsnet.org/ mems/challenges.html (accessed on 20 March 2003) [25] Manufacturing Advisory Service 2003 Webpage http://cybercut.berkeley.edu/mas2 (accessed on 20 March 2003)

7. Summary and future work

In this paper we have presented our preliminary work on the development of the knowledge intensive method and system for MEMS manufacturing process and material selection, which bring together engineering reference material and an informative education or learning tool over the Internet and WWW. The reference materials include basic process descriptions, special abilities, some simple design rules, and links to fabrication sites. The Java-based WebMEMS-MASS can provide a knowledge intensive intelligent dynamic environment for educating designers/students about the trade-offs available in different manufacturing processes. The applet of WebMEMS-MASS is potentially available to anyone, at any time and anywhere with a Java compatible browser so as to work in most computing environments. The system can be used for simple single parameter searches to select a process, and also process and material combined searches with secondary processes mapped to high-tolerance features. The rankings change in real time with user input but without the need to query a database for results [30]. The underlying WebMEMS-MASS databases and knowledge bases are extensible through administrator tools or via the Web which gives commercial manufacturing facilities the ability to update their own processes and rules. The designer/student can also submit on-line external manufacturing process and material selection knowledge bases for some specic new processes. A large amount of the future work will be dedicated to the enhancement of process and material databases and knowledge base and the further development of the system.

Acknowledgments
The authors would like to express their gratitude to anonymous reviewers of this paper for their insightful comments and suggestions that have helped to improve the paper.

521

X F Zha and H Du

[26] National Research Council of Canada (NRCC) 2003 Webpage http://ai.iit.nrc.ca/IR public/fuzzy/fuzzyClips/ fuzzyCLIPSIndex.html (accessed on 20 March 2003) [27] Pahl G and Beitz W 1996 Engineering DesignA Systematic Approach (New York: Springer) [28] Reithel B 2003 Challenges for future of MEMS technology Webpage http://faculty.bus.olemiss.edu/breithel/b620s02/ riley/Micro Electro Mechanical Systems.htm (accessed on 20 March 2003) [29] Senturia S D 1998 CAD challenges for microsensors, microactuators, and micro-systems Proc. IEEE 86 161126 [30] Smith C S 1999 Manufacturing advisory service: web based process and material selection PhD Thesis University of California at Berkeley, CA, USA [31] Su C 1999 Development of three MEMS devices: a microgripper, a micromechanism and a microaccelerometer MEng Thesis Nanyang Technological University, Singapore [32] Zha X F and Du H 1999 Knowledge intensive methodology for design and simulation of micro-electro-mechanical system (MEMS) devices Report-MEMS CAD/CAE/CE for Singapore National Science and Technology Board (NSTB, now A-STAR) project ARC 5/97 School of Mechanical

[33]

[34]

[35] [36]

[37] [38]

and Production Engineering, Nanyang Technological University, Singapore Zha X F and Du H 2000 Web-based knowledge intensive collaborative design modelling and decision support for mems Proc. Int. Conf. on Engineering and Technological Sciences (Beijing, China) vol 1 pp 8092 Zha X F and Du H 2001 A world wide web-based manufacturing consulting service system for processes/materials selection in concurrent design for manufacturing Proc. Int. Conf. on Materials for Advanced Technologies (ICMAT 2001) (Singapore) Zha X F and Du H 2001 Web-based knowledge intensive collaborative design framework for MEMS Proc. Int. Workshop on MEMS 2001 (Singapore) pp 50313 Zha X F and Du H 2001 A world wide web based database system for fabrication/packaging processes/materials selection in concurrent collaborative design of MEMS devices Proc. Int. Conf. on Materials for Advanced Technologies (ICMAT 2001) (Singapore) Zha X F and Du H 2002 Web-based knowledge intensive support framework for collaborative design of MEMS J. Micromech. Microeng. 12 51224 Kunchithapatham A 1996 A manufacturing process and materials design advisor, MS Thesis UC Berkeley

522