How Electronic Companies can Improve Employee Abilities in the
Fields of Precision and Automation
Q. C. Hsu1*, Y. C. Kao1, R. J. Lai1 and J. J. Sheu2
1 Department of Mechanical Engineering 2 Department of Die and Mould Engineering
National Kaohsiung University of Applied Sciences 415 Chien-Kung Rd., Kaohsiung 807, Taiwan, ROC.
The electronic companies of Taiwan are well known to have the capacity of producing great products in high volume. These companies, including IC and 3C enterprises have played a very important role in the world market, because of well-trained staff, innovative products and processes, and sufficient high performance facilities. However, due to increasing global competitiveness, in order to maintain an edge, these companies have realized that multi-disciplinary training and education of employees is of great importance. Therefore, cooperative work with educational institution has been paid more and more attention on both sides. To meet the needs of electronic companies and to improve their capabilities in precision and automation, the National Kaohsiung University of Applied Sciences (KUAS) has developed a series of training and education courses to help these companies in enhancing the capabilities of their employees. These courses consist of: (1) Systematic design, (2) Computer-aided geometry design, (3) Computer-aided engineering of precision trimming processes, and (4) Image processing and machine vision. The courses have been conducted successfully and have been recognized by the chief executive officers with awards for effectiveness.
Keywords : training, CAD/CAE, QFD, image processing, machine vision.
With regards to systematic design, the emphasis is put on the patent analysis and the systematic design methodology. For the patent analysis session, after a brief introduction of patent system and patent laws, the content documented in a patent specification is explained and the techniques of patent analysis are illustrated, including the “on-line search” for a specific topic within some available databases. What is presented in the session of systematic design methodology is based on the philosophy of quality function deployment (QFD). Beginning with identifying the customer needs, systematic approaches to establishing the product specification and the function
* Correspondence author: Q. C. Hsu
Associate Professor, Department of Mechanical Engineering
National Kaohsiung University of Applied Sciences, 415 Chien-Kung Rd., Kaohsiung 807, Taiwan, ROC. Tel: +886 7 3814526 Ext 5338, Fax: +886 7 3831373, E-mail: email@example.com
structure of the product required are introduced. The effort is focused on the conceptual design of a novel product.
The difficulty of representing components is a major obstacle in establishing an integrated CAD/CAM system. Computer aided geometry design is a basic and useful tool for building a part. The kernel of a CAD/CAM software is the geometric models. There are two categories of software according to the kernel geometric model adopted, namely: wire frame model and solid model. In the wire frame system, the boundary curves are used to make up surfaces. A geometry model is constructed via many surfaces. On the other hand, a solid model is built by using the basic primitives such as extruded cube or rotational part. The Boolean operations and features like rounds and chamfers are added to modify the solid geometric model. It is essential to realize the constraints and requirements of a valid and unique solid or wire frame model. Basic knowledge of curves, surfaces and solids is discussed to help the engineers use these elements correctly and efficiently.
The third course: “computer-aided engineering of precision trimming processes”, focuses on the basis of both theory and practice. The outline of this course consists of: (1) Plasticity theory and basis, (2) Formability of materials, (3) Machinery, (4) Basis on trimming processes, (5) Manufacturing process parameters, and (6) Case studies and practices. The sixth section was designed to meet the needs of the company through, firstly, brain storming with the shop floor engineers and technicians who attend the course, and secondly, the analysis and organization of the most representative problems. Appropriate commercialized finite element software was adopted to provide the problem-solving environment so that the problems could be resolved satisfying the industry’s standard and would provide reliable results. This means that some of the company’s real problems, which may have been bothering the employees for some time, was applied to realize the effectiveness of the designed course.
The technology of image processing and machine vision has been widely used in Electronic companies for precise positioning and inspection. The theoretical backgrounds of the fourth course consist of image components, image resolution, gray level, pixel level operation, and also image processing library and software. Moreover, the CCD camera, light source and lens quality are also introduced. Three levels of practices have been conducted, namely: image processing, morphology operation and high level processing. The case studies of factory machines for applying image processing and machine vision are also discussed.
2. Systematic Design
Systematic design is a kind of creative approach to finding the solution to a design problem. What is presented in the session of systematic design methodology is based on the philosophy of quality function deployment (QFD). As shown in Figure 1, the design process begins with the clarification of the customer’s needs to elaborate the product specification. The general statements from the customers’ wishes or preferences must be translated into the engineering requirements in terms of product attributes and characteristics. To establish the product specification, the relative importance of these attributes must be identified and a competitive benchmarking chart can be constructed. Then it goes through the conceptual design stage to establish the function structures and to search for solution principles as well as to combine the concept variants. Patents are a rich and public source of technical information for this purpose. To be patentable, the invention must be
new, it must involve an inventive step, (i.e. be non-obvious,) and it must be industrially applicable. Patent analysis, another section of the course, will be useful to see what other concepts have been protected and therefore must be avoided or licensed.
Engineering requirements Z Z Z Z Z Z … Benchmark Product Customers’ needs Relative impo rtance nn unit nn
unit unit unit
A B C … 1 XXX ◎ ◎ 2 XXX ☆ 3 △ 4 ☆ ○ … △ ◎ Benchmark product A B C Target values
Figure 1. The philosophy of QFD
For each feature or function essential to the product, one can list the possible means or solutions. Draw up a chart containing all possible combinations of sub-solutions, the morphological chart or function-means matrix will be obtained, which represents the solution space for the product. To evaluate the alternatives, a set of criteria must be listed based on the design objectives including technical and economic factors. The relative weighting of each objective will be assigned and the performance parameters or utility scores for each objective should be established. The utility values of each alternative are calculated and form the basis of comparison .
During the stage of embodiment design the layout and forms of a technical product or system will be determined, and finally, all the detail drawings and production documents are completed in the detail design stage. To obtain a satisfactory result, the stages are often carried out in an iterative way.
Patent Law, the application documents for an invention patent or a new utility model patent include a copy of specification and necessary drawings. The specification required to apply for an invention patent contains, in addition to the claims, the prior art, the purpose of the invention, the technical content, characteristics and effects, which would enable persons skilled in the art to understand the contents of and to practice the invention concerned. The claims shall describe the object, technical contents and characteristics of the invention . The description of prior arts, detailed description of invention and claims in the patent documents are, on the whole, the best source of information for the novice. Before gathering and analyzing the information of the related patents, one should be acquainted with the International Patent Classification (IPC) system  and the procedure of patent application as well as the content of the patent document. Nowadays, the efficiency of a patent search has been greatly improved with the help of on-line connection to the database supplied by the government office or certain companies. The search strategy and available patent databases such as USPTO, DEPATIS are illustrated. Finally, the patent maps, i.e. the visual representations of patent analysis, are selectively demonstrated. For example, the published patents in the field of ink-jet printing are searched and grouped according to the IPC system to analyze the key technical domains and the main competitive companies, as shown in Table 1.
Table 1. The key technical domains and the main competitive companies in the field of ink-jet printing (in part)
IPC Applicant Publication number
B 4 1 J 0 2 B 4 1 F 3 1 G 0 1 D 1 5 G 0 6 K 1 9 G 0 9 F 0 9 H 0 1 R 0 4 H 0 2 K 2 9 H 0 4 N 0 1 H 0 5 K 0 7 HP 466482 ＊ HP 414685 ＊ HP 442401 ＊ ＊ ＊ HP 345543 ＊ HP 245070 ＊ HP 128285 ＊ HP 117123 ＊ LEXMARK 352368 ＊ LEXMARK 367289 ＊ LEXMARK 386949 ＊ LEXMARK 386948 ＊ LEXMARK 412633 ＊ LEXMARK 419424 ＊ LEXMARK 458898 ＊ LEXMARK 425354 ＊ … …
3. Computer-Aided Geometry Design
To show the designer’s concept of what a product will look like, it is natural to use sketches. In order to present and discuss the real shape of a product, a CAD model that can be rotated and examined closely is essential. The course of computer-aided geometry design offers the attendants the basic mathematical theories of a CAD model. The students can learn the capabilities and the limitations of different modeling functions through this course. The CAD systems may have different modeling functions and utilities, but the basic theories are exactly the same. After a discussion of geometric modeling theories, the Pro/Engineer CAD system was used to train the students to create their models. Three-dimensional curves are constructed to form the characteristic surfaces. A solid model of the part is created using bounded surfaces and Boolean operations.
3.1 Surface modeling
Curves are the basic elements of the surface model while a parametric equation or a nonparametric equation can be used to represent them. Lines, conic arcs and splines are used to construct a simple or complex curve. The smoothness constrains the choice of the components of a curve. Cutting a cone with a plane can obtain the conic sections, such as a circle, an ellipse, a parabola, or a hyperbola. A regular shape can be established via the conic curves; but, to construct a more flexible and complicated shape, we need to employ higher order curves. Some basic curves are introduced here.
A parametric spline of degree 3 can be expressed by
] ′ ′ + − + − − + − = 1 0 1 0 3 2 3 2 3 2 3 2 2 3 2 2 3 1 ) ( P P P P u u u u u u u u u u P (1)
where u is the parameter; P ,0 P1,P′ ,0 P′1 are the two end points and their tangent vectors, respectively. The coordinates and tangent vectors are needed to specify the curve. For a spline curve with more than two points, boundary conditions of the end points and the continuity requirement of the other interior points are solved to obtain the tangent vectors of all of the fitting points. The requirement of input data is specific but not intuitive. The shape of a curve is very precise if the amount of fitting points is given enough.
Bezier selected the Bernstein polynomial functions as the blending function and in the form as follows: i i n i n i i n i u u P i n P u B u P − = − = = ( )
∑(1 ) ) ( 0 , (2)
where Bi,n(u) is the Bernstein polynomial function, and P is the coordinate of the ii th vertex.
The vertices that are used to construct a Bezier curve are called control vertices (points). The shape of a curve can be constructed and modified with the coordinate of the control points. It is
more intuitive and simple to catch the designer’s imagination of the product’s shape. While the degree of a Bezier curve is determined by the number of control points and any single control point can affect the shape of the entire curve.
To overcome the drawbacks of a Bezier curve, Cox  and Boor  suggested a blending function with the recursive formula as follow:
) ( ) ( ) ( , 1 1 0 − + = ≤ ≤ =
∑n ik k n i i t u t u N P u P (3)
where t are the knot values, and i
≤ ≤ = − − + − − = + − + + − + + − + − otherwise 0 ) ( 1 (u) ) ( ) ( ) ( ) ( ) ( 1 1 1 , 1 1 , 1 1 1 , , n k i i k i k i k i i k i k i i k i t u t N t t u N u t t t u N t u u N (4)
The (n+k+1) knot values from t to 0 tn+k (gap value 1 to each other) need to be determined to construct the (n+1) blending functions. The degree of a B-Spline curve can be selected regardless of the number of control points. The shape of a portion of the curve is affected only by a finite number of control points. The gap between neighboring knots can be uniform or non-uniform; as a result, we have uniform B-Spline and non-uniform B-Spline curves. In a non-uniform B-Spline curve the knot values are modified or deleted and non-uniform gaps are produced.
A non-uniform rational B-Spline ( NURBS) curve is similar to a non-uniform B-Spline curves but the control points are given in the form of homogeneous coordinate (hx, hy, hz, h) in stead of (x, y, z). The equation of a NURBS curve is represented as follows:
) ( ) ( ) ( , 0 , 0 u N h u N P h u P k i n i i k i n i i i
∑= = = (5)
The additional homogeneous coordinates h give us more versatile modification of a curve. i The conic curves can be exactly constructed via the NURBS curve instead of approximating procedure of the other types of curves.
A surface model is constructed by using the above-mentioned curves as its boundary lines. These curves are adopted in the u- and v- directions like the threads of a net. The interior points are interpolated using the bounded curves with respect to the u- and v- parameters, respectively. A B-Spline surface with ( n+1)×(m+1) control points can be constructed as follows :
) ( ) ( ) , ( , , 0 , 0 v N u N P v u P m ik jl j i j n i
∑= = = (6)
where Pi,j are the control points required to establish the surface. The flexibility of a surface model is the most important characteristic to meet the requirements of designers. The awareness of the limitations and operation methods of a surface model is essential to construct product geometry efficiently and correctly.
3.2 Solid modeling
Solid modeling systems have created a closed volume, called a solid. The information of a location is inside or outside of the solid is self-contained in the solid model. As a result, the calculation of volume, mass and other physical properties is very simple. The basic primitives, such as a block, a cylinder, a cone and a sphere, can be retrieved very quickly via the definition of geometric dimension. The generation of surface- bounded solid, such as sweeping a curve along a trajectory, can enable the ability of modeling a complex solid shape. These functions are similar to the surface modeling system and are called boundary modeling. The features, such as making a round or a chamfer, are very intuitive during the design phase. The method of creation of a physical feature is called feature-based modeling. The geometric data of a solid can also be defined with the geometric constraints, such as the tangency relations between the neighboring shape elements. This approach is called parametric modeling.
The Boolean operation is a method of combining the created primitives. The set theory is applied during the Boolean operations, i.e. each primitive solid is assumed to be a set of points. Thus, the results of a Boolean operation give a solid composed of the points and assure the validity of the solid model. The Boolean operations usually used in a solid modeling system are union, difference and intersection. The concept of half space allows the solid to be cut by a plane into two parts. Users can keep one or both of them.
There are three types of data structures used to store the solid modeling data , the first is the constructive solid geometry (CSG) representation, and the second is the boundary representation (B-Rep), the third is a decomposition model.
The CSG representation stores the tree-like history called CSG tree to build the solid model. The advantages of a CSG data structure are, first, the data structure is simple and compact, and secondly, the solid model in a CSG tree is always valid. But by using a Boolean operation alone for a CSG solid model system, the range of shapes that can be modeled is restricted. To derive the information of the boundary surfaces in a CSG tree requires a lot of computation.
On the other hand, the boundary and the connectivity information of a solid are stored in the B-Rep data structure. The data in any B-Rep system can be classified as either geometry information or topology information. The former defines the surfaces or curves of the boundary. The latter gives the interrelationships among the faces, edges and vertices. The surface information is always clearly defined in a B-Rep system. The data structure of B-Rep is most suitable for planar polyhedra. If a solid contains surfaces and curved edges, the data structure has to be modified to include these equations of surfaces and curves. As a result, the data structure is more complicated.
A decomposition model describes the solid model approximately as an aggregate of simple solids such as cubes. This type of model is simple and it’s easy to calculate the properties. To model the real shape of an object, the size of cubes or cells must be very small and therefore the memory space required increases dramatically.
4. Computer-Aided Engineering of Precision Trimming Processes
To enhance the capability of the company’s employees in computer aided engineering technologies, a customized course entitled “computer-aided engineering of precision trimming processes” was created to focus on the basis of both theory [7,8] and practice. This course consisted of six topics: (1) Plasticity theory and basis, (2) Formability of materials, (3) Machinery, (4) Basis on trimming processes, (5) Manufacturing process parameters, and (6) Case studies and practices.
The six topics listed above were formed based on intensive discussions during meetings with the managers of the company so as to meet their needs. The idea was to assist the company to build up their own core engineering capabilities on computer aided technologies especially on CAE so as to be competitive in the 21th century . Firstly, the basis on plasticity theory was arranged to give the employees the opportunity to learn the theory behind the practice. Plasticity is the basis in familiar with the trimming processes. Secondly, the formability of the commonly used materials was introduced to improve the ability on choosing the right materials and proposing appropriate parameters in processes. The topic on “machinery” then followed to give them the capability to select the most appropriate production facilities with regard to production time and cost. “Basis on trimming processes” and “Manufacturing processes parameters” was then realized through “case studies” in encouraging the attendees to sort out successful development and also problematic issues in their previous and current development.
The software DEFORM (Design Environment for FORMing)  was adopted in the processes simulation and therefore is a hands-on exercise in using the software to resolve practical design issues. A few groups were formed and each group was asked to prepare a real case in the company. The groups were encouraged to combine inter-disciplinary employees so as to offer the opportunity to engage in multi- disciplinary thinking in resolving a problem. Most of the groups have learned a systematic approach from the classes and also from the case study since production parameters must be well prepared so as to run a successful simulation. Groups that were formed mostly from engineering department performed much better than those mostly from the non-engineering background staff at the beginning. However, through the presentation organized at the end of this course to publish the achievement of their case studies, sufficient information interchange was fulfilled and all the attendees were believed to be more confident in joining in the design team on the development of new products in the future.
5. Image Processing and Machine Vision
An object is grabbed as a two-dimensional image with a fixed height and width. The digital image is divided into a lot of individual discrete points (namely pixels), each of them has a defined coordinates and brightness (namely gray level) as shown in Fig. 2. The image processing discussed in the class is focused only in gray scale image due to industrial application rather than color image in commercial application. There are three important and different resolution terms discussed, namely: space resolution, brightness resolution, and sample rate resolution . The space resolution of an image can be defined as pixels in a fixed length, for example, 300 ppi implies 300 pixels per inch. The concept of brightness resolution is concerned with how accurately the digital pixel brightness was compared to the original brightness at the same location in the image.
fully dark and 255 means fully bright. The frame rate is defined as the rate at which the image is updated. A high-end image processor often raises the frame rate to 1/100 of a second. The image processing is therefore employed to obtain useful information from an image.
Figure 2. The digital image composition: defined pixel coordinates and gray levels
After the previous definitions and descriptions of image processing, three levels of image operation are discussed, namely: low level operations, pixel group operations, and high-level operations. The low level operations of images include complement image operation, contrast enhancement, histogram equalization, binary operation, etc . The complement operation is used to make negative image from positive one. Contrast enhancement is a pixel level process involving the addition, subtraction, multiplication, or division of a constant value to every pixel within the image. Histogram equalization expands a narrow region of gray level in source image into a full region of brightness resolution and makes the image more uniformly distributed. Binary operation thresholds the gray scale image into binary image. Fig. 3 shows the results of low level processing.
Source image Binary operation
Complement image operation Histogram equalization
Figure 3. The results of low level processing
The most important pixel group operations are spatial convolution, edge detection, and morphological operation. The calculation scheme and results of pixel group operations are depicted in Fig. 4. By changing the convolution mask, the low pass filtering, the high pass filtering, the horizontal line detection, and the vertical line detection can be performed. The high level operations are related to specific industrial applications, for example, the analysis of particle composition, the measurement of object dimension, the matching of model image, the detection of manufacturing defects and the positioning of target object. The morphological operation includes erosion, dilation and skeleton. The normalized correlation model  is used for pattern matching in target image looking for model image.
∑− − − = 2 2 2 2 I N M M I N M I IM N r (7)
where r is the normalized correlation factor, I is gray level in target image, M is gray level in model image, and N is the total number of model image.
Calculation scheme of pixel group operations
Source image After high pass filtering
Horizontal line detection Vertical line detection
Morphological operation Morphological operation
Figure 4. The calculation scheme and results of pixel group operations
The block diagram of machine vision equipments is depicted in Fig. 5. The object being processed is carried in transfer mechanism. After proper lighting and lens magnification, the image of the object is captured and shown on the monitor. Finally, the image is manipulated and operated according to the built-in algorithms of image processing program in the computer. Table 2 shows the most important attributes that composed of image hardware system.
Frame Grabber transfer mechanism rotation and translation specimen Light PC, PLC, Interface CCD Lens
Figure 5. Block diagram of machine vision system
Table 2. The most important attributes of an image hardware system
Transfer mechanism Translation, Rotation
Direction: front, back, side
Pattern: directional, diffuse, ring, co-axial, polarized Lighting
Type: UV, infrared, normal lamp, flash lamp, LED
Lens Distortion, uniformity, parallelism, magnification ratio, depth of field, depth of focus, field of view, resolution, working distance
Type: CCD, CMOS Scanning method: array, line Image chip
Other: size, digital or analog, color or monochrome Frame grabber Size, color depth, frame rate, transfer rate, buffer size, resolution
Computer CPU speed, memory size, HD size
6. Results and Discussion6.1 Systematic design
Based on the systematic design, the customers’ needs are identified, objectively rather than subjectively, and are used to define the product specification during the product development, which is critical for a successful product design. On the other hand, the patent analysis serves as a stimulus to the generation of solution concepts, and at the same time brings in the consciousness of intellectual property to avoid possible patent infringement, which is of great value for a high tech designer, especially a novice.
6.2 Computer-aided geometry design
Fig. 6 shows the boundary curves and surfaces of a cell phone. The size and the basic shape of the product are determined at this stage. It is ambiguous by nature of the wire frame model. Fig. 7 shows the solid model with the extruded buttons and cutout screen. The similar buttons are created with the array function. The extruded buttons are cut and rounded to obtain a smooth product surface. Fig. 8 gives the final model of the product. Further cutting and rounding create the smooth surfaces of the product.
Figure 7. The solid model created by using extrusions, cuts and rounds.
Figure 8. The smooth surfaces of the product are created by further cutting and rounding 6.3 Computer-aided engineering of precision trimming processes
The major problem in introducing this course on computer-aided engineering of precision trimming processes was mostly on the familiarity of software simulation application. A metal forming processes simulation software DEFORM was adopted in this course extending the application of another course on computer-aided geometry design. The attendee needed to prepare billet and die geometry beforehand and the lecturer then assisted them in applying appropriate processes parameters for smooth simulation. The attendees were encouraged to take their real or potential problems for the case studies and practices. The threshold in learning the computer-aided engineering technologies was on the transformation from the practices to get the proper processes parameters for simulation. However, the attendees showed very promising achievement under the supervision of the lecturer through the courses.
Figure 9 illustrates what the attendees on the CAE for trimming process simulation had achieved. This was a process used to verify how many cutting blades should be made for the
trimming processes. Firstly, the attendees tried one blade and then compared the result with two blades, as shown in Fig. 9; the case on two blades with different height was also simulated; Eventually it was decided that the second case that had two blades at the same level obtained better results.
The simulated results were presented at meeting held by the company and the chief officer of the company was delighted to see an achievement that was beyond his expectation and therefore encouraged their staffs to propose further collaborative projects with the university.
(a) Initial simulation step
(c) Last step showing the billet has been cut off Figure 9. Case study from the CAE course by the attendees 6.4 Machine vision for matching, measurement, and inspection
The common problems of image matching are: (1) specimen being out of searching region, and (2) failure of match. The solution strategies are as follows: (1) checking the transfer mechanism function, (2) adjusting the searching region, (3) cleaning the surface of specimen and adjusting the light source, and (4) resetting the model image or adjusting the matching score. Fig. 10 shows the model images of pattern matching for precise positioning of LCD parts. Fig. 11 shows the measurement of pins’ dimensions, for example: the length and the angle. The common problems encountered are: (1) un-clear image for measurement, and (2) disturbance of environment light source. The proposed solving strategies are implementing secondary lighting and the shield of the environment light source.
Figure 11. The measurement of pins’ dimensions of semiconductor part (length and angle)
The context of these courses focuses on solving real problems in the real world rather than focusing only on theories. All of the four courses have been conducted successfully and have also been recognized by the chief executive officers with awards for effectiveness. In this paper, all of the details including planning, execution, results and discussion have been presented.
The authors would like to thank the Hitachi Kaohsiung Electronic Co., which organizes one-year training classes with the National Kaohsiung University of Applied Sciences.
 Pahl, G., W. Beitz, “Konstruktionslehre, Methoden und Anwendung,” 4. Auflage, Springer, 1997 (in Germans).
 Anon., Article 22, Section 2: Application, Chapter II Invention Patent, Taiwan Patent Law.  Anon., International Patent Classification (IPC), 7th Edition, WIPO, http://www.wipo.int/
 Cox, M. G., “The Numerical Evaluation of B-splines,” J. Inst. Maths. Applics., Vol. 15, 1972, pp 95-108.
 de Boot, C., “On Calculating with B-Spline,” J. of Approx. Theory, Vol. 6, 1972, pp 52-60.  Lee, K., “Principles of CAD/CAM/CAE systems,” Addison Wesley Longman Inc., 1999.  Lubliner, J., “Plasticity Theory,” Prentice Hall, 1998.
 Hughes, T. J. R., “The Finite Element Method: Linear Static and Dynamic Finite Element Analysis,” Dover Pubns, 2000.
 Wright, P. K., “21st Century Manufacturing,” Prentice Hall, 2001.
 Anon., “DEFORM 2D System User’s Manual,” Scientific Forming Tech. Co., 2000.
 Baxes, G. A., “Digital Image Processing – a Practical Primer,” Prentice Hall, New Jersey, 1984.  Gonzalez, R. C., R. E. Woods, “Digital Image Processing,” Prentice Hall, New Jersey, 2002.  Anon., “Matrox Imaging Library – User Guide,” Matrox Electronic Systems Ltd., 1996.