Elsevier

Applied Soft Computing

Volume 12, Issue 9, September 2012, Pages 2933-2947
Applied Soft Computing

Improving the lighting performance of a 3535 packaged hi-power LED using genetic programming, quality loss functions and particle swarm optimization

https://doi.org/10.1016/j.asoc.2012.04.023Get rights and content

Abstract

The lighting performance of a 3535 packaged hi-power LED (light-emitting diode) is mainly influenced by its geometric design and the refractive properties of its materials. In the past, engineers often determined the settings of the geometric parameters and selected the refractive properties of the materials through a trial-and-error process based on the principles of optics and their own experience. This procedure was costly and time-consuming, and its use did not ensure that the settings of the design parameters were optimal. Therefore, this study proposed a hybrid approach based on genetic programming (GP), Taguchi quality loss functions, and particle swarm optimization (PSO) to solve the multi-response parameter design problems. The feasibility and effectiveness of the proposed approach was demonstrated by a case study on improving the lighting performance of an LED. The confirmation results showed that all of the key quality characteristics of an LED fulfill the required specifications, and the comparison found that the proposed hybrid approach outperforms the traditional Taguchi method in solving this multi-response parameter design problem. The proposed hybrid approach can be extended to solve parameter design problems with multiple responses in various application fields.

Highlights

► A hybrid approach is proposed to solve the multi-response parameter design problems. ► The proposed approach is applied to improve the lighting performance of a 3535 packaged hi-power LED successfully. ► The proposed hybrid approach outperforms the traditional Taguchi method in solving multi-response parameter design problems.

Introduction

A light-emitting diode (LED) is a semiconductor light source initially created by Oleg Vladimirovich Losev in the 1920s [20]. In 1962, the first practical visible-spectrum (red) LED was invented by Nick Holonyak, Jr. while working at General Electric [11]. Early LEDs could only emit low-intensity red light; at present, various LEDs with very high brightness across the visible, ultraviolet and infra red wavelengths are available and extensively applied in diverse areas. For example, Fig. 1 illustrates a 3535 packaged hi-power LED, which is consisted of a lens and a single chip. Its lighting performance is mainly affected by (1) the geometric designs of the lens, chip, pad and base layer, and (2) the refractive properties of the lens and pad. These components require elaborate designs or settings to maximize the LED's overall lighting performance. In the past, the geometric designs and the selection of materials were determined by the design engineers based on the principles of optics and their own experience. The obtained optimal (but not real) design of the LED was then evaluated using optical simulation software, such as TracePro, to verify its optical characteristics. Based on the simulation results, the geometric designs and properties of the materials were revised and further fine-tuned through a trial-and-error process, in order to determine the final designs and settings. LED prototypes were fabricated in the last stage to verify the feasibility and lighting performance. This trial-and-error approach for designing an LED was costly and time consuming, and was unable to ensure that the geometric designs and the selection of materials were optimal. Therefore, it is important to solve the above problems arising in the design of a hi-power LED. Generally, the lighting performance of an LED is evaluated by several key optical characteristics (responses), thus, is considered a complicated multi-response parameter design problem. The Taguchi method is one of the traditional approaches to address such problems. For example, Su et al. [25] applied the Taguchi method to determine the optimal parameter settings in the fused process for fabricating fused bi-conic taper (FBT) couplers. According to the confirmation experiment and practical implementation results, the performance and reliability of 1% (1/99) single-window broadband tap couplers were improved significantly. However, some trade-offs had to be made through engineering judgments to deal with the conflict when selecting an optimal level of the control factor for simultaneously optimizing all responses. In other words, the Taguchi method would increase the decision-making uncertainty when solving the multi-response parameter design problems. Therefore, other studies have proposed various techniques to tackle multi-response parameter design problems. Kim and Lin [16] presented an approach based on exponential desirability functions to address the multi-response parameter design problem by optimizing the composition as well as the rolling and cooling conditions of JS-SS400-type steel. The comparison showed that the optimization results obtained by their proposed approach were better than those based on engineers’ experiences. Their proposed method was also applied to determine the effect of cysteine and calcium chloride on the texture characteristics of a dialyzed whey protein concentrate gel system. According to the desirability values, their approach significantly balanced all of the responses, in comparison with the results in Khuri and Conlon's study [15]. Tong et al. [27] proposed an integrated procedure based on principal component analysis (PCA) and the technique for order preference by similarity to ideal solution (TOPSIS) to deal with parameter design problems with multiple responses. They first used the signal-to-noise ratio (S/N) to evaluate the performance of each response, and applied PCA to the normalized S/N values to obtain a set of uncorrelated components. Certain principal components are retained based on the significance of the linear correlation between the responses and principal components and the cumulative variation of the responses accounted for by the selected principal components. The corresponding variation mode charts were then established to determine the optimization direction of the selected principal components. Finally, TOPSIS was conducted to obtain an overall performance index (OPI) for multiple responses, in order to determine the optimal factor/level combination. Their proposed solution procedure was demonstrated through a case study in which the chemical–mechanical polishing of copper (Cu-CMP) thin films from an integrated circuit manufacturer was optimized, and satisfactory results were obtained. Kovach and Cho [17] proposed an optimization approach, named the multidisciplinary–multiresponse robust design (MMRD), to solve the parameter design problems with multiple responses of different types. They utilized combined array designs to effectively incorporate noise factors into the robust design model. In addition, nonlinear goal programming techniques were employed to incorporate system specifications and desired target values as constraints and goals which were prioritized based on the inherent goal of robust design where the first goal was to minimize the variance. A case study on optimizing a chemical filtration process for measuring dosages was used to demonstrate their proposed methodology, and an adequate result that could address trade-offs between minimizing the variance and achieving the desired target value was achieved. Gaitonde et al. [9] proposed a method employing the Taguchi method and the utility concept to address the multi-response optimization problem. In their study, the optimal levels, as well as the importance of process parameters, were determined through the analysis of means (ANOM) and the analysis of variance (ANOVA) on a multi-response signal-to-noise ratio. The proposed methodology was demonstrated by a case study whose aim was to minimize surface roughness and specific cutting force by simultaneously optimizing the minimum amount of lubrication, cutting speed and feed rate during the turning of brass using a K10 carbide tool; acceptable results were obtained. Dubey and Yadava [6] presented a hybrid Taguchi method (TM) and response surface method (RSM), called TMRSM, in order to optimize multi-response problems. In their study, the Taguchi quality loss function was used to find the optimum level of input control factors which were then taken as the central values in RSM. The optimal settings of the parameters were determined by optimizing the second-order response model through RSM. Better results than those obtained using only the Taguchi approach were achieved in a case study aimed at improving a laser beam cutting process. Sibalija and Majstorovic [24] presented a systematic approach based on the Taguchi quality loss function, principal component analysis (PCA) and grey relational analysis (GRA) to solve a parameter design problem with multiple responses. The quality function that was then transformed into the proportion of quality loss (PQL) was adopted as a performance measure for multi-response optimization. The PCA was then performed on the normalized PQL (NPQL) to obtain the principal component score. Finally, GRA was employed to analyze the principal component score, in order to obtain the optimal parameter conditions. A thermosonic copper wire-bonding process at the bare copper leads was used to demonstrate their proposed approach, and the experiment confirmed the effectiveness of their method in reducing the uncertainty and complexity when determining the optimal parameter settings by the traditional Taguchi method. Hsu [12] integrated a back-propagation (BP) neural network, desirability functions, and genetic algorithms (GAs) to solve a multi-response parameter design problem involving printed circuit board (PCB) design improvements in the circuit from the north bridge (NB) to synchronous dynamic random access memory (SDRAM) in a personal computer. The BP neural model was used to estimate the functional relationships between major PCB design parameters and key quality characteristics, and the desirability functions were applied to evaluate the overall quality of the PCB design. In addition, GAs were utilized to acquire the (near) optimal PCB design parameters by exploring the mathematical model built by the BP neural network. The feasibility and effectiveness of the proposed procedure were demonstrated by a case study; the results showed that all of the quality characteristics very nearly approached their perfect states and achieved a 100% yield from the six sigma quality viewpoints.

The abovementioned studies proved that transforming multiple responses to an integrated quality index by some techniques, such as desirability functions, PCA, TOPSIS, multi-response S/N ratios and quality loss functions, was an effective way to address the multi-response parameter design problems. In this way, the original multi-response parameter design problem was converted into a parameter design problem with one response that could be tackled more easily. This study proposed a hybrid approach based on genetic programming (GP), Taguchi quality loss functions, and particle swarm optimization (PSO) to solve the multi-response parameter design problems. The GP technique was used to approximate the complex non-linear mathematical relationship between the input control factors and the output responses of a product/process. The overall quality of a product/process was then evaluated by Taguchi quality loss functions. Finally, the well-constructed GP models and Taguchi quality loss functions were further explored by PSO; the (near) optimal parameter settings of control factors for a product/process were thus found. The feasibility and effectiveness of the proposed procedure were demonstrated by a case study on improving the lighting performance of a 3535 packaged hi-power LED.

The remainder of this paper is organized as follows. Section 2 introduces the optimization techniques, including GP, Taguchi quality loss functions and PSO, needed for developing the hybrid approach; Section 3 presents the proposed approach; Section 4 evaluates the feasibility and effectiveness of the proposed approach by a case study of improving LED lighting performance; Section 5 discusses the comparison between the proposed approach and the Taguchi method in solving a multi-response parameter design problem; and finally, conclusions are given in Section 6.

Section snippets

Methodologies

This section introduces the techniques used for developing the integrated procedure to solve a multi-response parameter design problem.

Proposed hybrid approach

To solve the parameter design problems arising in the geometric designs and in the selection of the refractive properties of the materials for a 3535 packaged hi-power LED, this study proposed a hybrid approach based on GP, Taguchi quality loss function, and PSO. The principle of the proposed solution approach is depicted in Fig. 3, and the steps are detailed as follows:

  • Step 1.

    Identify the key quality characteristics (responses) of the product/process according to the improvement project.

  • Step 2.

    Identify the

Case study on improving LED lighting performance

In this section, a case study on improving LED lighting performance is presented to demonstrate the feasibility and effectiveness of the proposed hybrid approach. The details are described below, beginning with identifying the key quality characteristics.

Comparison

The Taguchi method (TM) is one of the most widely used traditional approaches for solving multi-response parameter design problems. To demonstrate the effectiveness of the proposed hybrid approach over the Taguchi method, the simulation results, as partially exhibited in Table 3, were further analyzed through TM. Table 10 summarizes the combinations of the design parameter settings which maximized the lighting performance of an LED based on the signal-to-noise ratio (S/N) regarding each quality

Conclusions

LEDs have been widely used in lighting environments. To optimize overall lighting performance, LEDs must be designed elaborately. In the past, design engineers often determined the geometric dimensions of the lens, chip, pad and base layer, and selected the materials for the lens and pad through a trial-and-error process based on the principles of optics and their own experience. This traditional approach for designing an LED was costly and time consuming, and could not guarantee that the

Acknowledgements

The author would like to thank the National Science Council, Taiwan, ROC, for its support of this research under Contract No. NSC 99-2221-E-159-018, and also to thank Raymond Huang for his valuable assistance during this study.

References (27)

  • I. Ciglarič et al.

    Computer-aided derivation of the optimal mathematical models to study gear-pair dynamic by using genetic programming

    Structural and Multidisciplinary Optimization

    (2006)
  • M. Clerc et al.

    The particle swarm-explosion, stability, and convergence in a multidimensional complex space

    IEEE Transactions on Evolutionary Computations

    (2002)
  • A.K. Dubey et al.

    Multi-objective optimisation of laser beam cutting process

    Optics & Laser Technology

    (2008)
  • Cited by (0)

    View full text