This article appeared in a journal published by Elsevier.

The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elseviers archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy

Engineering Failure Analysis 17 (2010) 369379

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Failure analysis and redesign of the evaporator tubing in a Kimchi refrigerator

Seong-woo Woo a,*, Dennis L. ONeal b, Michael Pecht c
a

SAMSUNG Electronics Co., Ltd., 272, Oseon-Dong, Gwangsan-Gu, Gwangju-City 506-723, South Korea Department of Mechanical Engineering, Texas A&M University, College Station, TX, USA c Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, College Park, MD, USA
b

a r t i c l e

i n f o

a b s t r a c t
A failure analysis was conducted on a pitted aluminum tubes in the evaporator used to cool, a Kimchi refrigerator. The root cause of the failure was a pitting corrosion which was traced to chlorine in a cotton adhesive tape used on the tubes. To reproduce the failure modes and mechanisms causing the tube pitting corrosions, a tailored set of accelerated life tests was applied to the evaporator tubing. Using chemical loads, the key noise parameters in the assembly, including a variety of chemical reaction formula, were analyzed. The failure modes and mechanisms found experimentally were identical to those of the failed sample. To correct the problem, the cotton tape in the cooling evaporator was replaced by a generic transparent tape. The B1 life of the new design is now guaranteed to be over 10 years with a yearly failure rate of 0.1%. Published by Elsevier Ltd.

Article history: Received 31 July 2009 Accepted 17 August 2009 Available online 21 August 2009 Keywords: Fitting corrosion Robustness Parameter design Accelerated life testing

1. Introduction Fig. 1 shows the Kimchi refrigerator with the newly designed cooling aluminum evaporator tubing. When a consumer stores the food in the refrigerator, the refrigerant ows through the evaporator tubing in the cooling enclosure to maintain a constant temperature and preserve the freshness of the food. To perform this function, the tube in the evaporator need to be designed to reliably work under the operating conditions it is subjected to by the consumers who purchase and use the Kimchi refrigerator. The evaporator tube assembly in the cooling enclosure consists of an inner case (1), evaporator tubing (2), Lokring (3), and adhesive tape (4), as shown in Fig. 1b. In the eld, the evaporator tubing in the refrigerators had been pitting, causing loss of the refrigerant in the system and resulting in the loss of cooling in the refrigerator. The data on the failed products in the eld were important for understanding the usage environment of consumers and pinpointing design changes that needed to be made to the product. Robust design techniques, including statistical design of experiment (SDE) and the Taguchi methods [1], have been developed by statisticians for use in improving designs in products. The Taguchi methods describe the robustness of a system for evaluation and design improvement also known as called quality engineering [2,3] or robust engineering [4]. Robust design processes include concept design, parameter design, and tolerance design [5]. Taguchis robust design method places the design in a position where random noise does not cause failure and helps to determine the proper design parameters [6]. The basic idea of parameter design is to identify, through exploiting interactions between control factors and noise factors, appropriate settings of control factors that make a systems performance robust in
* Corresponding author. Tel.: +82 62 950 6933; fax: +82 62 950 6807. E-mail address: twin.woo@samsung.com (S.-w. Woo). 1350-6307/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.engfailanal.2009.08.003

Author's personal copy

370 S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379

Nomenclature AF BX C1 C2 Cl% F(t) F F1 F0 h h* k KCP KNP LB n N1 r S S1 S0 ti Tf x acceleration factor durability index length of contraction tube type of adhesive tape chlorine concentration, PPM unreliability corrosive force, N corrosive force under accelerated stress conditions corrosive force under normal conditions testing times (or cycles) non-dimensional testing cycles, h h=LB P 1 5 Boltzmanns constant, 8.62 10 eV/deg key control parameter key noise parameter target BX life and x = 0.01, on the condition that x 6 0.2 number of test samples corrosive force of evaporator tubing under customer usage pattern pitted numbers of the evaporator tubing in the test stress mechanical stress under accelerated stress conditions mechanical stress under normal conditions test time for each sample, h time to failure, h required target, x = 0.01, on condition that x 6 0.2.

Greek symbols g characteristic life Superscripts b shape parameter in a Weibull distribution h i @ lnT f n the stress dependence, n @ lnS
T

Subscripts 0 normal stress conditions 1 accelerated stress conditions

Fig. 1. Kimchi refrigerator (a) and the cooling evaporator assembly (b).

Author's personal copy

S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379 371

relation to changes in noise factors. Thus, the control factors are assigned to an inner array in an orthogonal array, and the noise factors are assigned to an outer array. However, a large number of experimental trials array may be required for the Taguchi method because the noise array is repeated for every row in the control array. Alternative methods, such as the combined array approach, have been proposed [7,8]. However, for a simple mechanical structure, a lot of design parameters should be considered in the Taguchi methods robust design process. Products with the missing or improper minor design parameters may result in recalls and loss of brand name value. Based on an analysis of failed products in the eld, accelerated life testing (ALT) with the new concept (BX) and sample size, an alternative parameter study can be considered [9,10,11]. Failure analysis of the eld data helps pinpoint the missing key control parameters in the design process. ALT may help identify the missing key control parameters of the newly designed mechanical system and the proper choices for its levels. The key control parameters precipitated by ALT may not represent those occurring in the eld because of inconsistencies in the types and magnitudes of the loads applied during testing. Moreover, the numbers of test samples and the test durations are usually insufcient to uncover some occasional failure modes. ALT should be performed with sufcient samples and testing time, and with equipment designed to match expected product loads. The steps for the robust design of chemical / mechanical systems can be summarized as follows: (1) analysis of the problems identied in eld samples; (2) analysis of the loads in the dynamic system; (3) ALT to identify the key control parameters; and (4) formulation of a corrective action plan to determine the level of the key control parameters. The robust design of a newly designed evaporator tubing system in the cooling enclosure of a Kimchi refrigerator was investigated in this study. Because the evaporator assembly is a simple mechanical heat exchanger system including water condensation on the evaporator tubing and their corrosions, it is important to model the load on the basis of thermodynamics, heat transfer, and chemical reaction formula. ALT equipment can then be fabricated on the basis of load analysis. First, uncontrolled corrosive load conditions on the system were analyzed using design schematics and chemical equations. The analysis then proceeded to new robust methodologies for parameter designs. After a sequence of ALTs and corrective action plans, key control parameters and their levels were summarized in the BX life. Finally, the effectiveness of these methodologies was demonstrated by developing a robust design for the evaporator assembly. 2. Field design problems In the eld, the refrigerant tubes in the evaporator assembly in a Kimchi refrigerator were experiencing pitting and the refrigerant was leaking out of the tubes (Fig. 2). The specic usage conditions of the refrigerators by consumers in the eld were unknown. Field data indicated that the damaged products might have had design aws. The design aws combined with the repetitive loads could cause failure [12,13]. The pitted surfaces of a failed specimen from the eld were characterized by a scanning electron microscopy (SEM) and EDX spectrum (Fig. 3). We found a concentration of the chlorine in the pitted surface (Table 1). When Ion Liquid Chromatography (ILC) was used to measure the chlorine concentration, the result for the tubing having had the cotton adhesive tape was 14 PPM. In contrast, the chlorine concentration for tubing having had the generic transparent tape was 1.33 PPM. It was theorized that the high chlorine concentration found on the surface must have come from the cotton adhesive tape. 3. Load analysis The evaporator tubing assembly in the cooling enclosure of the Kimchi refrigerator consists of many mechanical parts. Depending on the consumer usage conditions, the evaporator tubing experienced repetitive thermal duty loads due to the normal on/off cycling of the compressor to satisfy the thermal load in the refrigerator. Because the refrigerant temperatures are often below the dew point temperature of the air, condensation can also form on the external surface of the tubing. Fig. 4 shows a robust design schematic overview of the cooling evaporator system. Fig. 5 shows the failure mechanism of the crevice (or pitting) corrosion that occurs because of the reaction between the cotton adhesive tape and the aluminum evaporator tubing. As a Kimchi refrigerator operates, water acts as an electrolyte and will condense between the cotton adhesive tape and the aluminum tubing and the crevice (or pitting) corrosion will begin. The number of Kimchi refrigerator operation cycles is inuenced by specic consumer usage conditions. In the Korean domestic market, the compressor can be expected to cycle on and off 2298 times a day to maintain the proper temperature inside the refrigerator. The crevice (or pitting) corrosion mechanism on the aluminum evaporator tubing can be summarized as: (1) passive lm breakdown by Cl attack; (2) rapid metal dissolution: Al ? Al+3 + 3e; (3) electro-migration of Cl into pit; (4) acidication by hydrolysis reaction: Al+3 + 3H2O ? Al(OH)3; + 3H+; (5) large cathode: external surface, small anode area: pit; (6) the large voltage drop (i.e., IR drop, according to Ohms Law V = I R, where R is the equivalent path resistance and I is the average current) between the pit and the external surface is the driving force for propagation of pitting (Fig. 6). Because the corrosion stress of the evaporator tubing depends on the corrosive load (F) that can be expressed as the concentration of the chlorine, the life-stress model (LS model) [14] can be modied as

T f ASn AFn ACl%n

Author's personal copy

372 S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379

Fig. 2. A damaged product after use.

The acceleration factor (AF) can be derived as

n  n  n  S1 F1 Cl1 % AF Cl0 % S0 F0
4. BX life and sample size The characteristic life g determined by maximum likelihood estimation can be dened as

g 

P r

tb i

nh r

As the product (or part) reliability improves, the number of failures decreases and there may be no failures in the laboratory. Thus, it is not appropriate to evaluate the characteristic life in Eq. (3). When the number of failed samples is below four, it is assumed to follow the Poisson distribution [15]. At a 60% condence level, the characteristic life can be redened as

gb

1 b nh r1

To introduce the BX life, the characteristic life in the Weibull distribution can be modied as

Author's personal copy

S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379 373

Fig. 3. SEM fractography showing a pitting corrosion on the evaporator tube.

Table 1 Chemical composition of the no pitting and pitting surfaces. No pitting Weight% O Al Cl Si Ca K Na Totals 11.95 87.29 0.33 0.42 Atomic% 18.65 80.74 0.23 0.38 Pitting Weight% 25.82 68.28 3.69 0.66 0.70 0.50 0.34 100.00 Atomic% 37.38 58.61 2.41 0.55 0.40 0.30 0.34

100.00

Lb x gb B

x b nh r1

To assess the BX life with about a sixty percent condence level, the number of test samples is derived from Eq. (5). That is,

 b 1 1 r 1 x h

with the condition that the durability target is dened as, h h=LB P 1.

Author's personal copy

374 S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379

Key Noise Parameters N1: Customer usage & load conditions N2: Environmental conditions Hot Temp Input Cooling Enclosure System Output Cold Temp

Key Control Parameters C1: Evaporator specifications

Fig. 4. Robust design schematic of a cooling enclosure system.

Fig. 5. An accelerating corrosion in the crevice due to low PH, high Cl concentration, depassivation and IR drop.

5. Laboratory experiments Generally, the operating conditions for the evaporator tubing in a Kimchi refrigerator are approximately 0 $ 43 C with relative humidity ranging from 0 $ 95%RH, and 0.2 $ 0.24 gs of acceleration. The compressor in a Kimchi refrigerator is expected to cycle on average 2298 times per day. With a life cycle design point of 10 years, the Kimchi refrigerator incurs 358,000 cycles (Table 2). The chlorine concentration of the cotton adhesive tape was 14 PPM. To accelerate the pitting of the evaporator tubing, the chlorine concentration of the cotton tape was adjusted to approximately 140 PPM by adding some salt. Using a stress dependence of 2.0, the acceleration factor was found to be approximately 100 in Eq. (2). The shape parameter was 6.41, and the test cycles and the numbers of samples [15] used in the ALT were calculated as follows:

 b 1 LB n r 1 x AF h

Fig. 6. Kimchi refrigerators in accelerated life testing.

Author's personal copy

S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379 Table 2 Operating cycles of the Kimchi refrigerator. Item Operating cycles (times) 1 day Normal Kimchi refrigerator 22 Worst 98 10 years Normal 80,300 Worst 357,700 375

20 Min F (140PPM)

. .
1 hour 4, 500 ALT Mission Cycles

Fig. 7. Duty cycles of repetitive corrosive load F on the evaporator tubing.

For B1 life, the required target x was 0.01. The test cycles and test sample numbers calculated in Eq. (7) were 4700 cycles and 18 units, respectively. The ALT was designed to ensure a B1 of 10 years life with about a sixty percent level of condence that it would fail less than once during 5200 cycles. Fig. 6 shows the Kimchi refrigerators in accelerated life testing and an evaporator tubing in the enclosure contained a 0.2 M NaCl water solution. Fig. 7 shows the duty cycles for the corrosive force (F) due to the chlorine concentration.

6. Results and discussion 6.1. Validity of the accelerated life test and failure analysis Fig. 8a and b shows the failed product from the eld and from the accelerated life testing respectively. In the photos in Fig. 8, the shape and location of the failure in the ALT were similar to those seen in the eld. Fig. 9 shows a graphical analysis of the ALT results and eld data on a Weibull plot. These methodologies were valid in pinpointing the weak designs responsible for failures in the eld and were supported by two ndings in the data. The location and shape also, from the Weibull plot, the shape parameters of the ALT, (b1), and market data, (b2), were found to be similar.

6.2. Parameter design with ALTs and corrective action plans The pitting of the evaporator tubing in both the eld products and the ALT test specimens occurred in the inlet/outlet of the evaporator tubing (Fig. 10). The design aw of the cotton adhesive tape resulting in high corrosive stress areas can be corrected by extending the length of the contraction tube from 50 mm to 200 mm and replacing the cotton adhesive tape with a generic transparent tape. The design improvements correspond to the missing key control parameter (KCP) as listed in Table 3.

Fig. 8. Failed products in eld and ALT.

Author's personal copy

376 S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379

Fig. 9. Field data and results of ALT on Weibull chart.

The repetitive corrosive force in combination with the high chlorine concentration of the cotton tape and the crevice between the cotton adhesive tape and the evaporator tubing contained the condensed water as an electrolyte may have been pitting. The parameter design criterion of the newly designed samples was more than the target life of B1 which was 10 years. The BX life of the sample can be calculated as:

Lb x B

n h AFb r1

Fig. 10. Structure of pitting the corrosion tubing in eld.

Author's personal copy

S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379 Table 3 Conrmed key parameters based on the marketplace data and ALTs. CTQ Pitting Parameters KNP KCP N1 C1 C2 Corrosive force (or chlorine concentration) Length of the contraction tube Type of the adhesive tape Unit N (or PPM) 377

Fig. 11. A redesigned evaporator tubing.

Fig. 12. Results of ALT plotted in Weibull chart.

The conrmed values of AF and b in Fig. 8 were 100.0 and 6.41, respectively. The recalculated test cycles and sample size in Eq. (7) were 5300 and 8 EA, respectively. Based on the BX and sample size, two ALTs were performed to obtain the design parameters and their proper levels. In the two ALTs the outlet of the evaporator tubing was pitted in the rst test and was not pitted in the second test.

Author's personal copy

378 Table 4 Results of ALT. First ALT Initial design In 5300 cycles, Corrosion of evaporator pipe is less than 1 1130 1160 1680 1680 cycles: cycles: cycles: cycles: 1/18 pitting 2/18 pitting 4/18 pitting 11/18 OK Second ALT First design iteration 5300 cycles: 8/8 OK S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379

Evaporator pipe structure

Material and spec.

Length of the contraction tube, C1: 50.0 mm ? 200.0 mm Adhesive tape type, C2: cotton type ? generic transparent tape

Fig. 11 shows a redesigned evaporator tubing with high corrosive fatigue strength. Based on the modied design parameters, corrective measures taken to increase the life cycle of the evaporator tubing system included: (1) extending the length of the contraction tube (C1) from 50.0 mm to 200.0 mm; (2) replacing the cotton adhesive tape (C2) with the generic transparent tape. With these modied parameters, the Kimchi refrigerator can reserve the food for a longer period without failure. Fig. 12 and Table 4 show the graphical results of ALT plotted in a Weibull chart and the summary of the results of the ALTs, respectively. Over the course of the two ALTs the B1 life of the samples increased by over 10.0 years. 7. Conclusions Robust methods were used to correct the failed cooling function of the evaporator tubing in a Kimchi refrigerator. Knowledge of the failure modes and mechanism to the pitted evaporator tubing in the eld were applied. The evaporator tubing was improved by changing the design parameters through failure analysis and a series of accelerated life testing. The following general conclusions were obtained: (1) Based on the products that failed both in the eld and in the controlled ALTs, pitting occurred both in the inlet and the outlet of the evaporator tubing. The design parameters of the failed evaporator tubing that were problematic were the cotton adhesive type and the length of the contraction tube. (2) Based on the failure analysis and two ALTs, the values for the design parameters were determined to meet the life cycle requirements. The yearly failure rate and B1 life of the redesigned evaporator tubing, based on the results of ALT, were less than 0.1% and 10 years, respectively. (3) Inspection of the failed product, load analysis, and two rounds of ALT indicated that the parameter design was greatly improved by using the new robust design methodologies for the mechanical evaporator tubing system.

References
[1] Taguchi G. Off-line and on-line quality control systems. In: Proceedings of the international conference on quality control. Tokyo, Japan; 1978. [2] Taguchi G, Shih-Chung T. Introduction to quality engineering: bringing quality engineering upstream. New York: American Society of Mechanical Engineering; 1992. [3] Ashley S. Applying Taguchis quality engineering to technology development. Mech Eng 1992. [4] Wilkins J. Putting Taguchi methods to work to solve design aws. Qual Prog 2000;33(5):559. [5] Phadke M. Quality engineering using robust design. Englewood Cliffs (NJ): Prentice Hall; 1989. [6] Byrne D, Taguchi S. The Taguchi approach to parameter design. Qual Prog 1987;20(12):1926. [7] Box G, Jones S. Designing products that are robust to the environment. Total Qual Manage 1987;3:26582. [8] Vining G, Myers R. Combining Taguchi and response surface philosophies: a dual response approach. J Qual Technol 1990;22:3845.

Author's personal copy

S.-w. Woo et al. / Engineering Failure Analysis 17 (2010) 369379 379

[9] Woo S, Pecht M. Failure analysis and redesign of a helix upper dispenser. Eng Fail Anal 2008;15(4):64253. [10] Woo S, ONeal D, Pecht M. Improving the reliability of a water dispenser lever in a refrigerator subjected to repetitive stresses. Eng Fail Anal 2009;16(5):1597606. [11] Woo S, ONeal D, Pecht M. Design of a hinge kit system in a Kimchi refrigerator receiving repetitive stresses. Eng Fail Anal 2009;16(5):165565. [12] Woo S, Ryu D, Pecht M. Design evaluation of a French refrigerator drawer system subjected to repeated food storage loads. Eng Fail Anal 2009;16(7):222434. [13] Woo S, Pecht M, ONeal D. Reliability design and case study of a refrigerator compressor subjected to repetitive loads. Int J Refrig 2009;32(3):47886. [14] McPherson J. Accelerated testing, packaging, electronic materials handbook. ASM International 1989;1:88794. [15] Ryu D, Chang S. Novel concept for reliability technology. Microelectron Reliab 2005;45(3):61122.

Dr. Woo has a BS and MS in Mechanical Engineering, and he has obtained PhD in Mechanical Engineering from Texas A&M. He major in energy system such as HVAC and its heat transfer, optimal design and control of refrigerator, reliability design of thermal components, and failure Analysis of thermal components in marketplace using the Non-destructive such as SEM & XRAY. In 1992.03 1997 he worked in Agency for Defense Development, Chinhae, South Korea, where he has researcher in charge of Development of Naval weapon System. Now he is working as a Senior Reliability Engineer in Side By Side Refrigerator Division, Digital Appliance, SAMSUNG Electronics, and focus on enhancing the life of refrigerator as using the accelerating life testing. He also has experience about Side-by-Side Refrigerator Design for Best Buy, Lowes, Cabinet-depth Refrigerator Design for General Electrics.

Dr. Dennis L. ONeal has a BS in Nuclear Engineering, Texas A&M University, MS in Mechanical Engineering, Oklahoma State University, and PhD in Mechanical Engineering, Purdue University. Currently he is a Department Head of Mechanical Engineering, Texas A&M University Professional Engineer, an ASHRAE Fellow. He has become the Holdredge/Paul Professor, Department of Mechanical Engineering in 2002. He has an interested in the areas of Heating, ventilating, and air conditioning; frost formation on heat exchangers; heat pump system defrost performance and dynamics; ventilation air heat pumps; and aerosol mixing in ventilation systems. He has been leading an energy system laboratory in Texas A&M and has written a several of heat-transfer related papers in the ASHRAE Transactions.

Dr. Michael Pecht has a BS in Acoustics, an MS in Electrical Engineering and an MS and PhD in Engineering Mechanics from the University of Wisconsin at Madison. He is a Professional Engineer, an IEEE Fellow and an ASME Fellow. He has received the 3M Research Award for electronics packaging, the IEEE Undergraduate Teaching Award, and the IMAPS William D. Ashman Memorial Achievement Award for his contributions in electronics reliability analysis. He has written eighteen books on electronic products development, use and supply chain management. He served as chief editor of the IEEE Transactions on Reliability for eight years and on the advisory board of IEEE Spectrum. He is chief editor for Microelectronics Reliability and an associate editor for the IEEE Transactions on Components and Packaging Technology. He is the founder of CALCE (Center for Advanced Life Cycle Engineering) and the Electronic Products and Systems Consortium at the University of Maryland. He is also a Chair Professor. He has been leading a research team in the area of prognostics for the past ten years, and has now formed a new Electronics Prognostics and Health Management Consortium at the University of Maryland. He has consulted for over 50 major international electronics companies, providing expertise in strategic planning, design, test, prognostics, IP and risk assessment of electronic products and systems.