- Email: [email protected]
Effect of muffler mounting bracket designs on durability
Engineering Failure Analysis 18 (2011) 1094–1107
Contents lists available at ScienceDirect
Engineering Failure Analysis journal homepage: www.elsevier.com/locate/engfailanal
Effect of muffler mounting bracket designs on durability Senthilnathan Subbiah, O.P. Singh ⇑, Srikanth K. Mohan, Arockia P. Jeyaraj TVS Motor Company Ltd., Research & Development, Hosur, Tamilnadu 635 109, India
a r t i c l e
i n f o
Article history: Received 16 September 2010 Received in revised form 6 February 2011 Accepted 6 February 2011 Available online 12 February 2011 Keywords: Automotive design Structural failures Welding Finite element analysis Fatigue life
a b s t r a c t Automotive industries perform durability tests on vehicles in the end-user environment to reduce failures and warranty costs in the end-user hands. In this paper we present the failure analysis of muffler mounting brackets of three-wheeler vehicles observed during the durability test. Cracks at the weld location between the engine cradle and brackets were observed in all the vehicles at an average distance of 10,000 km. Many possible causes of the failures are identified using fishbone diagram. The fishbone diagram is a graphical analysis tool that provides a systematic way of looking at effects and the causes that contribute to those effects. Statistical analysis of the failure data was conducted using Weibull distribution for durability life prediction. Further investigations were carried out on the design using finite element method (FEM). A FEM model was developed for the engine cradle assembly in which engine and muffler were modeled as point mass. Vertical forces were applied on the assembly using ‘4g’ criterion, where g is the acceleration due to gravity. The applied force accounts for the high impact forces that act on the structure during durability testing. Results show high magnitude of stresses and strain energy at the weld location. Analysis of the design suggests that bracket was acting as a cantilever beam with one-plane welding mounted on the engine cradle. Modified design, though eliminated the above failure, shifted the failure mode to the bush-bracket region. Various design modifications were carried out and its effect on durability has been discussed. FEM analysis on the final design shows significant reduction in stresses at the critical locations. The new design of the bush-bracket system passed the durability target of 1,00,000 km. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Durability and reliability of a product are equally important from a customer point of view. Reliability is the measure of unanticipated interruptions or unexpected failures during customer use [1]. During a reliability test, one important goal is to maximize the opportunities for observing unexpected failures, so that they can be fixed. Automotive industries are still riddled with significant warranty costs that incur due to premature failure of their products in the customer hands. The key to reducing the design & development expenses and warranty expenses is to subject the product for reliability and/or durability tests for failure modes investigation. The failure investigations of the muffler mounting brackets reported in this paper are observed during the durability tests under actual environment. The target durability of the component was 1,00,000 km whereas the average failures were observed at 10,000 km. One interesting part of this failure investigation is that solution of one failure mode resulted in the shift of failure mode to other part of the same component. The subsequent failure investigations are also presented in the paper.
⇑ Corresponding author. Address: TVS Motor Company Ltd., R&D (Design Analysis Group), P. Box No. 4, Harita, Hosur, Tamilnadu 635 109, India. Tel.: +91 4344 276780x3502; Direct: +91 4344 270502; fax: +91 4344 276649. E-mail addresses: [email protected], [email protected] (O.P. Singh). 1350-6307/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2011.02.009
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1095
Apart from reducing the amount of undesirable noise emitted by the internal combustion (IC) engine, mufflers play a critical role in reducing the emission of harmful gases to the environment. Mufflers are installed along with the exhaust pipe as a part of exhaust system of an IC engine. Depending upon the road conditions a muffler has to bear various magnitudes of dynamic and static forces. The vehicle components have to surmount the high impact forces and stresses induced by potholes & speed bumps and still maintain the structural integrity. Hence, designing mufflers specially its mountings for appropriate in-service durability are a challenge to design engineers. Design of mufflers mounting is important, as its failure leads to the failure of exhaust pipe and excessive vibration of the other components. Failure under investigation deals with the muffler mountings of a two-stroke 200 CC displacement auto rickshaw vehicle. During the durability testing, mufflermounting failures on all vehicles were observed at fillet weld location. Fig. 1a shows the Computer Aided Design (CAD) parts of the muffler mounting system. Both engine and mufflers are mounted on an engine cradle (Fig. 1b). The muffler is mounted on two brackets (1 & 2). The muffler adjoins the each bracket from one side and they are fastened together with plain bushes. These plain bushes are welded to the mounting brackets and these brackets are in turn welded to the engine cradle. Fig. 2 shows the failed mounting bracket 2. The brackets in Fig. 2a and b have failed at 3073 km and 14,402 km respectively. Similar failures were observed on other vehicles at this location. Weld cracks can be seen between the bracket and the cradle. The manufacturing of engine cradle assembly was made at the single supplier source. The material properties, welding parameters and other technical parameters were frequently checked against the specifications and ensured that these were within the limits. Hence, failure due to quality related issues is precluded from the investigation. 1.1. Durability testing background The durability testing consists of city, rural, torture and highway tracks. Each track has a specific range of road profile. The rural and torture track durability are accelerated road testing and the city and highway durability are very similar to the end user road load pattern. The durability driving patterns were arrived based on the customer usage patterns in the field. Each vehicle run consists of the different loading patterns for particular kilometers and in cumulative it is taken as one cycle. These cycles varies based on the type of durability testing. Though this type of testing is really time and energy consuming it simulates the very practical conditions. This type of accelerated testing is done to find the actual durability life of the vehi-
(a)
Engine
Muffler
(b)
Engine mountings holes Engine cradle 2 1
Muffler mounting brackets Resilient mounting Fig. 1. (a) CAD model of the muffler and engine mounting system on the engine cradle and (b) engine cradle showing the design of the two muffler mounting brackets.
1096
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Fig. 2. (a) Failed bracket at 3073 km and (b) at 14,402 km. Cracks at weld location between the engine cradle and bracket can be seen.
cle. It is to be noted that the durability target of 1,00,000 km in accelerated testing may even result in the durability of 10,00,000 km or higher in the customer hands depending upon the usage and driving pattern. The weld cracks were identified through a non-destructive method called dye penetrant inspection (DPI). DPI is based upon capillary action, where low surface tension fluid penetrates into clean and dry surface-breaking discontinuities. Penetrant was applied to the test component by spraying. After adequate penetration time has been allowed and the excess penetrant was removed, a developer was applied. The developer assists to draw penetrant out of the flaw. This makes the defects easily identifiable to the normal human eyes. The main advantage of this type of inspection is that they are portable and can be applied to any complicated surfaces at very low cost [2].
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1097
The paper is organized as follows. In Section 2 we discuss the root cause failure analysis using fishbone diagram and Weibull distribution. Finite element model (FEM) and results are presented in Section 3. Various design modifications and its effect on durability are discussed in Section 4 followed by conclusion in Section 5. 2. Root cause failure analysis Root cause failure analysis (RCFA) is not a single defined methodology; there are various tools, processes and philosophies of RCFA. It is a disciplined process used to investigate, rectify and eliminate equipment failures and it is most effective when directed at chronic breakdowns. There is always one root cause of the failure even though the failure appears to be contributed by various factors initially. To arrive at the logical conclusion at the root cause of the failure, we adopted two methodologies: fishbone diagram and statistical analysis, which are discussed in the following subsections. 2.1. Fishbone diagram Fishbone diagram or cause-effect diagram is a structured way of arriving at a few key sources that contribute most significantly to the problem being investigated [3–5]. These sources are then targeted for improvement. The fishbone not only facilitates design for assembly (DFA), but also helps designers conduct probable design failure modes and effects analysis (FMEA) and implement error proofing (Poka Yoke) measures. The basic concept in the Cause-and-Effect diagram is that the name of a basic problem of interest is entered at the right of the diagram at the end of the main bone. The main possible causes of the failure (the effect) are drawn as bones off the main backbone. Different names can be chosen to suit the problem at hand. Fig. 3 shows the fishbone diagram listing the various causes to the failure. The four main causes of the failures are identified as man, method, material and design. The first three causes viz. man, method and material would result in weld related defects like incomplete penetration, incomplete fusion, undercutting, bad weld design, metallurgical failures etc. [6]. Table 1 shows whether the causes of failure have a strong, weak or no relationship with the actual failure. Analysis of the relationships suggests that one-plane welding (design problem), bracket acting as a cantilever beam (design problem), concentricity between the two bushes (design problem), high thermal stresses due to radiation effect (design problem) and less CO2 weld penetration (man skill problem) has a strong influence on the weld failure. Detailed quality checks were conducted on weld materials and welding processes. Weld quality was as per standards. One-plane welding of bracket to cradle, which resulted
Fig. 3. Fishbone diagram showing the various causes of the bracket mounting failure.
1098
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Table 1 Relationship identification between the causes of failures.
a
Sl. no.
4 M0 s/4P0 s
Cause
Analysisa
1 2 3 4 5 6
Design Design Design Method Design Man
One-plane welding of bracket to cradle Brackets acts cantilever beam Location of right side bracket Concentricity between two bushes High thermal stresses due to radiation effect Less CO2 weld penetration
s 4
: Strong relationship, s: medium relationship, 4: weak/no relationship.
in bracket acting as a cantilever beam (see Fig. 7) would result in high stresses at the weld location. Further, since the muffler is placed in the proximity to the brackets, thermal stresses in the weld can develop due to high temperature. However, since the measured temperature of the weld and brackets were less than 80° C, failure investigation due thermal stresses was not considered. Concentricity between the two bushes is also important to avoid additional stresses induced during operation. However, the two bush axes were in same plane. Fishbone diagram analysis indicates that the design of the bracket could be the reason for the weld failure. In the next section we analyse the failure data using statistical method. 2.2. Statistical analysis The data collected from the field failures are examined statistically to draw insights into the failures and to determine the cause-effect patterns. The most widely used probability distribution in the reliability-engineering disciple is the Weibull distribution [7–9]. The Weibull method works with extremely small samples, even two or three failures. This characteristic is important especially in developmental testing with small samples. According to Nelson and Liu [10,11], if the item consists of many parts and the item falls in the experiment when the weakest part fails, then the Weibull distribution would be an acceptable model of that failure mode. The 3-parameter Weibull failure Probability Density Function (PDF) with time is represented as,
f ðtÞ ¼
b1 b tc
g
g
tc b
eð g Þ ;
ð1Þ
where f ðtÞ P 0, b > 0, g > 0, t P 0 or c, 1 < c < 1; b (beta) is the shape parameters or slope ; g (eta) is the scale parameter representing the characteristics life at which 63.2% of the population can be expected to have failed; and c is a location parameter or the minimum life of the component. Weibull b value is driven by the physics of failure. Weibull g values are measures of durability. The g values can be changed by grades of materials whereas the b values cannot be altered but are dependent upon the physics of failure. Fig. 4 shows the Weibull plot of the failed brackets with a 95% of confidence level. The 3-parameter Weibull plot is used to estimate the failure free operating period i.e. the minimum life of the component (c) and 99.00 90.00
Weibull W3 RR3, F=7 / S=5
3 parameter Weibull plot 2 parameter Weibull plot
% Failure
50.00
10.00 5.00
1.00 100.00
1000.00
10000.00
100000.00
Distance to failure (km) Fig. 4. Weibull probability plot of the failed samples. The parameter values are b = 0.8817, g = 8006.9, c = 2882.8.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1099
it is 2882 km. The 2-parameter plot shows the cumulative distribution function with c = 2882. The b value is 0.8817 (<1.0) signifying premature failure or ‘infant mortality’ of the components. The early failure could be due to high stresses induced in the weld zone and needs further investigation. The value of g is 8006.9 km, which is significantly less than the desired durability target (1,00,000 km). In the following section we analyze the failed design using finite element analysis (FEA).
3. The FEM model and results Finite element analysis (FEA) is one of the most popular engineering analysis methods for structural problems [12,13]. FEA requires a finite element mesh as a geometric input [14,15]. This mesh can be generated directly from a solid model for the detailed part model designed in a three-dimensional (3D) CAD system. Since the detailed solid model (see Fig. 1) is too complex to analyse efficiently, some simplification with an appropriate idealization process including detail removal and dimension reduction in the FE model is needed to reduce the excessive computation time. The engine cradle and the muffler mounting brackets are made of up thin metal sheets. For thin bodies, a different type of meshing approach is required. For sheet metal parts, we extracted and meshed the mid-surface using shell elements instead of 3D solid elements. These structures are represented much more efficiently using shell elements without compromising accuracy, with minimal solution time and less computational resources [16,17]. Fig. 5 shows the FEM model of the existing design. The existing design has 45° weld between the bush and bracket (Fig. 5b). Engine and muffler is represented by a point mass and they are connected to the engine cradle by rigid links. The mass is modeled by ANSYS element mass 21 [18] having rotary inertia. Mass element 21 is a point element having up to 6° of freedom: translations in the nodal x, y, and z directions and rotations about the nodal x, y, and z-axes. Elastic element shell 63 was used to model engine cradle, brackets, bush and the weld. Shell 63 element has both bending and membrane capabilities. Both in-plane and normal loads are permitted. The element has 6° of freedom at each node: translations in the nodal x, y, and z directions and rotations about the nodal x, y, and z-axes. A beam 4 element is used at the bush center. It is a uniaxial element with tension, compression, torsion, and bending capabilities. The element has 6° of freedom at each node. The beam is connected to the bush with rigid elements. A ‘4g’ criterion was used to apply the forces at the muffler location, where g is the acceleration due to gravity. It means that the force applied is four times the force produced by the muffler
(a) Rigid links Engine center of mass Muffler center of mass
Structural BC
(b)
Weld
Bush
45° weld
Fig. 5. (a) Finite element model of the failed muffler mounting system and (b) enlarged view showing weld and other details.
1100
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Fig. 6. (a) Predicted stress distribution using FEA and (b) predicted strain energy distribution. At weld location the magnitude of stress and strain energy are significantly high.
weight. This would induce maximum stresses on the components and it simulates the impact loads that act during the actual testing conditions. Fig. 6a shows the Von Mises stresses induced in the existing design of the brackets. The maximum stress is observed near the weld zone between the engine cradle and the bracket 1. The magnitude of the maximum stress (313 MPa) is too high at the weld location [19]. The bracket 1 is likely to fail first which can trigger the failure of bracket 2. Fig. 6b shows the strain energy contour plot for the same load case. The weld zone where failure was noticed (Fig. 2), absorbs the major part of the energy. This could be the reason of high stresses induced in the weld zone. The strain energy gives a measure of a material’s ability to absorb energy up to fracture. It signifies the ability of the material to absorb and release energy. In the following section, we analyze the existing design and various other design considerations and their effect on durability life.
1101
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
4. Design considerations and effect on durability 4.1. Existing design Fig. 7 shows the line diagram of the existing design. The engine cradle is mounted to the chassis structure through a resilient mount. Engine is mounted vertically on the cradle through the bushes. There are two mountings for mufflers whose sections are rectangular and welded all around in one plane with the cradle as shown in the Section B–B. The two bushes, which facilitate the muffler mountings, are welded to the rectangular section internally by 45° as shown in the area Section A–A. After analysis, we attribute the following factors for high stresses in the at the weld area (Fig. 6). First, the mountings design acts as a cantilever beam welded to the cradle. Second, welding zones between the cradle and the mountings (Section B–B)
Muffler mounting brackets
1
2 Cradle mounted to resilient mounting
Engine mounting
One plane welding 450 weld
Fig. 7. Line diagram of the existing muffler mounting bracket design. Section A–A is the area cross-section and section B–B is the total cross-section.
(a)
(b) After
Before
(c) Horizontal plane welding
(d) Weld
45° weld Engine cradle
Vertical plane welding
Engine cradle
Fig. 8. (a) Old design of the bracket, (b) ‘L’ type new design of the bracket, (c) side view: line diagram for the new ‘L’ type bracket design and (d) rear view.
1102
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
are in one plane and third, the muffler mounting length is too small to absorb bending loads. All these factors would induce higher stresses in the weld zone and subsequently, it can lead to the failure. This design has the durability life of 10,069 km. 4.2. New geometric model concepts 4.2.1. Concept design 1 A new design concept was developed that would facilitates two-plane welding at bracket-engine cradle interface instead of a single-plane welding as seen in the previous design. The rectangular bracket design is changed to the ‘L’ section as shown in the Fig. 8b. The bush is welded to the bracket internally on both sides for 45°. This was the maximum angle welding possible since weld gun was not approachable above 45°. In this design the failure mode of mounting bracket getting crack at the weld location between the cradle and bracket got totally eliminated. However, a new failure mode of bush weld crack on both sides of bracket 1 was noticed and it is shown in Fig. 9. The design has undergone the durability testing and the average life of the samples (bracket 1) tested were 26,580 km. No failure was observed in bracket 2. There was an improvement of 155 % from the previous design. 4.2.2. Concept design 2 From the failure mode of concept design 1 it is clear the weld area is insufficient at the bush locations. Therefore, the bush-bracket coupling becomes weaker not only because the loss of constrain generated by the incompleteness of the welded seam, but, moreover, due to the notch effect of the slot which produces stress concentration in this zone [20]. Design 2 is very similar to design 1 but to increase the weld ability of the bush the bracket was modified. An ‘L’ type relief (cut) was
Fig. 9. (a) The new ‘L’ type design of the bracket failed at new location and (b) enlarged view of the failed component.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1103
provided on the bottom of the bracket to increase the accessibility of the bush welding (Fig. 10). The weld area has now increased to 180° (Fig. 10c). Durability testing was done under the same condition. Failure in the bracket 1 was not observed. However, though interesting, the same failure mode was observed in bracket 2 with a bigger notch (cf. Figs. 9 and 11). It seems that bracket 2 is the next weaker part after bracket 1. FEM analysis (Fig. 6) also suggested that maximum stress in the second bracket was about 170 MPa, which is high. No significant improvement in durability life was observed with respect to the previous design.
4.2.3. Concept design 3 The final bracket design is the combination of the existing design and design 2. The manufacturing process of the cradle produces wrinkles (due to bend) in the inner surfaces of the rectangular cross-section near bracket 2 (see Fig. 12a). Using the same L-type bracket in this area may create a new failure mode of weld crack. In order to avoid it the existing design was used for bracket 2 and L-type design was incorporated for bracket 1. However to eliminate the bush welding crack the design of the bush was modified as shown in Fig. 12b. The existing plain bush was replace with collar bush. With this new collar
(a) Before
(b)
After
(c) Horizontal plane welding
180° welding Vertical plane welding Fig. 10. (a) Old ‘L’ type design of the bracket, (b) new ‘L’ type design with a ‘L’ type relief at the bottom and (c) weld area is increased by 180° in the new design.
Fig. 11. Image showing failed bracket at bush-bracket location during durability test.
1104
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Wrinkles at bend
(a) Before
2
1
Old plane bush
(b) After
New collar bush Fig. 12. (a) CAD model showing old design of the brackets and plane bush and (b) new design of brackets and collar bush.
(a)
Horizontal plane welding
360° welding Vertical plane welding
(b)
(c)
Bracket 1
Bracket 2
Fig. 13. (a) Line diagram of the final design of the bracket. Weld area increased to 360° at bush location, (b) bracket 1 with collar bush and (c) bracket 2 with collar bush.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1105
bush, it can now be welded to the bracket by 360° outside and 180° inside on the other side (Fig. 13a). The final design of the both the brackets with the new bush are shown in Fig. 13b and c. With 360° welds, stress concentration and notch effect is avoided. The increase in the bush length affects the overall assembly of the muffler and its mounting brackets. Minor changes in the location of mounting brackets in the muffler were also incorporated. Finite element analysis was carried out on the final designs of brackets with new bush. Fig. 14a shows the stresses developed at the critical failure zone. The location and magnitude of the maximum stress location has changed compared to the old design (Fig. 6). The magnitude of maximum stress developed in bracket 1 is now 143 MPa compared to 313 MPa in the old design, a reduction of 119% in maximum stress value. Further, the magnitude of strain energy observed by the weld has decreased by fourfold. In addition, the magnitude of maximum stress developed in bracket 2 was also reduced to about 70 MPa, which is about 140% less than the old design.
Fig. 14. (a) Predicted stress distribution using FEA and (b) predicted strain energy distribution. Magnitude of stresses and strain energy has reduced significantly compared to the old design (see Fig. 6).
1106
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Table 2 Values of stress range and life cycles for existing and final design of the brackets.
Bracket Bracket Bracket Bracket
1 2 1 2
(existing design) (existing design) (design concept 3) (design concept 3)
‘4g’ stress (MPa)
‘g’ stress (MPa)
rR (MPa)
N
313.00 92.80 143.00 37.38
32.69 22.20 14.50 7.55
281.31 70.60 128.5 29.83
200 12,504 2074 165,765
5. Discussion It is noticed that different design concepts has resulted in different durability life of the component and interestingly, failure was observed only at the weld locations. The initial crack propagation is likely to start at 45° to the direction of the applied load but would shortly change its course to run perpendicular to it [21]. Cracks can also propagate from the inclusions and discontinuities within the material. The propagation of a fatigue crack is difficult to predict as it depends on many factors including material properties, surface finish etc. The prediction of crack growth propagation is much easier once a crack has grown as material properties and surface finish has insignificant influence (except for Young’s modulus which is stress by strain ratio). Hence, only stress range variation or range, number of cycles, joint geometries and Young’s modulus are considered. The crack will eventually grow to a size when there will be insufficient material holding ensuing in plastic collapse or failure by brittle fracture (e.g. see Fig. 11). Small intrusions generally exist at the toes of a weld join at the interface between the parent materials usually at the base of weld undercut. Sometimes dressing the weld toes by grinding will improve fatigue/durability life, shot peening of the weld (setting outer fiber into compression) will give further improvements, but nevertheless the intrusions will exist in spit of surface quality, direction of loading or post weld heat treatment and will eventually propagate a fatigue crack under the suitable conditions. Accordingly, in the analysis of the fatigue of a welded structure it is assumed that a crack has already been originated and the failure mechanism is simply the propagation of that crack [22]. This ensures fatigue analysis much easier as it reduces the number of variables considerably. This is the reason that the fatigue graph of steel weld can be used to predict the fatigue life of aluminum weld just by making allowance for the difference in Young’s modulus [23]. The key challenge is to reduce the cyclic stress range acting at the critical locations, which will enhance the life of the components. In this paper we investigated the effect of design changes on durability life. Appearance of cracks at the weld locations (see Fig. 2) or complete fracture (see Fig. 11) was considered as failure and therefore, the crack initiation and propagation was not the part of this investigation due to the nature of durability testing conditions. The empirical formula for weld fatigue is given by [22]
rmR N ¼ C
ð2Þ
where rR is stress range, m is the Paris law constant, N is the number of complete cycles to failure or life of the component, and C is a constant. The stress range is the difference between the minimum and maximum nominal stresses at weld location. Maximum stresses were calculated in FEM using ‘4g’ vertical force and minimum forces using ‘g’ (static case) vertical force. For metal (steel and aluminum), m = 3; C = 1.6 1011 can be used as a worst case up to 107 cycles, this value is normally used for fillet welds containing no inherent defects or significant undercut [23]. Spadea and Frank [22] used C = 44 108 for fillet-welds with undercuts. We used this value in this analysis. Table 2 shows the values of stress range and N (cycles to failure). Allowance for peak stresses and stress concentration affects have not been taken into account. The residual stress will never exceed the yield stress of the material; if it does the material will relieve the excess stress by permanent deformation, which is the cause of weld distortion. Following points are noted: first, the stress range for the existing design of the brackets is significantly high resulting in low life cycle. This was also observed in the durability test reported previously. Second, the ratio N existing bracket 1 =N new bracket 1 ¼ 9:38 suggesting that the life of new bracket 1 (design concept 3) has increased by more than nine times when compared to the existing bracket 1. In durability test, the target was to increase the life of the bracket by about 10 times. Third, the ratio N existing bracket 2 =N new bracket 2 ¼ 12:26. This is an interesting observation. The design of the existing bracket 2 and new bracket 2 remains the same (see Fig. 12) except the bush and still, significant increase in life cycle was observed. The increase in life cycle is mainly due to the proper welding facilitated by the use of the collar bush, which in turn reduced the stress range. The new design is expected to increase the life significantly. Same durability test was conducted for this design. The new design passed in durability testing, which has the life of more than 1,00,000 km. This design was also confirmed in other durability vehicles and no failures were observed.
6. Conclusions Failure analysis of the muffler mounting brackets was performed using fishbone diagram, Weibull distribution, FEM analysis and durability tests. The muffler mounting bracket was developing cracks at weld location at an average distance of
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1107
10,000 km during the durability test. Various design concepts were examined. The final design passed the durability target of 1,00,000 km. The analysis results lead to the following conclusions: 1. The failed design of the bracket was such that it was acting as a cantilever beam welded to the engine cradle. Further, the whole welding zone between the bracket and engine cradle was in one plane. These factors lead to the high stresses in the weld zone, which resulted in the premature failure of the component. Such design should be avoided particularly for the welded components subjected to vibration and impact loads. 2. The bracket design was modified in such a way that two-plane welding in bracket-engine cradle zone could be possible. This design modification eliminated the weld failure at bracket-engine cradle junction. However, the failure mode shifted to bush-bracket zone. Analysis of the failure mode revealed that the lack of continuity of the welding seam around the bush-bracket junction resulted in the consequent failure. 3. Different design concepts of the brackets were examined in order to increase the weld seam length between the bushbracket zones. Use of collar bush instead of plane bush increased the weld length to 360°. FEM analysis showed considerable decrease in stresses at the critical locations. No failure was observed in the final design. 4. An important point to be taken from these failure investigations is that various factors such as weld plane, loading conditions, accessibility of the parts for weld gun should be considered while designing components joined together by welding. Further, this research suggests that the design of the components joined by weld play a significant role in durability and weld failures can be avoided by proper design of the components itself.
References [1] Woo Seong-woo, O’Neal Dennis L, Pecht Michael. Reliability design of a reciprocating compressor suction reed valve in a common refrigerator subjected to repetitive pressure loads. Eng Fail Anal 2010;17:979–91. [2] Kohan Anthony Lawrence. Boiler operator’s guide, 4th ed. McGraw-Hill Professional; 1997. [3] Grady R. Practical software metrics for project management and process improvement. Prentice-Hall Inc.; 1992. pp. 37, 79, 129, 130, 137–157. [4] Ishikawa K. A guide to quality control. Tokyo: Asian Productivity Organization; 1976. [5] Brussee W. Statistics for six sigma made easy. Cause-and-effect fishbone diagram. Kindle Book, McGraw Hill Company; 2004 [chapter 5]. [6] Pantazopoulos G, Sampani A. Analysis of a weld failure of a rolled Zn-alloy strip – a case study. Eng Fail Anal 2007;14:642–51. [7] Spindler W. How handy is Weibull? or, the use of failure distributions in the planning and analysis of battery experiments. J Power Sources 1986;17:269–78. [8] Liu Yongshou, He Xindang, Shao Xiaojun, Liu Jun, Yue Zhufeng. Analytical and experimental investigation of fatigue and fracture behaviors for antidouble dog-bone riveted joints. Eng Fail Anal 2010;17:447–1456. doi:10.1016/j.engfailanal.2010.05.006. [9] Shirani M, Härkegård G. Fatigue life distribution and size effect in ductile cast iron for wind turbine components. Eng Fail Anal 2011;18:12–24. doi:10.1016/j.engfailanal.2010.07.001. [10] Nelson W. Applied life data analysis. New York: John Wiley and Sons; 1982. [11] Liu Chi-chao. A comparison between the Weibull and lognormal models used to analyze reliability data. UK: University of Nottingham; 1997. [12] Singh OP, Mohan S, Venkata Mangaraju K, Jayamathy M, Babu R. Thermal seizures in automotive drum brakes. Eng Fail Anal 2010;17:1155–72. doi: org/10.1016/j.engfailanal.2010.02.001. [13] Cao YG, Shi YJ, Zhang SH, Xue SF, Sun XY. Research on strength degradation of oil transmission pipeline by third-party damages. Eng Fail Anal 2010;17:992–1007. doi:10.1016/j.engfailanal.2009.12.002. [14] Harnatkiewicz Piotr, Rusinski Eugeniusz, Moczko Przemyslaw. The lifetime prediction of a rotary screw compressor of a liquid cooler subjected to high pressure and high frequency loads. Eng Fail Anal 2010;17:1290–9. doi: org/10.1016/j.engfailanal.2010.03.003. [15] Hamzah MZ, Purbolaksono J, Inayat-Hussain JI, Nordin NF. Stability analysis of a gas turbine exhaust stack. Eng Fail Anal 2010;17:1466–74. doi: org/ 10.1016/j.engfailanal.2010.05.008. [16] Rezayat Mohsen. Midsurface abstraction from 3D solid models: general theory and applications. Comput Aided Des 1996;28(11):905–15. [17] Martin Tobias, Cohen Elaine. Volumetric parameterization of complex objects by respecting multiple materials. Comput Graph 2010;34:187–97. doi:10.1016/j.cag.2010.03.011. [18] ANSYS Version 10.0. ANSYS, Inc. Release 10.0 documentation. [19] Al-Fadhalah Khaled, Elkholy Ahmed, Majeed Majed. Failure analysis of Grade-80 alloy steel towing chain links. Eng Fail Anal 2010;17:1542–50. doi:10.1016/j.engfailanal.2010.06.005. [20] Toribio J, Kharin V, Ayaso FJ, González B, Matos JC, Vergara D, et al. Failure analysis of a lifting platform for tree pruning. Eng Fail Anal 2010;17:739–47. doi:10.1016/j.engfailanal.2009.08.017. [21] Maddox SJ. Fatigue strength of welded structures. UK: Woodhead Publishing Ltd.; 1991 [ISBN 1 85573 013 8]. [22] Spadea Joseph R, Frank Karl H. Fatigue strength of fillet-welded stiffeners with undercuts. Report no. FHWA/TX-05/0-4178-1, center for transportation research. The University of Texas; 2004. [23] Dynson J. Fatigue of welded joints. [retrieved 05.02.11].
Contents lists available at ScienceDirect
Engineering Failure Analysis journal homepage: www.elsevier.com/locate/engfailanal
Effect of muffler mounting bracket designs on durability Senthilnathan Subbiah, O.P. Singh ⇑, Srikanth K. Mohan, Arockia P. Jeyaraj TVS Motor Company Ltd., Research & Development, Hosur, Tamilnadu 635 109, India
a r t i c l e
i n f o
Article history: Received 16 September 2010 Received in revised form 6 February 2011 Accepted 6 February 2011 Available online 12 February 2011 Keywords: Automotive design Structural failures Welding Finite element analysis Fatigue life
a b s t r a c t Automotive industries perform durability tests on vehicles in the end-user environment to reduce failures and warranty costs in the end-user hands. In this paper we present the failure analysis of muffler mounting brackets of three-wheeler vehicles observed during the durability test. Cracks at the weld location between the engine cradle and brackets were observed in all the vehicles at an average distance of 10,000 km. Many possible causes of the failures are identified using fishbone diagram. The fishbone diagram is a graphical analysis tool that provides a systematic way of looking at effects and the causes that contribute to those effects. Statistical analysis of the failure data was conducted using Weibull distribution for durability life prediction. Further investigations were carried out on the design using finite element method (FEM). A FEM model was developed for the engine cradle assembly in which engine and muffler were modeled as point mass. Vertical forces were applied on the assembly using ‘4g’ criterion, where g is the acceleration due to gravity. The applied force accounts for the high impact forces that act on the structure during durability testing. Results show high magnitude of stresses and strain energy at the weld location. Analysis of the design suggests that bracket was acting as a cantilever beam with one-plane welding mounted on the engine cradle. Modified design, though eliminated the above failure, shifted the failure mode to the bush-bracket region. Various design modifications were carried out and its effect on durability has been discussed. FEM analysis on the final design shows significant reduction in stresses at the critical locations. The new design of the bush-bracket system passed the durability target of 1,00,000 km. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction Durability and reliability of a product are equally important from a customer point of view. Reliability is the measure of unanticipated interruptions or unexpected failures during customer use [1]. During a reliability test, one important goal is to maximize the opportunities for observing unexpected failures, so that they can be fixed. Automotive industries are still riddled with significant warranty costs that incur due to premature failure of their products in the customer hands. The key to reducing the design & development expenses and warranty expenses is to subject the product for reliability and/or durability tests for failure modes investigation. The failure investigations of the muffler mounting brackets reported in this paper are observed during the durability tests under actual environment. The target durability of the component was 1,00,000 km whereas the average failures were observed at 10,000 km. One interesting part of this failure investigation is that solution of one failure mode resulted in the shift of failure mode to other part of the same component. The subsequent failure investigations are also presented in the paper.
⇑ Corresponding author. Address: TVS Motor Company Ltd., R&D (Design Analysis Group), P. Box No. 4, Harita, Hosur, Tamilnadu 635 109, India. Tel.: +91 4344 276780x3502; Direct: +91 4344 270502; fax: +91 4344 276649. E-mail addresses: [email protected], [email protected] (O.P. Singh). 1350-6307/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2011.02.009
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1095
Apart from reducing the amount of undesirable noise emitted by the internal combustion (IC) engine, mufflers play a critical role in reducing the emission of harmful gases to the environment. Mufflers are installed along with the exhaust pipe as a part of exhaust system of an IC engine. Depending upon the road conditions a muffler has to bear various magnitudes of dynamic and static forces. The vehicle components have to surmount the high impact forces and stresses induced by potholes & speed bumps and still maintain the structural integrity. Hence, designing mufflers specially its mountings for appropriate in-service durability are a challenge to design engineers. Design of mufflers mounting is important, as its failure leads to the failure of exhaust pipe and excessive vibration of the other components. Failure under investigation deals with the muffler mountings of a two-stroke 200 CC displacement auto rickshaw vehicle. During the durability testing, mufflermounting failures on all vehicles were observed at fillet weld location. Fig. 1a shows the Computer Aided Design (CAD) parts of the muffler mounting system. Both engine and mufflers are mounted on an engine cradle (Fig. 1b). The muffler is mounted on two brackets (1 & 2). The muffler adjoins the each bracket from one side and they are fastened together with plain bushes. These plain bushes are welded to the mounting brackets and these brackets are in turn welded to the engine cradle. Fig. 2 shows the failed mounting bracket 2. The brackets in Fig. 2a and b have failed at 3073 km and 14,402 km respectively. Similar failures were observed on other vehicles at this location. Weld cracks can be seen between the bracket and the cradle. The manufacturing of engine cradle assembly was made at the single supplier source. The material properties, welding parameters and other technical parameters were frequently checked against the specifications and ensured that these were within the limits. Hence, failure due to quality related issues is precluded from the investigation. 1.1. Durability testing background The durability testing consists of city, rural, torture and highway tracks. Each track has a specific range of road profile. The rural and torture track durability are accelerated road testing and the city and highway durability are very similar to the end user road load pattern. The durability driving patterns were arrived based on the customer usage patterns in the field. Each vehicle run consists of the different loading patterns for particular kilometers and in cumulative it is taken as one cycle. These cycles varies based on the type of durability testing. Though this type of testing is really time and energy consuming it simulates the very practical conditions. This type of accelerated testing is done to find the actual durability life of the vehi-
(a)
Engine
Muffler
(b)
Engine mountings holes Engine cradle 2 1
Muffler mounting brackets Resilient mounting Fig. 1. (a) CAD model of the muffler and engine mounting system on the engine cradle and (b) engine cradle showing the design of the two muffler mounting brackets.
1096
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Fig. 2. (a) Failed bracket at 3073 km and (b) at 14,402 km. Cracks at weld location between the engine cradle and bracket can be seen.
cle. It is to be noted that the durability target of 1,00,000 km in accelerated testing may even result in the durability of 10,00,000 km or higher in the customer hands depending upon the usage and driving pattern. The weld cracks were identified through a non-destructive method called dye penetrant inspection (DPI). DPI is based upon capillary action, where low surface tension fluid penetrates into clean and dry surface-breaking discontinuities. Penetrant was applied to the test component by spraying. After adequate penetration time has been allowed and the excess penetrant was removed, a developer was applied. The developer assists to draw penetrant out of the flaw. This makes the defects easily identifiable to the normal human eyes. The main advantage of this type of inspection is that they are portable and can be applied to any complicated surfaces at very low cost [2].
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1097
The paper is organized as follows. In Section 2 we discuss the root cause failure analysis using fishbone diagram and Weibull distribution. Finite element model (FEM) and results are presented in Section 3. Various design modifications and its effect on durability are discussed in Section 4 followed by conclusion in Section 5. 2. Root cause failure analysis Root cause failure analysis (RCFA) is not a single defined methodology; there are various tools, processes and philosophies of RCFA. It is a disciplined process used to investigate, rectify and eliminate equipment failures and it is most effective when directed at chronic breakdowns. There is always one root cause of the failure even though the failure appears to be contributed by various factors initially. To arrive at the logical conclusion at the root cause of the failure, we adopted two methodologies: fishbone diagram and statistical analysis, which are discussed in the following subsections. 2.1. Fishbone diagram Fishbone diagram or cause-effect diagram is a structured way of arriving at a few key sources that contribute most significantly to the problem being investigated [3–5]. These sources are then targeted for improvement. The fishbone not only facilitates design for assembly (DFA), but also helps designers conduct probable design failure modes and effects analysis (FMEA) and implement error proofing (Poka Yoke) measures. The basic concept in the Cause-and-Effect diagram is that the name of a basic problem of interest is entered at the right of the diagram at the end of the main bone. The main possible causes of the failure (the effect) are drawn as bones off the main backbone. Different names can be chosen to suit the problem at hand. Fig. 3 shows the fishbone diagram listing the various causes to the failure. The four main causes of the failures are identified as man, method, material and design. The first three causes viz. man, method and material would result in weld related defects like incomplete penetration, incomplete fusion, undercutting, bad weld design, metallurgical failures etc. [6]. Table 1 shows whether the causes of failure have a strong, weak or no relationship with the actual failure. Analysis of the relationships suggests that one-plane welding (design problem), bracket acting as a cantilever beam (design problem), concentricity between the two bushes (design problem), high thermal stresses due to radiation effect (design problem) and less CO2 weld penetration (man skill problem) has a strong influence on the weld failure. Detailed quality checks were conducted on weld materials and welding processes. Weld quality was as per standards. One-plane welding of bracket to cradle, which resulted
Fig. 3. Fishbone diagram showing the various causes of the bracket mounting failure.
1098
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Table 1 Relationship identification between the causes of failures.
a
Sl. no.
4 M0 s/4P0 s
Cause
Analysisa
1 2 3 4 5 6
Design Design Design Method Design Man
One-plane welding of bracket to cradle Brackets acts cantilever beam Location of right side bracket Concentricity between two bushes High thermal stresses due to radiation effect Less CO2 weld penetration
s 4
: Strong relationship, s: medium relationship, 4: weak/no relationship.
in bracket acting as a cantilever beam (see Fig. 7) would result in high stresses at the weld location. Further, since the muffler is placed in the proximity to the brackets, thermal stresses in the weld can develop due to high temperature. However, since the measured temperature of the weld and brackets were less than 80° C, failure investigation due thermal stresses was not considered. Concentricity between the two bushes is also important to avoid additional stresses induced during operation. However, the two bush axes were in same plane. Fishbone diagram analysis indicates that the design of the bracket could be the reason for the weld failure. In the next section we analyse the failure data using statistical method. 2.2. Statistical analysis The data collected from the field failures are examined statistically to draw insights into the failures and to determine the cause-effect patterns. The most widely used probability distribution in the reliability-engineering disciple is the Weibull distribution [7–9]. The Weibull method works with extremely small samples, even two or three failures. This characteristic is important especially in developmental testing with small samples. According to Nelson and Liu [10,11], if the item consists of many parts and the item falls in the experiment when the weakest part fails, then the Weibull distribution would be an acceptable model of that failure mode. The 3-parameter Weibull failure Probability Density Function (PDF) with time is represented as,
f ðtÞ ¼
b1 b tc
g
g
tc b
eð g Þ ;
ð1Þ
where f ðtÞ P 0, b > 0, g > 0, t P 0 or c, 1 < c < 1; b (beta) is the shape parameters or slope ; g (eta) is the scale parameter representing the characteristics life at which 63.2% of the population can be expected to have failed; and c is a location parameter or the minimum life of the component. Weibull b value is driven by the physics of failure. Weibull g values are measures of durability. The g values can be changed by grades of materials whereas the b values cannot be altered but are dependent upon the physics of failure. Fig. 4 shows the Weibull plot of the failed brackets with a 95% of confidence level. The 3-parameter Weibull plot is used to estimate the failure free operating period i.e. the minimum life of the component (c) and 99.00 90.00
Weibull W3 RR3, F=7 / S=5
3 parameter Weibull plot 2 parameter Weibull plot
% Failure
50.00
10.00 5.00
1.00 100.00
1000.00
10000.00
100000.00
Distance to failure (km) Fig. 4. Weibull probability plot of the failed samples. The parameter values are b = 0.8817, g = 8006.9, c = 2882.8.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1099
it is 2882 km. The 2-parameter plot shows the cumulative distribution function with c = 2882. The b value is 0.8817 (<1.0) signifying premature failure or ‘infant mortality’ of the components. The early failure could be due to high stresses induced in the weld zone and needs further investigation. The value of g is 8006.9 km, which is significantly less than the desired durability target (1,00,000 km). In the following section we analyze the failed design using finite element analysis (FEA).
3. The FEM model and results Finite element analysis (FEA) is one of the most popular engineering analysis methods for structural problems [12,13]. FEA requires a finite element mesh as a geometric input [14,15]. This mesh can be generated directly from a solid model for the detailed part model designed in a three-dimensional (3D) CAD system. Since the detailed solid model (see Fig. 1) is too complex to analyse efficiently, some simplification with an appropriate idealization process including detail removal and dimension reduction in the FE model is needed to reduce the excessive computation time. The engine cradle and the muffler mounting brackets are made of up thin metal sheets. For thin bodies, a different type of meshing approach is required. For sheet metal parts, we extracted and meshed the mid-surface using shell elements instead of 3D solid elements. These structures are represented much more efficiently using shell elements without compromising accuracy, with minimal solution time and less computational resources [16,17]. Fig. 5 shows the FEM model of the existing design. The existing design has 45° weld between the bush and bracket (Fig. 5b). Engine and muffler is represented by a point mass and they are connected to the engine cradle by rigid links. The mass is modeled by ANSYS element mass 21 [18] having rotary inertia. Mass element 21 is a point element having up to 6° of freedom: translations in the nodal x, y, and z directions and rotations about the nodal x, y, and z-axes. Elastic element shell 63 was used to model engine cradle, brackets, bush and the weld. Shell 63 element has both bending and membrane capabilities. Both in-plane and normal loads are permitted. The element has 6° of freedom at each node: translations in the nodal x, y, and z directions and rotations about the nodal x, y, and z-axes. A beam 4 element is used at the bush center. It is a uniaxial element with tension, compression, torsion, and bending capabilities. The element has 6° of freedom at each node. The beam is connected to the bush with rigid elements. A ‘4g’ criterion was used to apply the forces at the muffler location, where g is the acceleration due to gravity. It means that the force applied is four times the force produced by the muffler
(a) Rigid links Engine center of mass Muffler center of mass
Structural BC
(b)
Weld
Bush
45° weld
Fig. 5. (a) Finite element model of the failed muffler mounting system and (b) enlarged view showing weld and other details.
1100
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Fig. 6. (a) Predicted stress distribution using FEA and (b) predicted strain energy distribution. At weld location the magnitude of stress and strain energy are significantly high.
weight. This would induce maximum stresses on the components and it simulates the impact loads that act during the actual testing conditions. Fig. 6a shows the Von Mises stresses induced in the existing design of the brackets. The maximum stress is observed near the weld zone between the engine cradle and the bracket 1. The magnitude of the maximum stress (313 MPa) is too high at the weld location [19]. The bracket 1 is likely to fail first which can trigger the failure of bracket 2. Fig. 6b shows the strain energy contour plot for the same load case. The weld zone where failure was noticed (Fig. 2), absorbs the major part of the energy. This could be the reason of high stresses induced in the weld zone. The strain energy gives a measure of a material’s ability to absorb energy up to fracture. It signifies the ability of the material to absorb and release energy. In the following section, we analyze the existing design and various other design considerations and their effect on durability life.
1101
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
4. Design considerations and effect on durability 4.1. Existing design Fig. 7 shows the line diagram of the existing design. The engine cradle is mounted to the chassis structure through a resilient mount. Engine is mounted vertically on the cradle through the bushes. There are two mountings for mufflers whose sections are rectangular and welded all around in one plane with the cradle as shown in the Section B–B. The two bushes, which facilitate the muffler mountings, are welded to the rectangular section internally by 45° as shown in the area Section A–A. After analysis, we attribute the following factors for high stresses in the at the weld area (Fig. 6). First, the mountings design acts as a cantilever beam welded to the cradle. Second, welding zones between the cradle and the mountings (Section B–B)
Muffler mounting brackets
1
2 Cradle mounted to resilient mounting
Engine mounting
One plane welding 450 weld
Fig. 7. Line diagram of the existing muffler mounting bracket design. Section A–A is the area cross-section and section B–B is the total cross-section.
(a)
(b) After
Before
(c) Horizontal plane welding
(d) Weld
45° weld Engine cradle
Vertical plane welding
Engine cradle
Fig. 8. (a) Old design of the bracket, (b) ‘L’ type new design of the bracket, (c) side view: line diagram for the new ‘L’ type bracket design and (d) rear view.
1102
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
are in one plane and third, the muffler mounting length is too small to absorb bending loads. All these factors would induce higher stresses in the weld zone and subsequently, it can lead to the failure. This design has the durability life of 10,069 km. 4.2. New geometric model concepts 4.2.1. Concept design 1 A new design concept was developed that would facilitates two-plane welding at bracket-engine cradle interface instead of a single-plane welding as seen in the previous design. The rectangular bracket design is changed to the ‘L’ section as shown in the Fig. 8b. The bush is welded to the bracket internally on both sides for 45°. This was the maximum angle welding possible since weld gun was not approachable above 45°. In this design the failure mode of mounting bracket getting crack at the weld location between the cradle and bracket got totally eliminated. However, a new failure mode of bush weld crack on both sides of bracket 1 was noticed and it is shown in Fig. 9. The design has undergone the durability testing and the average life of the samples (bracket 1) tested were 26,580 km. No failure was observed in bracket 2. There was an improvement of 155 % from the previous design. 4.2.2. Concept design 2 From the failure mode of concept design 1 it is clear the weld area is insufficient at the bush locations. Therefore, the bush-bracket coupling becomes weaker not only because the loss of constrain generated by the incompleteness of the welded seam, but, moreover, due to the notch effect of the slot which produces stress concentration in this zone [20]. Design 2 is very similar to design 1 but to increase the weld ability of the bush the bracket was modified. An ‘L’ type relief (cut) was
Fig. 9. (a) The new ‘L’ type design of the bracket failed at new location and (b) enlarged view of the failed component.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1103
provided on the bottom of the bracket to increase the accessibility of the bush welding (Fig. 10). The weld area has now increased to 180° (Fig. 10c). Durability testing was done under the same condition. Failure in the bracket 1 was not observed. However, though interesting, the same failure mode was observed in bracket 2 with a bigger notch (cf. Figs. 9 and 11). It seems that bracket 2 is the next weaker part after bracket 1. FEM analysis (Fig. 6) also suggested that maximum stress in the second bracket was about 170 MPa, which is high. No significant improvement in durability life was observed with respect to the previous design.
4.2.3. Concept design 3 The final bracket design is the combination of the existing design and design 2. The manufacturing process of the cradle produces wrinkles (due to bend) in the inner surfaces of the rectangular cross-section near bracket 2 (see Fig. 12a). Using the same L-type bracket in this area may create a new failure mode of weld crack. In order to avoid it the existing design was used for bracket 2 and L-type design was incorporated for bracket 1. However to eliminate the bush welding crack the design of the bush was modified as shown in Fig. 12b. The existing plain bush was replace with collar bush. With this new collar
(a) Before
(b)
After
(c) Horizontal plane welding
180° welding Vertical plane welding Fig. 10. (a) Old ‘L’ type design of the bracket, (b) new ‘L’ type design with a ‘L’ type relief at the bottom and (c) weld area is increased by 180° in the new design.
Fig. 11. Image showing failed bracket at bush-bracket location during durability test.
1104
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Wrinkles at bend
(a) Before
2
1
Old plane bush
(b) After
New collar bush Fig. 12. (a) CAD model showing old design of the brackets and plane bush and (b) new design of brackets and collar bush.
(a)
Horizontal plane welding
360° welding Vertical plane welding
(b)
(c)
Bracket 1
Bracket 2
Fig. 13. (a) Line diagram of the final design of the bracket. Weld area increased to 360° at bush location, (b) bracket 1 with collar bush and (c) bracket 2 with collar bush.
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1105
bush, it can now be welded to the bracket by 360° outside and 180° inside on the other side (Fig. 13a). The final design of the both the brackets with the new bush are shown in Fig. 13b and c. With 360° welds, stress concentration and notch effect is avoided. The increase in the bush length affects the overall assembly of the muffler and its mounting brackets. Minor changes in the location of mounting brackets in the muffler were also incorporated. Finite element analysis was carried out on the final designs of brackets with new bush. Fig. 14a shows the stresses developed at the critical failure zone. The location and magnitude of the maximum stress location has changed compared to the old design (Fig. 6). The magnitude of maximum stress developed in bracket 1 is now 143 MPa compared to 313 MPa in the old design, a reduction of 119% in maximum stress value. Further, the magnitude of strain energy observed by the weld has decreased by fourfold. In addition, the magnitude of maximum stress developed in bracket 2 was also reduced to about 70 MPa, which is about 140% less than the old design.
Fig. 14. (a) Predicted stress distribution using FEA and (b) predicted strain energy distribution. Magnitude of stresses and strain energy has reduced significantly compared to the old design (see Fig. 6).
1106
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
Table 2 Values of stress range and life cycles for existing and final design of the brackets.
Bracket Bracket Bracket Bracket
1 2 1 2
(existing design) (existing design) (design concept 3) (design concept 3)
‘4g’ stress (MPa)
‘g’ stress (MPa)
rR (MPa)
N
313.00 92.80 143.00 37.38
32.69 22.20 14.50 7.55
281.31 70.60 128.5 29.83
200 12,504 2074 165,765
5. Discussion It is noticed that different design concepts has resulted in different durability life of the component and interestingly, failure was observed only at the weld locations. The initial crack propagation is likely to start at 45° to the direction of the applied load but would shortly change its course to run perpendicular to it [21]. Cracks can also propagate from the inclusions and discontinuities within the material. The propagation of a fatigue crack is difficult to predict as it depends on many factors including material properties, surface finish etc. The prediction of crack growth propagation is much easier once a crack has grown as material properties and surface finish has insignificant influence (except for Young’s modulus which is stress by strain ratio). Hence, only stress range variation or range, number of cycles, joint geometries and Young’s modulus are considered. The crack will eventually grow to a size when there will be insufficient material holding ensuing in plastic collapse or failure by brittle fracture (e.g. see Fig. 11). Small intrusions generally exist at the toes of a weld join at the interface between the parent materials usually at the base of weld undercut. Sometimes dressing the weld toes by grinding will improve fatigue/durability life, shot peening of the weld (setting outer fiber into compression) will give further improvements, but nevertheless the intrusions will exist in spit of surface quality, direction of loading or post weld heat treatment and will eventually propagate a fatigue crack under the suitable conditions. Accordingly, in the analysis of the fatigue of a welded structure it is assumed that a crack has already been originated and the failure mechanism is simply the propagation of that crack [22]. This ensures fatigue analysis much easier as it reduces the number of variables considerably. This is the reason that the fatigue graph of steel weld can be used to predict the fatigue life of aluminum weld just by making allowance for the difference in Young’s modulus [23]. The key challenge is to reduce the cyclic stress range acting at the critical locations, which will enhance the life of the components. In this paper we investigated the effect of design changes on durability life. Appearance of cracks at the weld locations (see Fig. 2) or complete fracture (see Fig. 11) was considered as failure and therefore, the crack initiation and propagation was not the part of this investigation due to the nature of durability testing conditions. The empirical formula for weld fatigue is given by [22]
rmR N ¼ C
ð2Þ
where rR is stress range, m is the Paris law constant, N is the number of complete cycles to failure or life of the component, and C is a constant. The stress range is the difference between the minimum and maximum nominal stresses at weld location. Maximum stresses were calculated in FEM using ‘4g’ vertical force and minimum forces using ‘g’ (static case) vertical force. For metal (steel and aluminum), m = 3; C = 1.6 1011 can be used as a worst case up to 107 cycles, this value is normally used for fillet welds containing no inherent defects or significant undercut [23]. Spadea and Frank [22] used C = 44 108 for fillet-welds with undercuts. We used this value in this analysis. Table 2 shows the values of stress range and N (cycles to failure). Allowance for peak stresses and stress concentration affects have not been taken into account. The residual stress will never exceed the yield stress of the material; if it does the material will relieve the excess stress by permanent deformation, which is the cause of weld distortion. Following points are noted: first, the stress range for the existing design of the brackets is significantly high resulting in low life cycle. This was also observed in the durability test reported previously. Second, the ratio N existing bracket 1 =N new bracket 1 ¼ 9:38 suggesting that the life of new bracket 1 (design concept 3) has increased by more than nine times when compared to the existing bracket 1. In durability test, the target was to increase the life of the bracket by about 10 times. Third, the ratio N existing bracket 2 =N new bracket 2 ¼ 12:26. This is an interesting observation. The design of the existing bracket 2 and new bracket 2 remains the same (see Fig. 12) except the bush and still, significant increase in life cycle was observed. The increase in life cycle is mainly due to the proper welding facilitated by the use of the collar bush, which in turn reduced the stress range. The new design is expected to increase the life significantly. Same durability test was conducted for this design. The new design passed in durability testing, which has the life of more than 1,00,000 km. This design was also confirmed in other durability vehicles and no failures were observed.
6. Conclusions Failure analysis of the muffler mounting brackets was performed using fishbone diagram, Weibull distribution, FEM analysis and durability tests. The muffler mounting bracket was developing cracks at weld location at an average distance of
S. Subbiah et al. / Engineering Failure Analysis 18 (2011) 1094–1107
1107
10,000 km during the durability test. Various design concepts were examined. The final design passed the durability target of 1,00,000 km. The analysis results lead to the following conclusions: 1. The failed design of the bracket was such that it was acting as a cantilever beam welded to the engine cradle. Further, the whole welding zone between the bracket and engine cradle was in one plane. These factors lead to the high stresses in the weld zone, which resulted in the premature failure of the component. Such design should be avoided particularly for the welded components subjected to vibration and impact loads. 2. The bracket design was modified in such a way that two-plane welding in bracket-engine cradle zone could be possible. This design modification eliminated the weld failure at bracket-engine cradle junction. However, the failure mode shifted to bush-bracket zone. Analysis of the failure mode revealed that the lack of continuity of the welding seam around the bush-bracket junction resulted in the consequent failure. 3. Different design concepts of the brackets were examined in order to increase the weld seam length between the bushbracket zones. Use of collar bush instead of plane bush increased the weld length to 360°. FEM analysis showed considerable decrease in stresses at the critical locations. No failure was observed in the final design. 4. An important point to be taken from these failure investigations is that various factors such as weld plane, loading conditions, accessibility of the parts for weld gun should be considered while designing components joined together by welding. Further, this research suggests that the design of the components joined by weld play a significant role in durability and weld failures can be avoided by proper design of the components itself.
References [1] Woo Seong-woo, O’Neal Dennis L, Pecht Michael. Reliability design of a reciprocating compressor suction reed valve in a common refrigerator subjected to repetitive pressure loads. Eng Fail Anal 2010;17:979–91. [2] Kohan Anthony Lawrence. Boiler operator’s guide, 4th ed. McGraw-Hill Professional; 1997. [3] Grady R. Practical software metrics for project management and process improvement. Prentice-Hall Inc.; 1992. pp. 37, 79, 129, 130, 137–157. [4] Ishikawa K. A guide to quality control. Tokyo: Asian Productivity Organization; 1976. [5] Brussee W. Statistics for six sigma made easy. Cause-and-effect fishbone diagram. Kindle Book, McGraw Hill Company; 2004 [chapter 5]. [6] Pantazopoulos G, Sampani A. Analysis of a weld failure of a rolled Zn-alloy strip – a case study. Eng Fail Anal 2007;14:642–51. [7] Spindler W. How handy is Weibull? or, the use of failure distributions in the planning and analysis of battery experiments. J Power Sources 1986;17:269–78. [8] Liu Yongshou, He Xindang, Shao Xiaojun, Liu Jun, Yue Zhufeng. Analytical and experimental investigation of fatigue and fracture behaviors for antidouble dog-bone riveted joints. Eng Fail Anal 2010;17:447–1456. doi:10.1016/j.engfailanal.2010.05.006. [9] Shirani M, Härkegård G. Fatigue life distribution and size effect in ductile cast iron for wind turbine components. Eng Fail Anal 2011;18:12–24. doi:10.1016/j.engfailanal.2010.07.001. [10] Nelson W. Applied life data analysis. New York: John Wiley and Sons; 1982. [11] Liu Chi-chao. A comparison between the Weibull and lognormal models used to analyze reliability data. UK: University of Nottingham; 1997. [12] Singh OP, Mohan S, Venkata Mangaraju K, Jayamathy M, Babu R. Thermal seizures in automotive drum brakes. Eng Fail Anal 2010;17:1155–72. doi: org/10.1016/j.engfailanal.2010.02.001. [13] Cao YG, Shi YJ, Zhang SH, Xue SF, Sun XY. Research on strength degradation of oil transmission pipeline by third-party damages. Eng Fail Anal 2010;17:992–1007. doi:10.1016/j.engfailanal.2009.12.002. [14] Harnatkiewicz Piotr, Rusinski Eugeniusz, Moczko Przemyslaw. The lifetime prediction of a rotary screw compressor of a liquid cooler subjected to high pressure and high frequency loads. Eng Fail Anal 2010;17:1290–9. doi: org/10.1016/j.engfailanal.2010.03.003. [15] Hamzah MZ, Purbolaksono J, Inayat-Hussain JI, Nordin NF. Stability analysis of a gas turbine exhaust stack. Eng Fail Anal 2010;17:1466–74. doi: org/ 10.1016/j.engfailanal.2010.05.008. [16] Rezayat Mohsen. Midsurface abstraction from 3D solid models: general theory and applications. Comput Aided Des 1996;28(11):905–15. [17] Martin Tobias, Cohen Elaine. Volumetric parameterization of complex objects by respecting multiple materials. Comput Graph 2010;34:187–97. doi:10.1016/j.cag.2010.03.011. [18] ANSYS Version 10.0. ANSYS, Inc. Release 10.0 documentation. [19] Al-Fadhalah Khaled, Elkholy Ahmed, Majeed Majed. Failure analysis of Grade-80 alloy steel towing chain links. Eng Fail Anal 2010;17:1542–50. doi:10.1016/j.engfailanal.2010.06.005. [20] Toribio J, Kharin V, Ayaso FJ, González B, Matos JC, Vergara D, et al. Failure analysis of a lifting platform for tree pruning. Eng Fail Anal 2010;17:739–47. doi:10.1016/j.engfailanal.2009.08.017. [21] Maddox SJ. Fatigue strength of welded structures. UK: Woodhead Publishing Ltd.; 1991 [ISBN 1 85573 013 8]. [22] Spadea Joseph R, Frank Karl H. Fatigue strength of fillet-welded stiffeners with undercuts. Report no. FHWA/TX-05/0-4178-1, center for transportation research. The University of Texas; 2004. [23] Dynson J. Fatigue of welded joints.