This document presents a methodology for probabilistic fatigue modelling and reliability prediction. It discusses:
1) Replacing deterministic fatigue variables with random variables described by statistical distributions.
2) Using the S-N curve approach and linear damage accumulation model to calculate fatigue damage probabilistically.
3) Predicting reliability using stress-strength interference methods, assuming normal distributions for stress and strength.
It then demonstrates the methodology on fatigue data from a Post-Panamax crane, estimating the reliability at three different stress levels.
This document presents a methodology for probabilistic fatigue modelling and reliability prediction. It discusses:
1) Replacing deterministic fatigue variables with random variables described by statistical distributions.
2) Using the S-N curve approach and linear damage accumulation model to calculate fatigue damage probabilistically.
3) Predicting reliability using stress-strength interference methods, assuming normal distributions for stress and strength.
It then demonstrates the methodology on fatigue data from a Post-Panamax crane, estimating the reliability at three different stress levels.
This document presents a methodology for probabilistic fatigue modelling and reliability prediction. It discusses:
1) Replacing deterministic fatigue variables with random variables described by statistical distributions.
2) Using the S-N curve approach and linear damage accumulation model to calculate fatigue damage probabilistically.
3) Predicting reliability using stress-strength interference methods, assuming normal distributions for stress and strength.
It then demonstrates the methodology on fatigue data from a Post-Panamax crane, estimating the reliability at three different stress levels.
This document presents a methodology for probabilistic fatigue modelling and reliability prediction. It discusses:
1) Replacing deterministic fatigue variables with random variables described by statistical distributions.
2) Using the S-N curve approach and linear damage accumulation model to calculate fatigue damage probabilistically.
3) Predicting reliability using stress-strength interference methods, assuming normal distributions for stress and strength.
It then demonstrates the methodology on fatigue data from a Post-Panamax crane, estimating the reliability at three different stress levels.
PROBABILISTIC MODELLING OF FATIGUE FOR RELIABILITY
PREDICTION MINOR RESEARCH
BY: ANUNG GRAHITO 23614301
DEPARTMENT OF AERONAUTICS AND ASTRONAUTICS
FACULTY OF MECHANICAL AND AEROSPACE ENGINEERING
INSTITUT TEKNOLOGI BANDUNG
BANDUNG 2016 I. INTRODUCTION
Fatigue is a failure mode due to high number of cycle load that cause crack to initiate and propagate, the crack then continues to propagate until fracture occurs. Fatigue loading is recognized as one of the most common failure modes in mechanical components. Fatigue failure analysis is an important issue for reliability analysis and structures design in many industries including power generation industry, automotive industry, aerospace industry and many others.
Earlier models of fatigue damage in literature focus on the deterministic nature
of the process, while in reality, its nature is stochastic. Conventional fatigue damage models also tend to be conservative and produce undefinable levels of risk in components. This has led to a need for finding new methods to predict the reliability and remaining life of components. The advantage of the new or improved methodology will enhance the confidence in final product, as well as cost-saving.
The stochastic in fatigue loading may result from couple of things, including manufacturing process, geometry of the component, loading process and more others. In manufacturing process, different manufacturers will have different tolerances in producing the material and thus same material will have slightly different mechanical properties. Different environment also has influence; a controlled laboratory environment will give more precise experimental result compared to laboratory where many distractions are present. Because of these, source of scatter, even under the same experimental condition such as constant amplitude loading, fatigue test may give random number of fatigue life with specific distribution.
This research is trying to provide methodology of modelling probabilistic
damage to predict reliability of a component. The methodology then need to be tested with some examples to prove its validity. II. PROBABILISTIC FATIGUE Probabilistic modelling is based on statistical simulation of a deterministic model. The variables in deterministic model of fatigue calculation are replaced by random variables, which is described by a statistical distribution. There are some important aspects to model probabilistic fatigue analysis. First, the deterministic approach of fatigue modelling, whether it is stress-life approach, strain-life, or fracture mechanics approach. Second, cumulative damage accumulation model. And the third is an appropriate statistical simulation model.
2.1 S-N Approach
S-N curve approach is one of the most classic approach of fatigue modelling. It is based on data of fatigue test (S-N curves) that shows the relation between fatigue life (cycles to failure, N) and cyclic stress amplitude, Sa, and assumes that fatigue damage accumulation is linear. The basic equation that represents the S-N curve is given by
= (1)
where Nf = number of cycles to failure, A is fatigue strength constant and m
represents slope of the S-N curve. Figure 1 below shows probabilistic S-N curve PDFs of fatigue life
S1
= S2 S3
Nf1 Nf2 Nf3 Number of Cycles,
Nf Figure 1 Probabilistic S-N Curve
2.2 Cumulative Damage
One thing that is very important for modelling probabilistic fatigue damage is an accurate damage accumulation model to predict the fatigue damage. Damage accumulation is a phenomenon where variable loading will cause damage in a component, and it gradually accumulates over time until eventually the component fails. One of the most widely used damage accumulation model is the linear damage accumulation model, also known as Palmgren-Miner rule. The equation is as follows:
= (2)
Similarly, for multi-stress levels, damage accumulation can be expressed as:
= =1 = =1 (3)
where D is the total accumulated fatigue damage of the material, Di is the damage accumulated when the component is subjected to stress level S, ni is the number of cycles at stress level S, and Nfi is fatigue life at stress level S. Failure is assumed to happen when the damage accumulation is equals to 1
1 (4)
The number of cycles to failure at any stress level S, can be obtained from S- N curve. A linear relationship between S-N curve model and linear damage accumulation model can be obtained by combining equation (1) and (2)
=
= (5)
where C represents a reciprocal fatigue strength constant, S is constant amplitude
stress level, m represents slope of the S-N curve, and n is number of cycles at stress level S. And for multi-stress levels, the linear damage accumulation models can be expressed as:
= =1 = =1 (6)
2.3 Reliability Prediction
One of the most well-known tools of structural reliability analysis is the stress-strength interference method. In this method, failure is assumed to occur if the stress(load) exceeds the strength of the component. One of the assumption that is used in this research is the stress and strength distribution follows normal distribution. So, the reliability of the components can be stated as:
= 1 ( ) ( 7)
= 1 ( 2 + 2 ) (8)
where D is damage accumulation, is threshold damage, 2 is the mean and is
the variance. When the usage cycle is equal to failure life (n = Nf), then the standard deviation of damage accumulation will be the same with standard deviation of threshold damage accumulation ( = ). Substituting equation (6) to (8):
( =1 ) = 1 ( ) (9) 2 + ( ( 2 =1 / ))
The equation above will be able to provide reliability prediction for a
component, it will assess component reliability by showing the reliability degradation in relation to stress level and number of cycles. III. CASE STUDY A case study is needed to test the proposed model validity. For this research, fatigue data from Post-Panamax crane that is already operating for industry is used. The fatigue data below shows the cycles to failure at a certain stress level.
Using data from the above table, the parameter in S-N curve can be obtained,
(1)(2) = (2)(1) ; m = 3,825 ( 10 )
= ; A = 8,62 x 1013 ( 11 )
1 = ; C = 1,16 x 10-14 ( 12 )
And then three stress level from the fatigue data will be tested and the reliability will be estimated for each of those three stress level. For the first analysis, the stress level 118 Mpa is chosen and the variability of threshold damage is calculated using this formula:
= ( ) ( 13 )
Using the aforementioned parameter, the variability of threshold damage at stress level 118 Mpa is, = 0,142491. After is obtained, then using equation () to estimate reliability at certain stress level for any given time:
(1 =1 ) =1 0,1429412 + =1( ( / ))2 ( )
Figure 2 will show the reliability plot for stress level 118 Mpa
1,00 Reliability Plot 0,95 0,90 Reliability, R
0,85 0,80 0,75 118 Mpa 0,70 0,65 0,60 0,55 0,50 0,E+00 5,E+05 1,E+06 2,E+06 2,E+06 Number of Cycles
Figure 2 Reliability Plot for stress level 108 Mpa
The plot above shows the degradation of reliability with the increase of life cycles. In the beginning when the number of cycles are close to zero, the reliability is almost at 100%, then it continues to decrease until the component reach number of cycle that leads to failure. Reliability estimation is also conducted for stress level 103,8 Mpa and 96,7 Mpa to show the difference of trend between them. Reliability Plot 1,00 0,95 0,90 0,85 Reliability, R
Figure 3 Reliability Plot for Different Stress Level
The figure above shows the difference in degradation of reliability in relation
to stress level. It shows that the higher the stress level is applied to a component; the reliability loss is faster. For industrial purpose, the acceptance of failure probability is less than one percent, figure below shows the probability of failure plot. Probability of Failure Plot 1,E+00
1,E-02 Probability of Failure
1,E-04 118 Mpa 1,E-06 103 Mpa 96 Mpa 1,E-08
1,E-10
1,E-12
Number of Cycles, Mpa
Figure 4 Probability of Failure plot for different stress level
It clearly shows that the higher the stress level is; the probability of failure is also higher. The table below shows the maximum number of cycles when the probability of failure is less than 1 percent. If the number of cycles exceeds these numbers, then the probability of failure will steadily increase until the material/component fails.
For probability of failure < 1%
Stress Amplitude Max Number of (MPa) Cycles 118,0 540.000 103,8 940.000 96,7 1.080.000 IV. CONCLUSION A methodology for modelling probabilistic damage for reliability prediction was provided in this research, which is an extension of earlier deterministic studies carried out by the authors. The methodology then is validated by using case study from fatigue data from Post-Panamax crane that is available. The result of this research are based on assumption that damage accumulation is linear phenomenon. The result of the methodology shows a plot of reliability trend in relation with stress level and number of cycles. Five stress level and cycles to failure data is available in this research, and three of them are used to shows the difference in reliability plot and probability of failure. The plot of reliability for the difference stress level suggests that the higher the stress is applied to a component; the reliability loss is faster. REFERENCES
Chapman & Hall. 2. Fatemi, A. (2011, May 19). eFatigue. Diambil kembali dari https://www.efatigue.com/
3. Lukens, R.W. (1983). PROBABILISTIC FATIGUE LIFE PREDICTIONS OF STRUCTURAL
COMPONENTS IN HIGH-CYCLE FATIGUE REGIMES. California: Naval Postgraduate School 4. Rathod,V.,Yadav,P.O.,Rathore,A.,Jain,R. (2011). PROBABILISTIC MODELING OF FATIGUE DAMAGE ACCUMULATION FOR RELIABIITY PREDICTION. International Journal of Quality, Statistics, and Reliability
