Statistical Methods for Verifying Component Reliability

Seminar

About once a year, our department offers a two-day seminar on the topic of  »Statistical Methods for Verifying Component Reliability«. This seminar is held in German.

 

Contents

Special statistical methods are required in order to statistically verify the reliability of components. In a two-day compact seminar, we provide you with the necessary knowledge to be able to answer key questions with confidence:

  • Do you determine the probability of failure against the variable loads of the entire array or do you design against a defined design case?
  • Should the test scenario derived from this include an additional increase compared to the assessment basis?
  • Is it better to carry out many short or few long tests to prove reliability?
  • How to handle censored data correctly?

As a prerequisite for the seminar, basic knowledge of statistics is helpful, but not essential. The mathematical presentation will be reduced to a reasonable level and all topics will be illustrated even better with practical examples.

Location

Fraunhofer ITWM, Fraunhofer-Platz 1, 67663 Kaiserslautern

 

Costs

The participation fee is € 1300 and includes the conference documents as well as lunch and drinks.

 

Next date

The seminar is expected to take place again in May 2025.

If you are interested in an in-house seminar, please do not hesitate to contact us.

Program

The seminar schedule is usually as follows. Changes will be announced on the event page of the current seminar.

  • Basic concepts of descriptive statistics
    • Random variable, mean value, variance
    • Quantiles and median
  • Distribution and density of random variables
    • Important service life distributions
    • Failure rates
  • Stress against strength
    • Design against variable customers
    • Design against defined test scenario
  • Estimation of distributions
    • Probability paper
    • Maximum Likelihood
    • Confidence intervals
  • Statistics and machine learning
    • Unsupervised learning: correlation analysis, cluster analysis, analysis of variance (ANOVA), principal component analysis (PCA)
    • Special features of service life data
      • Dealing with censored data
      • Bootstrap method for small sample sizes
    • Statistical test methods
      • Formulation of hypotheses
      • Consumer and producer risk
    • Planning of service life tests
      • Weibull vs. lognormal
      • Many short or few long attempts?
      • Success runs and Weibull analysis
      • Prior knowledge/Bayes' method
    • Evaluation of Wöhler tests
      • Evaluation of fatigue strength data
      • Efficient Wöhler models

Speaker

  • Dr. Klaus Dreßler,  Division Director »Mathematics for Vehicle Engineering« and Head of Department »Dynamics, Loads, and Environmental Data« 
  • Dr. Sascha Feth, Department »Dynamics, Loads and Environmental Data«, Fraunhofer ITWM
  • Dr. Michael Speckert, Deputy Head of Department »Dynamics, Loads and Environmental Data«, Fraunhofer ITWM
  • Dr. Jochen Fiedler, Department »Dynamics, Loads and Environmental Data«, Fraunhofer ITWM