
Derive the joint Gaussian distribution of dependent variables Y1 and Y2 as linear combinations of independent x1 and x2, using the Jacobian and determinant, with Gaussian marginals at R=0.
Explore fatigue damage theory using fds and psd to compare non-gaussian and gaussian signals with a research graphical user interface, highlighting duration and frequency effects on damage estimates.
Estimate the power spectral density from a single signal by dividing it into blocks, applying discrete Fourier transform, and averaging periodograms using methods like Welch.
Understand how a signal's Fourier series and its moments yield a Gaussian distribution when phases are uniformly distributed, illustrating the central limit theorem in stochastic processes.
show how to derive the integral (1/π) ∫_0^π cos(ω t + φ)^k dt for even k using Fourier series and complex coefficients, with ω tied to the period.
Explore ergodic signal properties, linking time and ensemble averages through C0 and moments m_k. Derive the sinusoid's probability density from its even moments, using Fourier representations and delta-function techniques.
Derive the distribution of a sinusoidal signal x = a cos(ω t + φ) by a change of variables, revealing the arcsine density p(x) = 1/(π sqrt(a^2 - x^2)) on [-a, a].
This lecture derives a damage formula for a deterministic sine added to gaussian noise, using an integral of peak rate per unit time and the confluent hypergeometric function.
Real time signals force g(ω) to be conjugate symmetric, with |g(ω)| even and φ(ω) odd, yielding a real g(t) via a cosine integral.
Derive Sterling's formula from the factorial integral using x = n y, second-order Taylor expansion near y = 1, and Gaussian approximation to yield n! ≈ n^n e^{-n} sqrt(2π n).
Fatigue Damage, Random Vibrations and Accelerated Life Testing
Fatigue damage is one of those engineering topics where mathematics becomes very concrete.
A component may survive a single load without any visible problem, but fail after thousands, millions, or even billions of cycles. In many real applications, the loading is not a simple sinusoid. It is random, broadband, sometimes non-Gaussian, and often measured from real operating conditions.
This course is about understanding the mathematics behind that process, and how those ideas can be used in vibration testing and reliability engineering.
The material is closely connected to my PhD dissertation, Advanced Mission Synthesis Algorithms for Vibration-based Accelerated Life-testing. The dissertation became publicly available after a two-year embargo, because some of the algorithms developed during the research could not be disclosed immediately.
During that research activity, I also developed graphical user interfaces in collaboration with companies interested in the project. These tools were created to show how the equations can be implemented in practice to generate vibratory signals for accelerated fatigue testing.
The aim of this course is therefore not only to introduce formulas, but to show how mathematical ideas can become engineering tools.
What the Course Is About
The course introduces fatigue-life estimation tests designed to reproduce, in a shorter time, the fatigue damage that a component would experience during its operational life.
This idea is central in accelerated life testing: instead of waiting for the real operating life of a component, we try to design laboratory tests that have the same damage potential as the real environment, but compressed into a shorter duration.
To do this properly, we need mathematics.
We need to understand stochastic processes, probability distributions, random vibrations, single-degree-of-freedom systems, Power Spectral Density, and the statistical behavior of maxima in random processes.
We also need to understand how a measured or target vibration environment can be connected to fatigue damage, and how test signals can be generated in a controlled way.
Main Topics Covered
The course begins with an introduction to fatigue damage and its role in engineering applications.
We then move to stochastic processes and probability distributions, because random vibration testing requires a statistical description of the input signals.
Single-degree-of-freedom systems are introduced as a fundamental model for understanding how mechanical systems respond to vibration.
A key part of the course is the connection between Power Spectral Density and fatigue damage. This is essential for understanding how vibration profiles can be designed and compared in terms of their damaging effect.
The course also discusses the probability density of the maxima of a random process, which is important when studying fatigue under random loading.
From there, we move toward the synthesis of test signals with the same damage potential as real environmental conditions.
Both Gaussian and non-Gaussian signals are discussed. This distinction matters because many real measured signals contain peaks, bursts, or other features that deviate from a purely Gaussian model. In those cases, a more refined approach may be needed to reproduce the relevant fatigue damage correctly.
Course Content
The course includes:
Introduction to fatigue damage.
Stochastic processes and probability distributions.
Random vibration concepts.
Single-degree-of-freedom mechanical systems.
Probability density of the maxima of a random process.
Fatigue-life estimation and accelerated testing.
Tailoring vibration tests to specific applications.
Generating signals with the same damage potential as measured environments.
The relation between Power Spectral Density and fatigue damage.
Gaussian and non-Gaussian signal generation.
Applications to vibration testing and reliability engineering.
Course Approach
This is a mathematical engineering course.
Some parts are theoretical, because the equations matter. Other parts are practical, because the final goal is to understand how those equations can be used to design meaningful tests.
I try to make the mathematics as intuitive as possible, always keeping the connection with real engineering applications visible.
The course is not only about learning fatigue formulas. It is about understanding why those formulas are introduced, how they are connected to random vibrations, and how they can be used in accelerated life testing.
Who This Course Is For
This course is intended for engineering students, mechanical engineers, reliability engineers, researchers, and anyone interested in vibration testing and fatigue damage.
It may be especially useful for learners who want to understand how probability, stochastic processes, vibration theory, and fatigue models come together in real engineering problems.
It is also suitable for students who enjoy seeing advanced mathematics applied to practical mechanical systems.
Course Materials
The course is delivered online and can be followed at your own pace.
In addition to the video lectures, students may also consult reading material related to my PhD dissertation for deeper study.
Some graphical tools developed during the research project are also shown, so that students can see how the mathematical theory can be translated into computational procedures for signal generation and test design.
Time Commitment
The video material can be completed in a relatively compact amount of time, but the course should not be rushed.
Several ideas require reflection, especially the connection between random vibration, spectral descriptions, probability distributions, and fatigue damage.
A reasonable estimate is around 10 to 12 hours of total study time if you want to follow the lectures carefully and think through the main concepts.
Final Note
Fatigue damage is not just a practical engineering problem, and it is not just a mathematical abstraction.
It lies exactly between the two.
This course is meant to help students see that connection: how random vibrations can be described mathematically, how fatigue damage can be estimated, and how accelerated tests can be designed to reproduce the damaging effect of real operating conditions in a shorter time.