TEACHING
STATISTICS AND MACHINE LEARNING IN CIVIL ENGINEERING
Course objectives:
In the first parts of the course, the purpose is to provide the student with knowledge of probability and statistics, including the nature of random variables, fitting probability distributions to data sets (e.g., loads, material properties, temperature, etc.), and selecting statistical models for engineering analysis. The final parts of the course aim at providing insight into the use of statistical methods for the evaluation of safety and introducing the concepts of machine learning for classification and regression. Thus, the course focuses on statistical methods and the basics of machine learning applied for modelling in civil and architectural engineering.
​
Learning Outcomes:
At the end of the course, the student is expected to be able to:
Describe the concepts of uncertainty and probability
Estimate and interpret simple summary statistics, such as mean, standard deviation, variance, median and quartiles as well as correlation
Apply simple graphical techniques, including histograms, qq-normal plots, and box plots
Identify and describe probability distributions
Apply statistical concepts, such as the formulation of models, parameter estimation, construction of confidence intervals, hypothesis testing and model selection
Apply and interpret simple statistical methods and machine learning within regression and analysis of variance
Describe and apply the frequency analysis of extreme events
Apply and interpret techniques of simulation for design
Apply statistical tools for basic reliability analysis
Assess statistics and machine learning in civil and architectural Engineering.
RISK AND RELIABILITY IN CIVIL ENGINEERING
Course Objective:
This course aims to introduce the basic principles and fundamental techniques of risk and reliability analysis in engineering. After the course, the students will have a refreshed knowledge of probability theory and statistics to model uncertainties in engineering problems. The students will be able to do reliability analysis of engineering problems and to use risk assessment methods for decision making under uncertain conditions.
​
Learning Outcomes:
At the end of the course the student is expected to be able to:
- explain the concepts uncertainty, reliability, safety, and risk
- explain the various methods of component and system reliability and risk analysis
- explain the basic concepts of machine learning and information theory for uncertainty quantification and risk analysis
- model physical, statistical, model, and measurement uncertainties
- model the reliability by FORM/SORM methods surrogate models and simulation
- model system behavior and estimate the reliability of systems
- exemplify Bayesian statistical methods
- exemplify risk and reliability methods for probabilistic design of engineering problems
- evaluate reliability analysis of engineering problems and use risk assessment methods for decision making under uncertain conditions
- evaluate modeling of uncertainties and assessment of reliabilities and risk for engineering problems.
​
Contents:
PART O: OVERVIEW
Lecture O: (i) Introduction to Reliability, Risk and Resilience, (ii) Reliability Methods, Bar Example, (iii) Course Overview
​
PART A: UNCERTAINTY QUANTIFICATION
Lecture A1: (i) Preliminary Data Analysis, (ii) Probability Theory
Lecture A2: (i) Random Variables, (ii) Probability Distributions, (iii) Parameter Estimation
Lecture A3: (i) Model Selection - Maximum Likelihood Estimation, (ii) Probability Plot, (iii) Basics of Information Theory and Akaike Information Criterion (AIC)
Lecture A4: (i) Multiple Random Variables, (ii) Multivariate Gaussian, (iii) uncorrelated, (iv) correlated
​
PART B: STRUCTURAL RELIABILITY
Lecture B1: (i) Structural Reliability, (ii) Crude Monte Carlo Simulation, (iii) MCS bar example
Lecture B2: Reliability Index (i) Cornell, (ii) Hasofer-Lind, (iii) Design point and sensitivity analysis
Lecture B3: First-Order Reliability Method (FORM) (i) Uncorrelated Gaussian, (ii) Correlated Gaussian, (iii) Uncorrelated NonGaussian, (iv) Correlated Gaussian
Lecture B4: System Reliability, (i) Introduction (Fault tress, event trees, bayesian networks), (ii) MCS, FORM for System reliability
​
PART C: RISK ANALYSIS
Lecture C1: (i) Risk-informed Decision Analysis, (ii) Value of Information
Lecture C2: (i) Risk-Based Design and risk-management, (ii) Basics of Random Vibrations
​
HYBRID SYSTEM DESIGN FOR SMART CITY
Summer 2017 - Tsinghua Berkeley Shenzhen Institute