Search Results for regression analysis

Statistical analysis of mechanical property data is the most reliable method for determination of minimum design allowables. This article describes the general procedures used to determine design allowables. It provides information on the determination of a distribution form. The article presents statistical methods which help in determining design allowables. These methods include direct computation for normal distribution, direct computation for an unknown distribution, computation of derived properties, and regression analysis. The article concludes with information on low- and elevated-temperature design properties.

This article presents typical problems encountered in the analysis of experimental creep and creep-rupture data and the possible solutions to these drawbacks. It provides information on planning the test and creep strain/time relationships. The exponential creep equation and the rational polynomial creep equation are discussed. The article also describes the dependence of stress and temperature on equation parameters and explains the lot-centered regression analysis.

This article presents definitions, equations, and notational conventions that are used widely in statistical analysis. It provides information on a basic framework for the selection of a statistical test. The article outlines the basic elements common to all statistical tests. Illustrations of specific statistical tests and the concept of confidence limits are introduced. The article provides an overview of sampling issues, with guidance on both selecting a sampling method and determining the number of units to sample. One-way analysis of the variance, correlation, and simple regression are also illustrated.

This article discusses some of the statistical aspects of design from an engineer's perspective. It reviews the commonly used statistical terms and distributions for providing some guidance on the practical engineering applications of these distributions. The article describes the basic statistical procedures that can be used to address variability and uncertainty in an engineering analysis. It contains a table that lists the relevant statistics standards published by the American Society of Testing and Materials.

A corrosionist refers to a corrosion engineer, a corrosion technician, a corrosion scientist, a chemist, a physicist, an electrical engineer, a mechanical engineer, a coatings or plastics salesperson, a corrosion consultant, or a plant operator. This article presents an overview of statistical inference and addresses the commonly used statistical tools and tests. It describes the science and engineering of materials, including metals, polymers, and ceramics. The article explores the principles of various surface-sensitive techniques and the usefulness and limitations of these techniques. The techniques are divided into those that provide insight into surface topography and surface structure, and those that provide understanding of chemical nature and identity. The article presents a list of web sites and print media addressing corrosion and related topics in five different areas: societies and associations; corrosion standards, specifications, and recommended practices; sources of corrosion information; corrosion databases and data compilations; and other web resources.

Neural-network (NN) modeling is most suitable for simulations of correlations that are hard to describe or cannot be accurately predicted by physical models. This article describes the principles and procedures of NN modeling. It discusses the use of NN modeling in general organization of software and graphical user interfaces. The article also provides information on the ways to improve and upgrade the NN models.

Measuring the thickness of thin films can be accomplished in many ways. This article focuses on the optical method of single-wavelength ellipsometry, two multiple-wavelength methods of reflectometry and spectroscopic ellipsometry for measuring the thickness of thin films. The general capabilities, principles and applications of ellipsometry and reflectometry are discussed in terms of nondestructive methods.

This article addresses the issue of the implementation of composite damage tolerance requirements as it relates to military aircraft. It presents a brief introduction on the durability impact threat, damage tolerance impact threat, and other damage tolerance damage threats. The article summarizes damage tolerance criteria and durability criteria for military aircraft. It discusses the damage tolerance design philosophy for metallic structures and composite structures of the aircraft. The article describes the implementation of a damage tolerance analysis methodology in terms of the mechanics based model, the regression algorithm, and the semi-empirical analysis.

This article discusses statistical concepts that form the basis of most of the following articles on specific areas of applied statistics and data analysis. It reviews some of the basic concepts that must be understood to successfully apply the statistical procedures, including probability, random variables, degrees of freedom, confidence limits and intervals, and reliability. Descriptive statistics, measures of central tendency, confidence limits and intervals, and degrees of freedom are also discussed.

This article describes methods that are oriented toward the analysis of planned cooperative experiments that involve multiple factors, blocks, and/or treatment levels. The experiments include factorial, fractional factorial, randomized block, and incomplete block experiments.

Hardenability is a composition-dependent property of steel and depends on carbon content and other alloying elements as well as the grain size of the austenite phase. This article provides an overview of a wide range of testing procedures used to determine and quantify hardenability of shallow-hardening, low-carbon, plain carbon, and low-alloy medium-carbon steels ranging from classical fracture and etching, Grossmann hardenability, and Jominy end-quench testing to manual and computerized computational methods. The article then uses this as a backdrop for the implementation of the core concepts of hardenability in a variety of predictive tools for calculating hardenability. The Caterpillar 1E0024 Hardenability Calculator, a personal computer-based program, calculates the Jominy curve based on the steel composition. The article also describes the method for boron and nonboron steels, with calculation examples for 8645 steel and 86B45 steel.

This article details factors that have been used for evaluating the susceptibility of alloys to stress-corrosion cracking. Many considerations impacting the validity and accuracy of information gathered from laboratory testing programs are reviewed. The article highlights the main characteristics of probability distributions, such as normal distribution, log-normal distribution, exponential distribution, Poisson distribution, and extreme-value distribution. It also provides information on the statistical concepts to produce effective test programs.

There are two parts to deal with uncertainty in fatigue design: determining the distributions of possible values for all uncertain inputs and calculating the probability of failure due to all the uncertain inputs. This article discusses the sources of uncertainty in a fatigue analysis, such as the material properties, distribution of applied stress levels within a given environment, environments or loading intensities, and modeling or prediction. It presents a probabilistic approach for analyzing the uncertainties and determining the level of reliability (probability of failure).

Bend tests are conducted to determine the ductility or strength of a material. This article discusses the different bend tests with emphasis on test methods, apparatuses, procedures, specimen preparation, and interpretation and reporting of results. The types of bend tests discussed are bending ductility tests, bending strength tests (ASTM E 855), bend tests as per EN 12384 and JIS 3130, and computer-aided bending tests. The three standard bending strength tests are the cantilever beam bend test, the three-point bend test, and the four-point bend test. European Standard EN 12384 specifies a bend test to determine the modulus of elasticity in bending. Japanese Industrial Standard JIS 3130 specifies two tests to determine the elastic limit of spring plate or strip: the repeated deflection spring test and the moment type spring test.

Neural networks permit the discovery of fundamental relationships and quantitative structure within vast arrays of ill-understood data. This article provides an overview of neural network modeling method, describing its overfitting nature. It discusses the use of neural networks in modeling noise and uncertainties in conducting experiments. The article also presents examples of the application of neural-network modeling to the behavior of metals.

This article reviews the planning of fatigue experiments, including the structure of a test plan, randomization, and nuisance variables. The statistical characterization of the S/N (stress/life) or e/N (strain/life) response of a single material tested under a single condition is discussed. The techniques for defining a mean fatigue curve and evaluating scatter or variability about that mean are explained. The article presents the standard techniques for statistical characterization of the fatigue strength or fatigue limit of a single material by use of the Probit method, the up-and-down (staircase) method, and two-point procedures. Stress-level selection methods are also presented. The article discusses the comparison of the fatigue behavior of two or more materials for data generated at a single stress or strain level. Treatments to compare data generated over a range of stress or strain levels are included. The article also summarizes the consolidation of fatigue data generated at different conditions.