5 - NumPy for Engineering Numerical Computation — Topics & Learning Outcomes

Module Topics

Numpy Functions

• Mathematical and Arithmetic Functions — NumPy provides a rich set of universal functions (ufuncs) for performing element-wise mathematical operations on arrays. These include basic arithmetic as well as advanced mathematical computations commonly used in engineering analysis.
• Statistical Functions — NumPy includes a comprehensive suite of statistical functions that allow engineers to analyze datasets, compute descriptive statistics, and understand data distributions. These functions operate directly on arrays and support axis-based computations.
• Linear Algebra Functions — NumPy's numpy.linalg module provides essential linear algebra operations used extensively in structural analysis, control systems, and signal processing. These functions handle matrix operations that are fundamental to solving engineering problems.
• Array Manipulation Functions — NumPy offers a wide range of functions for reshaping, joining, splitting, and transforming arrays, enabling engineers to prepare and restructure data for computational workflows. These functions are essential for adapting data to the required format of algorithms and simulations.
• Sorting and Searching Functions — NumPy provides functions to sort array elements and search for values meeting specific criteria, which are frequently needed in data preprocessing and engineering signal analysis. These functions support multi-dimensional arrays and offer control over sorting algorithms.
• Input/Output and Data Loading Functions — NumPy supports saving and loading array data through a set of I/O functions that interface with binary and text file formats, facilitating data persistence and interoperability in engineering projects. These functions allow engineers to store simulation results and reload experimental datasets efficiently.

Introduction to NumPy and the ndarray

• What Is NumPy? — NumPy is an open-source Python library that provides a high-performance multidimensional array object and a collection of routines for fast mathematical operations on those arrays.
• The ndarray Object — The ndarray (n-dimensional array) is NumPy's primary data structure, storing elements of a single data type in a contiguous block of memory.
• NumPy Arrays vs. Python Lists — While Python lists are flexible containers that can hold mixed types, NumPy arrays are typed, contiguous, and designed for numerical batch operations.
• Key Array Attributes — Understanding an array's metadata attributes is essential before performing any computation or manipulation.

Creating NumPy Arrays

• Creating Arrays from Python Sequences — The most direct way to create a NumPy array is by passing a Python list or tuple to np.array(), which infers the shape and dtype automatically.
• Range and Linspace Functions — np.arange() and np.linspace() generate arrays of evenly spaced values and are essential for creating domain variables in engineering simulations.
• Constant-Value Initialization Functions — NumPy provides convenience functions to create arrays pre-filled with zeros, ones, or arbitrary constants, which are commonly needed to initialize result buffers in iterative engineering algorithms.
• Identity and Diagonal Matrices — For linear algebra workflows, NumPy provides direct constructors for identity and diagonal matrices that appear frequently in engineering mathematics.
• Random Array Generation — NumPy's random module enables generation of arrays populated with random values drawn from various statistical distributions, critical for Monte Carlo simulations and testing.

Element-Wise Operations and Universal Functions

• Vectorized Arithmetic — Arithmetic operators applied to NumPy arrays act element-wise, computing operations across all elements simultaneously without any Python-level loop.
• Universal Functions (ufuncs) — Ufuncs are NumPy functions that operate element-wise on arrays and support additional features such as output array specification and reduction methods.
• Comparison and Boolean Operations — Comparison operators applied to arrays return boolean arrays of the same shape, which can be used directly for filtering and conditional logic.
• Aggregate Reduction Functions — NumPy provides functions that reduce an array along one or more axes, computing values such as sum, product, minimum, and maximum.

Broadcasting

• What Is Broadcasting? — Broadcasting is NumPy's set of rules for stretching arrays with smaller shapes to match larger arrays during arithmetic, enabling operations between arrays of different dimensions.
• Broadcasting Rules — NumPy aligns shapes from the trailing (rightmost) dimension and applies three rules to determine whether two arrays can be broadcast together.
• Common Broadcasting Patterns in Engineering — Broadcasting naturally handles many recurring engineering patterns, such as subtracting a mean vector from every row of a matrix or scaling columns by different factors.
• Avoiding Broadcasting Errors — Mismatched shapes are a frequent source of bugs; understanding shape mismatches and how to inspect and fix them prevents runtime errors.

Indexing, Slicing, and Fancy Indexing

• Basic Integer Indexing — Individual array elements are accessed with integer indices, using zero-based indexing and supporting negative indices to count from the end.
• Slicing Arrays — Slices extract contiguous subarrays using the start:stop:step syntax and return a view into the original array, not a copy.
• Boolean (Mask) Indexing — A boolean array of the same shape as the target can be used as an index to select elements where the mask is True, returning a flattened 1-D array of matching elements.
• Fancy (Integer Array) Indexing — Fancy indexing uses an array of integer indices to select arbitrary elements or rows in any order, enabling non-contiguous and reordered selections.
• np.where and Conditional Selection — np.where() provides a vectorized conditional expression that selects elements from two arrays based on a boolean condition, similar to a ternary operator applied element-wise.

Reshaping, Transposing, and Array Manipulation

• Reshaping Arrays — The reshape() method returns a new view of the array with a different shape, provided the total number of elements remains the same.
• Transposing Arrays — Transposition permutes the axes of an array; for 2-D arrays this swaps rows and columns, and is accessed via the .T attribute or np.transpose().
• Flattening and Raveling — flatten() and ravel() both return 1-D versions of a multidimensional array, but differ in whether they return a copy or a view.
• Stacking and Splitting Arrays — NumPy provides functions to combine multiple arrays along existing or new axes, and to split a single array into multiple sub-arrays.
• Adding and Removing Dimensions — np.newaxis and np.expand_dims() insert size-1 dimensions to adjust array rank for broadcasting or function compatibility.

Linear Algebra with NumPy

• Matrix Multiplication — NumPy supports matrix multiplication via the @ operator and np.matmul(), which perform the standard dot product across rows and columns, distinct from element-wise multiplication.
• Solving Linear Systems — np.linalg.solve(A, b) efficiently solves the linear system Ax = b for x using LU factorization, which is the standard approach for engineering systems of equations.
• Eigenvalues and Eigenvectors — np.linalg.eig() and np.linalg.eigh() compute eigenvalues and eigenvectors of square matrices, used extensively in structural analysis, vibration problems, and principal component analysis.
• Matrix Decompositions — NumPy provides LU (via solve), QR, SVD, and Cholesky decompositions through the np.linalg submodule, enabling factorization-based approaches to engineering computations.
• Norms and Determinants — np.linalg.norm() computes vector and matrix norms used for convergence checks and condition assessment, while np.linalg.det() computes the scalar determinant.

Statistical and Mathematical Functions

• Descriptive Statistics — NumPy provides array-level functions for computing mean, median, variance, and standard deviation along specified axes, forming the basis for engineering data characterization.
• Min, Max, and Sorting — Functions for finding extrema and sorting arrays are fundamental to data analysis, threshold detection, and ranked ordering of engineering results.
• Trigonometric and Exponential Functions — NumPy includes a comprehensive set of element-wise mathematical functions for trigonometry, exponentials, and logarithms, essential for evaluating engineering formulas over arrays.
• Correlation and Covariance — np.corrcoef() and np.cov() compute correlation and covariance matrices between variables, enabling assessment of relationships in multi-channel engineering data.
• Polynomial Evaluation and Fitting — np.polyval() evaluates a polynomial with given coefficients over an array of values, and np.polyfit() fits a least-squares polynomial to data, supporting common engineering curve-fitting tasks.

Student Learning Outcomes

By the end of this module, students will be able to:

Course Outcomes (reference)