AI for Mechanical Engineers Course

Lesson 1 • History and Evolution of Engineering AI

Traces AI development from rule-based expert systems to modern neural networks. Contextualises current tools within decades of engineering automation progress.

Lesson 2 • What AI Means in Engineering

Defines AI, machine learning, and deep learning with engineering-specific examples. Grounds abstract terminology in familiar mechanical systems and workflows.

Lesson 3 • Data as Engineering Raw Material

Frames data as the foundational input for all AI systems, analogous to material properties in design. Introduces data types, sources, and quality considerations in mechanical contexts.

Lesson 4 • Types of Machine Learning

Covers supervised, unsupervised, and reinforcement learning with mechanical engineering analogies. Enables engineers to select the correct learning paradigm for a given problem.

Lesson 1 • Dataset Splitting and Sampling Strategies

Covers train-validation-test splits, cross-validation, and handling imbalanced datasets. Prevents data leakage and ensures reliable model performance estimates.

Lesson 2 • Data Cleaning and Validation

Teaches detection and correction of missing values, outliers, and sensor drift. Directly impacts model accuracy by ensuring input integrity before training.

Lesson 3 • Engineering Data Sources and Formats

Surveys data origins including sensors, PLCs, CAD files, and simulation logs. Establishes awareness of format diversity engineers encounter before modelling.

Lesson 4 • Data Pipelines and Versioning

Introduces automated pipelines and dataset versioning for reproducible engineering experiments. Connects data preparation to professional MLOps practices introduced later.

Lesson 1 • Clustering for Pattern Discovery

Applies k-means, hierarchical, and density-based clustering to find structure in unlabeled data. Useful for grouping failure modes, operating regimes, and material behaviours.

Lesson 2 • Hyperparameter Tuning Techniques

Covers grid search, random search, and Bayesian optimisation for model configuration. Improves model performance systematically without manual trial-and-error.

Lesson 3 • Model Evaluation and Selection

Establishes rigorous evaluation using cross-validation, confusion matrices, and error metrics. Ensures engineers choose models based on engineering performance criteria, not just accuracy.

Lesson 4 • Classification for Fault Detection

Uses decision trees, SVMs, and ensemble methods to classify system states and faults. Directly applicable to condition monitoring and quality inspection workflows.

Lesson 1 • Convolutional Networks for Engineering Images

Applies CNNs to surface defect detection, weld inspection, and component classification. Connects image processing theory to practical visual quality control applications.

Lesson 2 • Training Neural Networks

Covers backpropagation, gradient descent variants, and loss function selection for engineering tasks. Enables engineers to train stable, converging models on mechanical datasets.

Lesson 3 • Regularisation and Preventing Overfitting

Applies dropout, batch normalisation, and early stopping to improve generalisation. Critical for engineering models trained on limited or noisy industrial datasets.

Lesson 4 • Neural Network Architecture Fundamentals

Explains neurons, layers, activation functions, and forward propagation from first principles. Provides the structural vocabulary needed to design and modify networks.

Lesson 1 • Remaining Useful Life Prediction

Builds regression and LSTM models to estimate time-to-failure from degradation signals. Directly supports maintenance scheduling and spare parts planning.

Lesson 2 • Fault diagnosis and classification

Classifies bearing, gear, and shaft faults using frequency-domain features and ML classifiers. Reduces diagnostic time and dependence on expert technician availability.

Lesson 3 • Condition monitoring fundamentals

Reviews vibration, temperature, and acoustic emission signals used to assess machine health. Establishes the physical basis for data-driven maintenance decisions.

Lesson 4 • Maintenance decision integration

Connects AI predictions to maintenance scheduling, cost modelling, and work order systems. Ensures AI outputs drive actionable decisions rather than remain isolated analytics.

Lesson 1 • Multi-objective optimisation algorithms

Uses genetic algorithms and Pareto-front methods to balance competing design objectives. Enables simultaneous optimisation of weight, stress, cost, and thermal performance.

Lesson 2 • Topology and generative design

Applies AI-assisted topology optimisation to generate lightweight, load-bearing structures. Connects additive manufacturing constraints to AI-generated geometry outputs.

Lesson 3 • Surrogate modelling for simulation

Builds Gaussian process and neural network surrogates to replace expensive FEA and CFD runs. Dramatically reduces computational cost while preserving design space accuracy.

Lesson 4 • AI-augmented CAD and simulation workflows

Integrates AI recommendations into existing CAD and FEA toolchains without full workflow replacement. Bridges AI outputs with engineer-controlled design validation processes.

Lesson 1 • Image acquisition for industrial inspection

Covers camera selection, lighting design, and image preprocessing for consistent industrial imaging. Proper acquisition is the foundation of reliable vision model performance.

Lesson 2 • Vision system deployment and validation

Covers edge deployment, latency optimisation, and statistical validation of vision systems. Ensures production-ready performance under real lighting and throughput conditions.

Lesson 3 • Object detection and localisation

Applies YOLO and Faster R-CNN architectures to locate defects and components in images. Enables real-time detection on production lines with bounding box outputs.

Lesson 4 • 3D vision and point cloud processing

Introduces structured light, LiDAR, and stereo vision for dimensional inspection. Extends 2D vision capabilities to volumetric and geometric quality verification.

Lesson 1 • Model packaging and serving

Packages trained models into APIs and containerised services for integration with plant systems. Bridges the gap between notebook prototypes and production-grade deployments.

Lesson 2 • Model monitoring and drift detection

Tracks prediction accuracy, data drift, and concept drift in live engineering deployments. Prevents silent model degradation that leads to incorrect maintenance or quality decisions.

Lesson 3 • CI/CD pipelines for ML models

Applies continuous integration and delivery principles to automate model testing and deployment. Reduces manual errors and accelerates safe model updates in production.

Lesson 4 • Responsible AI in engineering systems

Addresses safety, fairness, explainability, and accountability requirements for engineering AI. Ensures deployed systems meet professional engineering ethics and regulatory expectations.