Question 3 of 6: Artificial Intelligence and Computer-Aided Process Planning

Question 3: Artificial Intelligence and Computer-Aided Process Planning (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

(a) Effective Applications of Artificial Intelligence in Manufacturing

Artificial-intelligence techniques are most valuable in manufacturing wherever a decision would otherwise depend on codified expert judgement or on recognizing patterns in noisy data. Generative process planning uses an expert-system rule base to synthesize an operation sequence, machine selection, and cutting parameters directly from a part's geometric and technological features, replicating the reasoning of an experienced process planner without needing an existing similar plan on file. Diagnosis and troubleshooting of equipment faults is a classic expert-system application: a knowledge base of symptom–cause rules, built from maintenance-technician experience, can guide an operator to the likely fault far faster than trial-and-error. Machine-vision-based quality inspection applies pattern-recognition/AI classifiers to detect surface defects, verify assembly completeness, or read part identification, at speeds and consistency beyond manual inspection. Production scheduling and job-shop dispatching benefit from AI heuristic search (and increasingly machine learning) to find good sequences among a combinatorial explosion of possibilities that classical dispatching rules handle only approximately. Robot vision and path planning use AI to identify and locate randomly oriented parts (bin-picking) and to plan collision-free trajectories. Finally, design-for-manufacture advisory systems can flag features of a CAD model likely to cause manufacturing difficulty, again by encoding expert manufacturing knowledge as rules the system applies automatically.

(b) Variant vs. Generative CAPP: Illustrative Examples

The variant system is desirable when the plant regularly produces parts that fall into well-defined part families with only modest geometric or dimensional variation — for example, a machine shop that produces a family of similar stepped shafts (varying only in diameter, length, and a few keyway/thread details) can group-technology-code each new shaft, retrieve the standard process plan already on file for that family, and have the planner make small edits (feed/speed, a dimension) rather than plan the part from first principles. This is inexpensive to implement (it needs only a classification/coding scheme and a library of standard plans) and works well precisely because the family's process logic rarely changes.

The generative system is desirable where parts are highly varied and do not cluster into pre-existing families — for example, an aerospace or prototype job shop producing one-off or very-low-volume custom brackets and housings, each with a distinct geometry, cannot rely on a library of prior plans and instead needs the system to reason from the part's features (holes, pockets, tolerances, material) directly to a new process plan every time. Generative CAPP costs far more to develop up front (the manufacturing logic must be captured as rules or algorithms covering the full range of features the plant might see) but then scales to arbitrary part variety without needing a matching precedent on file, which is exactly what a low-volume, high-variety shop needs and a variant system cannot provide economically.