22-Mec-B4 Integrated Manufacturing Systems · May 2017
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Paper format. National Exams, May 2017 — 16-Mec-B4 Integrated Manufacturing Systems. Three hours, open book, any non-communicating calculator permitted. Six questions are printed and any five constitute a complete paper; all questions are of equal value, so each is treated below as a 20-mark question. Every question is solved here, because the set is a study resource rather than a sitting.
Reference texts. M. P. Groover, Automation, Production Systems, and Computer-Integrated Manufacturing, 5th ed. (process planning, CAPP, group technology, discrete control and programmable logic controllers); R. B. Chase and F. R. Jacobs, Operations and Supply Chain Management, 16th ed. (forecasting); S. Nahmias and T. L. Olsen, Production and Operations Analysis, 7th ed. (inventory models and lot sizing); C. E. Ebeling, An Introduction to Reliability and Maintainability Engineering, 3rd ed. (series and parallel reliability); S. Kalpakjian and S. R. Schmid, Manufacturing Engineering and Technology, 8th ed. (machinability data and cutting conditions).
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.
Process planning converts a design into the sequence of operations, machines, tools, fixtures, cutting conditions, times and costs that will actually make the part. Three levels of automation are recognised. Manual planning depends entirely on the planner's experience. Variant (retrieval) planning codes the part with a group-technology classification, finds the part family it belongs to, retrieves the standard plan stored for that family, and edits it. Generative planning synthesises a plan from first principles for each new part, using a formal description of the part's features together with a manufacturing knowledge base and decision logic, with no stored plan to start from. The semi-generative approach is the hybrid that sits between the last two: a family plan supplies the skeleton, and generative decision logic fills in and adapts everything that is genuinely specific to the part in hand, with the human planner in the loop wherever the logic is inconclusive.
In a semi-generative system the part is first coded or its features are extracted from the solid model, and the classification is used to retrieve a standard route for its family — the operation sequence, the datum and setup strategy, the class of machine at each stage. That skeleton is not simply edited by hand. The system then applies generative logic to it: machine and tool selection against the current shop's capability file, cutting speeds and feeds looked up in a machinability database, tolerance and surface-finish checks that add or delete a finishing operation, allowance and stock calculations, standard times from a work-measurement file, and cost roll-up. Where the rules run out — an unfamiliar feature interaction, a tolerance the standard route cannot hold, a machine that is down — the system raises the case to the planner in an interactive dialogue instead of failing silently. The planner's decision is captured and can be promoted back into the family plan or the rule base.
The problems of a fully generative plan are, in practice, why almost no commercially deployed system is purely generative. Automatic recognition of machining features from a solid model is still unreliable when features intersect. The knowledge base required is enormous, plant-specific and perishable: every machine's capability, every fixture, every tolerance capability and every shop rule has to be formalised, and it changes whenever the shop changes. Sequencing is a combinatorial problem once precedence, datum and tolerance-stack constraints are imposed, so search time grows quickly with part complexity. The development effort cannot be justified by a plant with a limited part spectrum. Worst of all, the behaviour is brittle: a part outside the encoded rules yields no plan or a plan that is feasible on paper and wrong on the floor, because the system knows nothing about current machine loading, tooling actually on hand, or operator skill.
The semi-generative approach attacks each of these. The retrieved family plan supplies the sequence, so the combinatorial sequencing search largely disappears and what remains is the well-behaved parametric part of the problem — selection and calculation — which is exactly what decision tables, decision trees and a machinability database do reliably. The knowledge-capture burden falls to the decisions that repeat often enough to be worth formalising, so a useful system can be deployed after coding one or two part families and extended incrementally, giving payback long before a full generative system would have been finished. Human intervention gives graceful degradation: a novel part produces a partial plan and a prompt rather than a failure, and the planner supplies shop-condition knowledge the rule base does not hold. Consistency is still obtained, because the standard route and the cutting data are common to every planner, and every override is recorded and can be fed back.
The practical shape of this in a Canadian machine shop making, say, a family of valve bodies is that the planner enters the part number, the system returns the family route with machines, tools, speeds, feeds and standard times already filled in, and flags the two operations where the drawing tolerance is tighter than the family standard. The planner resolves those two, and the routing sheet, the operation instructions and the cost estimate are issued automatically. That is a fraction of the manual planning time, with none of the development risk of a fully generative system.
