For a part to be produced (a CNC milled part, a cast-iron block, a moulded item), which operations does it go through, in which order, on which machine, with which tool and fixture, and in which setup arrangement. Academic name: Computer-Aided Process Planning (CAPP); foundational Alting-Zhang (1989) and Khoshnevis-Chen (1991), modern survey Xu-Wang-Newman (2011).
In plain words
Sound familiar?
- We are a tooling shop running 500-5,000 new part orders per year, or an automotive Tier-1 part supplier; the customer sends a CAD model and the production engineer answers the 'how do we make this' question from intuition + experience + a similar-part card — there is no written, auditable process planning procedure.
- We are an aerospace spare-part manufacturer; every part has a process plan file but the sequencing decision is made on the basis of a 20-30-year-veteran engineer's experience; alternative routings (e.g. milling + drilling in one setup vs two setups) are not evaluated.
- Two engineers in the same shop write two different process plans for the same part; there is no model that numerically compares which is better (total setup time + tool changes + transport).
- We acquired a new CNC machine; the alternative routings have changed but existing parts' process plans were not updated — they still run on the old route, the new machine's capability is not exploited.
- Setup time (tool/fixture change + adjustment + cleaning) is 20-40% of the total operation time; but which parts can have their setups combined is not analysed.
- When the customer asks 'in how many days can you deliver this part?' the price and lead time are estimated 'roughly'; the real process plan + machine schedule comes later and drifts from the quote.
- Our CAD/CAM software does 'feature recognition' (recognises holes, pockets, slots, profiles in the part) but the process plan decision is still made manually; the automatic routing (generative CAPP) module is either absent or not used.
Why it matters
How it's solved
Technical depth
How it's solved
Technical depthIn one sentence: From the part’s CAD model recognise the geometric features (holes, pockets, slots, profiles); list operation alternatives per feature (drill/EDM/mill, etc.); under precedence rules (heat-treatment before precision milling, coating last) let a math model pick which operation, on which machine, in which setup group, in which order — minimising total time + setup time + tool changes.
This problem is studied in the Operations Research (the discipline that uses math and computers to solve business decisions) and manufacturing engineering literature as Computer-Aided Process Planning (CAPP). CAPP is distinct from classical job-shop scheduling (#001): #001 solves the machine sequence of multiple parts under an existing process plan; CAPP produces the process plan itself — which operations, in which order, on which machine, with which tool. Distinct from parallel machine scheduling (#071): #071 distributes similar jobs onto similar parallel machines; CAPP designs the operation chain for a single part. Distinct from cell formation (#057): #057 partitions the physical shop layout into cells; CAPP, under that layout, produces the process plan for each part — it is the bridge from design to make. Three stages:
1. Modelling. Input data: (a) part data set — CAD model (3D solid geometry), part material (aluminium, steel, casting, composite), dimensions, tolerances, surface-quality requirements, heat-treatment requirements (e.g. hardening, stress-relief); from the CAD model feature recognition identifies geometric features (holes, pockets, slots, profiled surfaces, threads), (b) operations library — milling (rough/finish), turning, drilling (manual/CNC), grinding (surface/cylindrical), heat-treatment, coating, assembly — for each operation a machine-type compatibility, tool requirement, typical time formula (a function of cutting parameters), tolerance capability range, (c) machine fleet — for each machine the type (3-axis mill, 5-axis mill, CNC lathe, EDM, grinder, heat-treat furnace), capacity, tool magazine, fixture catalogue, typical setup-time distribution, hourly cost, (d) tool and fixture catalogue — for each tool a type (insert, cutter, drill), size, material compatibility, life, cost; for each fixture compatible part families, setup time, (e) precedence rules — geometric (e.g. pocket before face-mill base), technological (heat-treatment may be required before precision milling, coating is always last), tolerance (tight-tolerance operations after rough), (f) alternative routing table — for each feature a set of operation alternatives (e.g. hole: drill + ream, or drill + boring, or EDM; thin pocket: CNC mill or EDM). Decision variables: (i) the operation chosen for each feature, (ii) the operation sequence (topological order under precedence + setup groups), (iii) machine choice for each operation (from a compatible machine pool), (iv) tool and fixture for each operation, (v) setup groups (operations that can run in one setup). Constraints: feature precedence (geometric + technological + tolerance), machine-operation compatibility, tool-machine compatibility (tool-magazine capacity), fixture compatibility (part positioning), tolerance capability (machine + tool + setup combination must meet the target tolerance). Objective: weighted sum — total operation time + total setup time + tool-change count × cost + transport (inter-machine movement) + quality-risk penalty.
2. Solver-driven decision. Three core approaches: (a) Variant CAPP — from existing part-family templates by similarity; a new part inherits the process plan of the closest matching part family, adapted by small variations; conceptually close to group technology (#057), fast but weak for genuinely new-type parts. (b) Generative CAPP — routing built from scratch; feature recognition + operations library + precedence rules + machine/tool catalogue feed an automatic process plan generator; rule-based / expert systems (classical) or MIP (Mixed-Integer Linear Programming — optimisation with some 0/1 variables and some continuous) / metaheuristics (modern). (c) Hybrid CAPP — variant approach seeds the template, generative approach adapts via an optimisation step; common in modern industrial use. Algorithm choice scales with size: small part (10-30 features, 5-15 operation alternatives) — MIP direct; mid part (30-100 features, many alternatives) — CP (Constraint Programming — a paradigm that lets complex rules be stated directly) or decomposition (first operation selection, then sequencing, then machine/tool); large part (100+ features, heat-treatment + assembly integrated) — metaheuristic (intelligent search methods — genetic algorithm, simulated annealing, tabu search). Alternative process plan evaluation: 3-10 feasible plans are produced for one part, compared on the objective, the field-best is selected — the practitioner takes the first feasible, academic CAPP picks the optimum from alternatives. Setup planning sub-problem: operations that can be done in one setup are grouped (the Group Technology idea applied inside a single part), tool change + fixture change is reduced. Integrated CAPP + JSP: after the process plan decision the shop-level job-shop scheduling (#001) is triggered, with feedback potentially revising the process plan.
3. Field integration. Output is four-layered: (a) process plan file (route card) — the approved operation sequence for the part: which operation, on which machine, with which tool, in which fixture, in which setup, expected time, quality-control points, (b) CAD/CAM hand-off — the process plan is exported to the CAD/CAM software, the CAM program (G-code) is generated operation-by-operation, (c) shop-level job-shop trigger — after the process plan is approved, this part enters the #001 (job-shop scheduling) pool and takes its slot on the machine schedule, (d) cost and lead-time estimate — from the process plan + setup estimate the per-part cost and delivery date are computed numerically and bound to the customer quote. Upstream integration: ERP (part master data, customer orders), PLM platform (CAD/CAM/CAE data management), MES (production tracking), quality management system (tolerance + inspection). Quarterly production-engineering committee: process plan library growth, alternative-routing usage ratio, setup-time trend (actual vs plan), tool-life realisation, scrap + rework rate by part family, new-part intake performance.
Alternatives
Manual + production engineer experience + similar-part card
FreeZero licence
Who it fits: Small shop (50-300 new parts/year), single dominant part family
- + Zero software cost
- + Production engineer experience stays in front
- + A simple similar-part memory is enough at low variety
- − Alternative routings not evaluated (first feasible is picked)
- − Engineer knowledge is lost when they leave
- − Setup grouping intuitive; no multi-operation tool optimisation
- − Process planning takes 2-8 hours for a new-type part
- − $200K-700K/year missed optimisation at 500+ new parts/year
CAD/CAM software with basic routing table
Enterprise$10K-50K licence + $3K-15K/year maintenance (regional SMB market)
Who it fits: Mid-size shop (200-1500 new parts/year), feature recognition in the existing CAD/CAM software
- + CAD geometry + basic feature recognition integrated
- + Standard route-card templates
- + CAM (G-code) generation automatic
- + Local-language UI and support
- − Generative CAPP module typically absent — routing still manual
- − Alternative routing comparison absent
- − Setup planning + multi-operation tool optimisation absent
- − Limited integration with the academic CAPP literature
International PLM platform + CAPP module
Enterprise€800K-4M licence + €200K-1M/year maintenance (large PLM platforms)
Who it fits: Large manufacturer (2000+ new parts/year), multi-site, aerospace / automotive Tier-1
- + Mature CAPP module — variant + generative + hybrid
- + Feature recognition + operations library + tool catalogue integrated
- + Alternative routing comparison + optimum selection
- + Integrated CAPP + JSP (linked to shop scheduling)
- + Part-family classification catalogue (OPITZ-style coding)
- − High licence + long (18-30 month) rollout
- − Customisation to a local manufacturing ecosystem extends the project
- − Tool/fixture catalogue must be rebuilt locally
- − Broad engineering training programme
Open-source solver + academic CAPP reference implementation + consulting
Open SourceLicence free; in-house build 30-50 weeks or $200K-700K consulting
Who it fits: Manufacturer with a tech team, CAD/CAM integration, a one-off 'CAPP transformation' project
- + No licence cost
- + Template-based, generative and hybrid CAPP approaches well-defined for open-source solvers
- + Variant + generative + hybrid CAPP flexibly designed
- + 15+ TR theses (YÖK), reference implementations available
- − In-house OR specialist + production engineer + software team required
- − Academic prototype to production: 12-18 months
- − Feature recognition modules limited in open source (CAD format diversity)
- − Maintenance and version updates in-house
Recommendation
Ask in the meeting
- Which approach does the CAPP module support — template-based (variant), from-scratch (generative, rule-based / MIP), or hybrid? Is there a generative reference for non-rotational (non-prismatic) parts?
- Which formats does feature recognition (CAD-to-features: holes, pockets, slots, profiles) support (STEP, IGES, native CAD)? Which feature types are recognised — only prismatic, or also free-form + complex geometry?
- Is alternative process-plan generation and comparison supported (e.g. 3-5 feasible plans per part, ranked on the objective) or does the system produce a single plan?
- How is the setup-planning sub-problem modelled (grouping operations that can run in one setup, multi-operation tool optimisation) — is the Group Technology idea applied inside a single part?
- How are precedence rules (geometric + technological + tolerance) defined in the system — entered by the user, inferred automatically, or inherited from a part-family template?
- Is an integrated process-planning + shop-scheduling approach supported (process plan decision then triggers shop scheduling, with feedback revising the plan)? Are the process plan and the shop schedule solved in the same model?
- In a pilot with real operational data (12-16 weeks), what savings report can be produced against the previous manual process planning — setup time, total production time, engineer preparation time, alternative-routing usage ratio?
- If the contract ends, in which standard format (STEP, JSON, CSV) can we export the part-family classification catalogue, process plan library, tool/fixture catalogue and CAPP analysis history?
Technical details
Editor’s note
In plain speech this problem is called “work routing”, “production routing” or “how to make the part”. The academic name is Computer-Aided Process Planning (CAPP). Foundational survey: Alting and Zhang (1989) International Journal of Production Research; first generative system for non-rotational parts: Khoshnevis and Chen (1991); modern (2000-2011) survey: Xu, Wang and Newman (2011) International Journal of Computer Integrated Manufacturing; CIRP perspective: ElMaraghy (1993); planning-scheduling integration gap: Shobrys and White (2002).
Distinction from #001 (Job-Shop Scheduling): #001 solves the machine sequence of multiple parts under an existing process plan — each part’s process plan is given as input, the question is “which part on which machine in which order”. CAPP produces the process plan for a single part — the question is “which operations does this part go through, in what order, on which machine”. It is the bridge between design and make. The two are integrated: first CAPP, then JSP — Shobrys and White (2002) quantify the integrated planning-scheduling need.
Distinction from #057 (Group Technology Cell Formation): #057 partitions the physical shop layout into cells — it groups machines, families parts. CAPP, under that layout, produces a process plan for each part. The Group Technology idea is foundational for CAPP’s variant approach: similar parts are templated in the same family and a new part inherits from the closest family template.
Distinction from #071 (Parallel Machine Scheduling): #071 distributes jobs across similar (identical or uniform) parallel machines. CAPP designs the operation chain for a single part (each operation on a different machine type, sequentially). #043 (Assembly Line Balancing) divides work on a fixed serial line; CAPP is the operation-sequencing decision that comes before that.
Most-skipped point in the sector: alternative process plan evaluation. The practitioner production engineer, having inspected the part, produces the first feasible process plan and approves it. But a part typically has 3-10 different feasible operation sequences — examples: can CNC milling + drilling be done in one setup or must they be in two; heat-treatment first or last; pocket machining by rough CNC or by EDM; hole completion by reaming or by boring. If these alternatives are compared on the objective (total setup + tool change + transport + quality-risk), the real optimum emerges — total production time typically drops 15-30% (Khoshnevis-Chen 1991). In the field this comparison is not made because “there is a working plan already”, but an alternative-routing table is worth $150K-500K/year for a 500-2,000 new part/year shop.
Second skipped point: CAPP-JSP decoupling. If the process plan is optimised in isolation, shop-level job-shop scheduling (#001) is still optimal under that plan — but the global optimum is lost. Example: CAPP decides “for this part a 5-axis mill is best”; but in the shop that machine is busy and a 3-axis mill is idle. Under integrated CAPP + JSP (Shobrys-White 2002), the alternative routing (3-axis + an extra setup) is chosen against the actual shop load. Third skipped point: the limit of feature recognition. CAD-to-feature recognition handles holes, pockets, slots and other prismatic features automatically; free-form + complex geometry (turbine blade, mould surface) requires manual definition by the engineer. Knowing this limit matters during pilot scoping — feature recognition should not be over-claimed.
A step-by-step path for an SMB
Stage 1 — Measure first, plan later. At least 12 months of process-planning data: new-part count per month, process-plan preparation time per part (engineer-hours), actual total production time per part (vs estimate), setup time per part, scrap + rework rate. Existing part-family classification (if any) or retrospective extraction (similar geometry + similar material + similar tolerance). Operations library inventory: for each operation a machine-type compatibility, tool requirement, time formula, tolerance capability range. Machine fleet inventory: for each machine the type, capacity, tool magazine, fixture, hourly cost.
Stage 2 — Build the knowledge capital. Part-family templates (variant CAPP foundation): the typical process plan of similar parts, alternative routings, setup groups. Feature library: typical feature types (holes, pockets, slots, profiled surfaces, threads) and the catalogue of alternative operation paths for each. Precedence rules: geometric (pocket before base), technological (heat-treatment before precision milling), tolerance (tight-tolerance after rough). Tool/fixture catalogue: multi-operation tools (one tool can process several features in one setup), fixture-part-family compatibility.
Stage 3 — Pilot. 12-16 weeks. Run variant CAPP template + alternative-routing evaluation + setup optimisation on a single part family (e.g. the 50-100 most-frequently-produced parts). The decision stays with the production engineer; CAPP recommends. Success criterion in writing, beforehand: in the pilot family, total production time -15% minimum, setup time -25% minimum, engineer preparation time -30% minimum.
Stage 4 — Rollout. 12-24 months to extend across the whole part catalogue + integrated CAPP + JSP. New-part intake procedure: which family template is closest, how much adaptation is needed, are new features required. Quarterly production-engineering committee: process plan library growth, alternative-routing usage ratio, setup-time trend (actual vs plan), tool-life realisation, new-part intake performance.
Risks — what can go wrong
Feature recognition error (CAD geometry reading). Geometric feature recognition from a CAD model is limited: prismatic features (holes, pockets, simple slots) are recognised well; free-form + complex geometry (turbine blade, mould surface, organic form) requires manual definition by the engineer. An unrecognised feature is ‘invisible’ in the CAPP module and the process plan is wrong. Fix: scope the pilot to part families where feature recognition is strong, define a manual-override procedure for complex parts.
Tool / fixture mismatch (reality-model gap). A tool the CAPP module shows in the catalogue is physically not available in the shop, or the tool-magazine capacity is insufficient; a fixture does not position the part as expected. The model produces an ‘optimum’ plan that the shop cannot execute. Fix: tool/fixture catalogue synced bidirectionally with MES + machine configuration, monthly physical inventory audit.
Alternative routings not used (intuitive first-feasible). The most critical risk: the CAPP module is installed but the production engineer, out of habit, picks the first feasible; alternative routings are not evaluated. Fix: alternative-routing comparison is a mandatory step, at least 2-3 alternatives are displayed, the objective score is shown; a quarterly KPI report tracks ‘alternative usage ratio’.
Single-supplier CAD/CAM/PLM lock-in. Without a contract clause for “annual standard-format export (STEP, JSON, CSV) of the part-family classification catalogue, process plan library, tool/fixture catalogue and CAPP analysis history”, leaving the system means losing years of production-engineering memory. The PLM platform contract should explicitly cover ownership of the part-family templates + feature library + process plan library.
Solution method — technical view
| Approach | Typical scale | Solve time | Guaranteed optimum? |
|---|---|---|---|
| Manual + production engineer experience | Small shop, <300 parts/year | 1-8 hours/part | No, 40-60% optimum |
| Variant CAPP (part-family template) | Mid, 300-1500 parts/year | minutes/part | No, family-template quality |
| Generative CAPP (rule-based / expert) | Mid, 200-1000 parts/year | minutes/part | No, rule-set dependent |
| Generative CAPP (MIP, small part) | Small part (10-30 features) | minutes-hours | Yes (within bound) |
| Constraint programming (mid part) | Mid part (30-100 features) | hours | Yes (within bound) |
| Metaheuristic (genetic, SA, tabu) — large part | Large part (100+ features) | hours | No, good practical quality |
| Hybrid variant + generative + MIP | Mid-large manufacturer | hours | Good practical + near-optimum |
| Integrated CAPP + JSP (Shobrys-White) | Whole-shop integrated | hours-hours | Good upper-lower bound |
Objective function choice:
- Objective 1 — Minimum total production time: Speed-focused; fits a tight delivery commitment.
- Objective 2 — Minimum total setup time: Machine-capacity-efficient; fits setup-intensive parts.
- Objective 3 — Minimum total cost: Operation time × hourly machine cost + tool-change cost + fixture-change cost + transport.
- Objective 4 — Minimum quality-risk: Tight-tolerance operations after heat-treatment, precision operations after rough, tight-tolerance routed via the most-capable machine + tool combination.
Multi-objective: weighted sum (time + cost + quality-risk) or hierarchical (precedence + compatibility constraints first, then quality-risk, then time + cost).
CAPP variants — pick by the field:
- Variant CAPP (classical): Part-family-template-based, fast but weak for genuinely new-type parts.
- Generative CAPP (Khoshnevis-Chen 1991): Routing from scratch, feature recognition + rule-based / MIP.
- Hybrid CAPP (Xu-Wang-Newman 2011): Variant template + generative optimisation, the modern industrial standard.
- Feature-based CAPP: CAD-feature-recognition-centric, operation-alternative catalogue for each feature.
- Integrated CAPP + JSP (Shobrys-White 2002): Process plan decision + shop scheduling integrated.
- Knowledge-based CAPP: Expert system + academic rules + machine learning for new-part similarity inference.
Academic references
Listed in the page frontmatter under sources. Alting-Zhang (1989) and Khoshnevis-Chen (1991) foundational; ElMaraghy (1993) CIRP perspective; Xu-Wang-Newman (2011) modern survey; Shobrys-White (2002) the planning-scheduling integration need.
Sources
- Alting, L. and Zhang, H.-C. (1989). Computer aided process planning: the state-of-the-art survey. International Journal of Production Research, 27(4), 553-585. Foundational survey of the CAPP field.
- Khoshnevis, B. and Chen, Y.-P. (1991). An automatic process planning system for non-rotational parts. International Journal of Production Research, 29(9), 1851-1873. Reference generative CAPP system for non-rotational parts.
- Xu, X., Wang, L. and Newman, S. T. (2011). Computer-aided process planning — A critical review of recent developments and future trends. International Journal of Computer Integrated Manufacturing, 24(1), 1-31. Modern (2000-2011) CAPP literature survey.
- ElMaraghy, H. A. (1993). Evolution and future perspectives of CAPP. CIRP Annals, 42(2), 739-751. CIRP perspective on the evolution and future of CAPP.
- Shobrys, D. E. and White, D. C. (2002). Planning, scheduling and control systems: why cannot they work together. Computers & Chemical Engineering, 26(2), 149-160. Analysis of the missing planning-scheduling integration.
- YÖK Thesis Center — keyword: ‘süreç planlaması’ or ‘CAPP’ or ‘işlem sırası’ — 15+ theses from TR academia. tez.yok.gov.tr
Glossary
- Process Planning
- Manufacturing OR problem of deciding, for a part to be produced, the optimal sequence of machining/forming/assembly operations and the specific machine, tool and fixture for each operation.
- Setup Reduction
- OR / IE practice of minimising the cumulative setup time (tool change, fixture change, cleaning) across a part's process plan via sequence ordering, multi-operation tools, family setups and SMED internal/external separation.
- MIP
- An optimization model where some decision variables are forced to be whole numbers (e.g. number of trucks, number of shifts).
- Job Shop
- Many distinct jobs flowing through many distinct machines, each job in its own routing.
Related problems
Backward from the Customer Order — Which Raw Material, When, and How Much Should I Order?
For mid-size manufacturing SMBs running 50-500 end items and 200-2,000 raw materials and sub-components (automotive Tier-2, white-goods component shops, furniture and assembly, machinery). With each product's parts list (BOM) and each part's procurement or production lead time on file, the question is: a customer orders 200 units of product X with a fixed delivery week — which raw materials in what quantity must be ordered which week, and which sub-parts must enter assembly when? Material requirements planning (academic name MRP) answers this question backwards from the delivery date and is the core logic underneath every ERP. Done on paper or by 'my supplier always takes 3 weeks' rules of thumb, you either run out and stop the line, or burn cash on excess stock.
Five Suppliers Quote the Same Part — How Much Do I Buy from Each?
You run a manufacturer with 20-200 staff and buy the same raw material or component from 4-8 different suppliers. Each supplier differs on price, quality, lead time, capacity, payment terms, and financial stability; deciding who stays on the approved list and how much of each order goes to whom are two separate calls. A wrong pick can trigger a months-long quality crisis or stop the line when a single supplier fails; the 'cheapest bid wins' approach hides quality, late-delivery, and compliance costs and ends up 15-40% more expensive. Scoring 5-15 bids across 8-12 criteria by hand in a spreadsheet becomes inconsistent fast.