Manufacturing Optimization in Aerospace: What to Optimize First

Disclosure: M365.FM host Mirko Peters works at iFAKT GmbH. This page describes the company factually; paid partnerships and links are labelled as advertising.

By Mirko Peters · 2026-11-12

Manufacturing optimization in aerospace starts with hard constraints, clear goals and explainable plans. A guide for managers choosing optimization software.

Manufacturing optimization in aerospace means making better decisions about what to build, where, when and by whom, under constraints that you can't bend: qualified people, approved processes, the right tooling, certified material and complete records. It isn't mainly about running machines faster. This article is for operations managers who are considering optimization software and want to know what to define first, how the methods work in plain terms and how to test a tool before committing.

Key points

• Manufacturing optimization in aerospace means better plans under hard constraints such as qualified people, approved processes, tooling and traceability, not only faster machines.
• AS9100 builds on ISO 9001 and adds topics such as configuration management, traceability and product safety, so an optimizer must never trade records for speed.
• Pick one primary goal, such as on-time delivery, and treat the others as limits; optimizing utilization alone can delay the urgent order.
• Most scheduling problems are NP-hard, so ask vendors how they find feasible plans quickly and how a planner can see why a plan looks the way it does.
• Test any optimization software by replaying a past month of your own orders and comparing its plan with what actually happened.

What makes aerospace optimization different?

Three things stand out in aerospace compared with a general metalworking shop.

The rules are part of the product. The AS9100 quality standard is published by the International Aerospace Quality Group and is built on ISO 9001. According to ANSI, it adds aerospace-specific requirements such as risk management, product safety, configuration management, counterfeit parts prevention, traceability and supply chain oversight. So an optimizer can't suggest a shortcut that skips a record or an inspection. The plan has to produce the documentation as a by-product.

Some steps can only happen in approved places. Nadcap is an accreditation program for special processes such as heat treating, nondestructive testing, welding, chemical processing and coatings. If an order needs a step that only certain accredited lines or suppliers may perform, that resource is a hard constraint and often the bottleneck. A schedule that ignores it is fiction.

Volume is low and variety is high. Many different part numbers share the same machines and the same few qualified people. That makes the sequence matter more than the speed of any single operation. Waiting time between steps can be a large share of the lead time, so sequencing decisions deserve as much attention as cycle times.

What should you optimize?

Before any software, write down the objective in one sentence, and separate the goal from the limits.

Typical goals in a manufacturing plant of this kind:

• on-time delivery against customer dates,
• shorter total lead time,
• less work in process,
• lower overtime and fewer expedited jobs,
• better use of scarce resources such as a special-process line.

These goals pull against each other. Maximum machine utilization tends to fill every gap, which leaves no room for an urgent order. Short lead time for one customer may push another one late. Choose one primary goal and treat the others as limits you watch, for example "maximize on-time delivery without exceeding the agreed overtime".

Then list the hard constraints, because they decide whether a plan can run at all:

1. Operator qualifications and certifications for each operation.
2. Approved machines, lines and external processors for each process.
3. Tooling, fixtures and gauges, including calibration status.
4. Material availability and the certificates that must come with it.
5. Shift calendars, maintenance windows and holidays.
6. The order of operations, including mandatory inspections and holds.

Anything you don't write down will be broken by the plan, and you'll find out on the floor.

Example: utilization against delivery

This is an invented example, not a measured result. A plant has one heat-treatment furnace. An optimizer set to maximize utilization batches parts from many orders to fill each load. The furnace is busy all week, and three urgent orders wait for their loads. The same furnace with on-time delivery as the primary goal runs smaller or partly filled loads for the urgent parts and shows lower utilization on paper. Which result is better depends on the goal you chose, which is why you must choose it first.

How does optimization work in plain terms?

You don't need the mathematics, but you should know what you're paying for.

Allocating tasks to resources over time, with limits on resources, precedence between steps and setup times, is the core of scheduling. The PyJobShop introduction notes that most scheduling problems are NP-hard, which means finding the provably best answer can be very difficult as the problem grows. In practice, software uses heuristics, metaheuristics or exact methods, including constraint programming, and returns a good feasible plan within a useful time.

What this means for you:

• Feasible first. A plan that respects all hard constraints beats a theoretically better plan that breaks one.
• Speed matters. If a re-plan takes an hour, planners will stop using it when a machine goes down at 2 pm.
• "Optimal" is relative to your model. If the model lacks a constraint, the optimum is wrong. Garbage in, polished garbage out.
• Explanations build trust. A planner should be able to see why a job sits where it does and which constraint blocks it.

This is the territory of finite capacity scheduling, which plans with real resource limits instead of assuming that capacity is unlimited.

What data does an optimizer need?

Optimization is only as good as its inputs. Check these before you talk to vendors:

• routings with realistic process and setup times, not just the standard times in the ERP,
• resources and their calendars,
• a qualification matrix: who may do what,
• tooling and fixture availability,
• open orders with due dates and priorities,
• current stock and expected deliveries of material,
• the status of the work already in process.

Be honest about what you have. If the qualification matrix lives in someone's head, build it first. That work is useful whichever tool you choose.

How do you choose optimization software for an aerospace plant?

Ask questions that reveal how a tool handles your constraints, not how it looks in a demo.

1. Constraints. Can it model operator qualifications, approved resources, tooling and mandatory inspection steps? Ask the vendor to show one of each on your data.
2. Goals. Can you change the primary goal and see the effect? Can the tool compare two plans side by side?
3. Disruptions. What happens when a machine fails, material arrives late or a rush order comes in? How long does a re-plan take and what does the planner see afterwards?
4. Explanations. Can a planner trace why a job is where it is?
5. Overrides. Can planners lock jobs and change priorities by hand, and does the tool respect those decisions in the next run?
6. Records. How does the plan connect to the system that holds your traceability and configuration records? The optimizer shouldn't be a second place where order status is kept.
7. Data connection. How does the tool get ERP and shop floor data, and how often?
8. Effort. How many days of model setup and data cleanup are typical before the first useful plan, and who does the work?

Don't ask for a ranking of products. Write your constraints and your goal, and let candidates prove themselves on them.

How do you test it before you commit?

A replay is more convincing than a demo.

1. Pick one cell or value stream. Not the whole plant.
2. Take one past month of real orders with their actual start times, finish times and disruptions.
3. Record the baseline: on-time delivery, lead time and work in process as they happened.
4. Let the tool plan the same orders from the same starting state, with the same constraints.
5. Compare the plan with what actually happened, and let the planners criticize it. Their objections show the missing constraints.
6. Run it in shadow mode for a few weeks beside the current method before it drives real decisions.

Record the preparation effort as well. A tool that wins on plan quality but needs a specialist for every change may lose in daily use.

How do you bring planners on board?

Involve them from the first day, because they hold the knowledge the model lacks. Let them name the five orders they worried about most last month, then check whether the tool's plan handles those. Give them a way to say "this plan is wrong because..." and turn each answer into a constraint or a data fix. After a few rounds the plan usually stops surprising them, and that's the moment when trust starts. Without this loop, even a good tool gets blamed for every late order that the model never knew about.

Common mistakes

• Optimizing before defining the goal. The software optimizes whatever number you hand it.
• Leaving out soft facts. Qualifications, tooling and approved processes are the constraints that planners know best and models forget first.
• Judging by the demo plant. A clean example proves nothing about your data.
• Treating the optimizer as the record. Order status and traceability data must stay in one controlled place.
• Ignoring the planners. If they can't understand or override the plan, they'll go back to spreadsheets.
• Starting plant-wide. One cell is enough to learn what the tool needs from you.

Where Polaris fits

If you'd like to include a planning optimizer in your test, iFAKT Polaris is one option. iFAKT lists aerospace and aviation among its key industries and describes Polaris as AI plus a digital twin for production. The Polaris Optimizer module turns the product-process-resource model into an optimized production plan based on ERP data, resource availability and constraints such as resources, shifts, zones, skills and capacities. According to iFAKT, it also analyzes the impact of disruptions and re-optimizes. Judge it with the replay above, on your own orders.

Related topics to explore

• Qualification and skills matrices. How to keep them current so a planner can trust them.
• Special-process capacity. How to plan scarce, accredited resources such as furnaces and inspection stations.
• Re-planning after disruptions. Which changes should trigger a new plan and which should wait for the next cycle.
• Digital twins for production planning. What a model of your plant adds to a planning tool and what it costs to maintain.
• Measuring plan adherence. Which few measures show whether the floor follows the plan.

Next step

Manufacturing optimization in aerospace pays off when the constraints and the goal are written down before any tool is chosen. Write your top six hard constraints and one primary goal on a single page, and use them as the test sheet for every vendor conversation.