Why “Custom” Doesn’t Mean “From Scratch Every Time”
Every custom machine builder has been there. A customer walks in with a part nobody on earth has run before, asks for a cycle time that sounds physically aggressive, and wants it on the floor in fourteen weeks. You nod, shake hands, and walk back to a blank SolidWorks screen wondering where the hell to start.
I’ve been on both sides of that desk. Over the last decade I’ve led design packages for everything from a 3-axis gantry that picked 2,000 parts an hour off a stamped reel to a rotary indexing station that did 14 assembly operations on a medical connector. The honest truth is this: a custom machine design process only feels chaotic when you treat every project like a brand-new problem. Once you’ve shipped enough, you realize 70 percent of the decisions are the same ones you made last time. The other 30 percent is where the real engineering happens.
This article lays out the process I actually use in the shop — not the academic version with twelve phase-gates and a PowerPoint for every bolt. It’s the sequence I’d want a new engineer on my team to read on day one, and the one I still scribble on the whiteboard before kickoff.
Phase 1: Understand the Requirement Before You Touch CAD
The single biggest reason custom machines go sideways is not bad mechanics — it’s that nobody wrote down what the machine was actually supposed to do. By week six, the customer is arguing about a feature they assumed was included, and you’re arguing about whether it’s a change order.
The Specification Document That Protects You
Before a single part is modeled, I want a one-to-two page spec locked down. It doesn’t need to be pretty. It needs to answer:
- What is the part? Send a real drawing, not a photo. If the part is still changing, flag it. A machine designed around a moving part is a machine you will rebuild.
- What is the cycle time, measured how? “Fast” is not a number. I need the target takt, the allowed refeed time, and whether the clock starts on operator load or on first motion.
- What is the upstream and downstream interface? Does the infeed come from a bowl feeder, a drawer, a robot? Does outfeed drop into a tray, a bagger, or a customer’s existing conveyor? This defines half your layout.
- What are the operating conditions? Floor space, utility drops (compressed air in bar, 24V, 480V 3-phase), ambient temperature, washdown or IP rating, and whether the machine sits inside a safety fence or on the open shop floor.
- What does success look like in three months? Uptime target, first-pass yield, MTBF — whatever the customer actually cares about on the production floor.
Rule of thumb: If the spec doesn’t have a number for cycle time, part variation, and floor space, you don’t have a spec. You have a conversation.
Visit the Customer’s Floor
I know it costs money and time. Go anyway. The photo in the email never shows the 6-inch step at the doorway, the overhead conduit that will collide with your gantry, or the operator who will load the machine standing on a milk crate. Those surprises become punch-list items in month three. Two hours on site in week one saves you two weeks of rework later.
Phase 2: Concept Layout — The Top-Down Block Model
This is where most engineers jump straight to detailed part modeling and get trapped. Resist the urge.
Start with a top-down assembly skeleton. In SolidWorks, that means a layout sketch in the top-level assembly with the floor footprint, the part travel path, and the envelope for each major station. Everything hangs off that skeleton. When the customer says “can we move the infeed 300 mm to the left?” you change one sketch, not forty parts.
The Layout Questions That Determine Everything
| Decision | What It Drives | When to Lock It |
|---|---|---|
| Inline vs. rotary vs. pallet | Indexing method, floor space, cycle time ceiling | Week 1 concept |
| Number of stations | Frame length, tooling budget, changeover complexity | Week 1 concept |
| Manual load vs. auto feed | Safety guarding, operator ergonomics, cost | Week 1 concept |
| Indexer brand (cam, servo, pneumatic) | Cost, accuracy, noise, lead time | Week 2 |
| Control platform (PLC brand, HMI) | Panel layout, software, spare parts commonality | Week 2 |
The concept model doesn’t need fasteners. It doesn’t need sheet metal. It needs enough geometry that you can stand in front of the customer and say “this is what it looks like, this is how big it is, this is what it costs.” If the customer can’t picture it, they can’t sign off on it — and that sign-off is the gate that keeps you from redesigning the frame after the laser has cut.
Phase 3: Detailed Design — Where the Non-Standard Automation Machine Takes Shape
Once the layout is approved, the work breaks into parallel workstreams. A small team might run these sequentially; a team of three or four should run them in parallel.
Frame and Structure
Start with the weldment. The frame is the longest-lead mechanical item after the control panel. Tube steel (typically 80×80 or 100×100 mm, RHS or SHS) is the default for most custom machines. Cast aluminum extrusion is tempting because it’s fast to assemble, but it flexes under load and gets expensive at machine heights above 1.5 m. I default to welded tube for the main structure and extrusion only for guarding and light sub-frames.
Two things people forget on the frame:
- Levelling and floor mounting. Show the mount, not just the tube. A frame without a foot and a levelling pad becomes a fight at install.
- Cable management from day one. Add the cable tray rack to the weld model, not as an afterthought. The photo of the machine with wires draped like spaghetti is the photo that ends up in the customer’s email to your boss.
Tooling and Fixturing
This is the part that actually touches the customer’s product. Tolerances here matter more than anywhere else in the machine. Rule of thumb: locate the part on two datums, clamp on a third, and never clamp on a datum surface you also use for location. It sounds obvious until you’re debugging a part that shifts 0.05 mm every cycle because the clamp is pushing it off the locator.
Use hardened and ground inserts (typically A2 or D2 tool steel, 58–62 HRC) for surfaces that contact the part in production. Soft 6061 locators look fine on day one and look like Swiss cheese by month six when the part has been punched against them a million times.
Motion and Actuation
Pick cylinders and actuators from your standard catalog, not from the supplier’s wish list. Every extra brand on the bill of materials is an extra spare part on the shelf, an extra valve manifold, an extra part number the purchasing team has to source, and an extra component the electrician has never wired before. I keep a short list — one cylinder brand, one linear guide brand, one ball screw brand — and I only deviate when the application actually demands it.
Phase 4: Verification Before You Cut Anything
This is the phase that separates shops that ship on time from shops that ship angry. Three checks, in order:
Interference Check (Digital)
Run a full collision detection on the assembly. Set the tolerance to 1 mm, not zero. Then walk the mechanism through its full motion envelope — every station, every open/close cycle, every clamp extension. A static assembly doesn’t show you what happens when the gripper opens while the feeder is indexing.
Design Review (Human)
A design review with two engineers who didn’t draw the thing catches things the drawer is blind to. I run a simple checklist: tooling access for the operator, service access for maintenance, pinch points, fastener accessibility, and whether the part can actually be loaded and unloaded without an arm extension.
Supplier Quote Review (Commercial)
Send the BOM out before you release the drawings. The 3-week lead time on that servo drive you picked becomes 8 weeks when the supplier’s warehouse is out. Catching that in week four beats catching it in week eight, when the welder is already asking for the next project.
Phase 5: Build, Debug, and Install
The design phase is over when the drawings go to the shop. But the engineer isn’t done. The first time the machine runs under its own power, you will find things the CAD model didn’t tell you about — a sensor that sees the wrong surface, a gripper that fights the part eject, an air line that kinks at full stroke.
Stay on the floor. The engineers who disappear back to the office after release are the ones who get called back for every revision. The ones who stand next to the technician during debug are the ones whose next machine has the lessons already designed in.
A Practical Timeline for a Typical Custom Build
| Week | Activity | Gate |
|---|---|---|
| 1 | Spec lock, site visit, concept layouts | Customer signs off on footprint and cycle time |
| 2–3 | Top-down skeleton, station layout, supplier long-lead quotes | Layout review with customer |
| 4–6 | Detailed design: frame, tooling, actuators, panel | Internal design review, BOM out for quote |
| 7 | Drawing release, material order | Customer approves final drawings |
| 8–11 | Shop build, panel build, PLC programming in parallel | — |
| 12 | Internal debug and run-off | Internal acceptance test |
| 13 | Customer run-off in your facility | Customer signs FAT |
| 14 | Ship, install, SAT at customer site | Customer signs SAT |
That fourteen-week plan is tight but realistic for a 4–6 station machine. It only works if the gates are real — if the customer doesn’t sign off on the concept, you don’t start detailed design. The gate is not bureaucracy; it’s the thing that keeps you from building a machine the customer never actually wanted.
The Long Game: Standardize What You Can, Engineer What You Must
A custom equipment build doesn’t mean every part is custom. The most efficient shops I’ve worked in treat custom design like this: the interfaces between stations are standardized, the mounting patterns are standardized, the control architecture is standardized, and only the actual tooling that touches the product is truly one-off. The frame is from a previous project’s weldment. The safety circuit is the same as last quarter. The HMI screens are copied from the build before, with the customer’s logo dropped in.
That’s not cutting corners. That’s the difference between a shop that quotes a 14-week machine and one that quotes 22. The customer doesn’t pay you to reinvent the linear guide mounting hole every time. They pay you to solve the part-handling problem that nobody else has solved.
Build the process around that distinction. The next time a customer walks in with a weird part, you’ll spend your week one on the part, not on the frame.
