Automation Is Not Only for Large Plants

The companies that benefit most from a custom machine usually assume they are too small for one: a 20-person shop assembling by hand, a food producer filling containers with a scoop, a contract packager whose margin is entirely labor. A single semi-automatic station that removes one operator from one repetitive step often pays for itself in under 18 months.

What makes these projects fail is almost never the mechanics. It is starting with a machine concept instead of with a process, and discovering during commissioning that the real production problem was somewhere else.

Decide What Not to Automate First

Automation multiplies whatever process you point it at, including a bad one. Before scoping any machine, be honest about four screening questions.

  • Is the process stable? If operators do the step differently on different days, or if the incoming material varies, a machine will jam on the variation a human absorbs without noticing. Stabilize and document the manual process first.
  • Is the volume real and durable? A machine sized for a product that gets redesigned in nine months is scrap. High-mix, low-volume work usually wants flexible fixtures and better tooling, not dedicated automation.
  • Does the step require judgment? Inspection tasks where a person decides whether something looks acceptable are the most expensive to automate. Machine vision can do them, but the specification effort is significant.
  • Is the bottleneck actually there? Automating a station that is not the constraint produces inventory, not throughput.

Often the honest answer is that a set of well-made fixtures fixes 70 percent of the problem for a few thousand dollars. Printed and machined jigs are a legitimate first step, not a consolation prize, and the approach is covered in 3D printed jigs and fixtures for the production line.

Three Tiers, Three Budgets

Tier one: an operator-assisted station. A fixture, an actuator or two, a pneumatic press or dispenser, a foot switch, and a simple controller. The operator loads and unloads; the machine does the precise or forceful part. Typically $15,000 to $60,000 and three to five months. Best return per dollar in the whole category.

Tier two: a collaborative robot cell. A cobot arm handles pick, place, screwdriving, dispensing, or machine tending next to a person, usually without a full guarding cage after a risk assessment. Installed cost typically $50,000 to $150,000 including the arm, end effector, fixtures, integration, and safety validation. Cobots changed the arithmetic for small manufacturers because the integration cost, not the arm price, used to be the barrier.

Tier three: a dedicated automated line. Indexing conveyor or rotary table, multiple stations, vision inspection, and a supervisory controller. $250,000 to well over $1 million, nine to eighteen months, justified only by volume that runs for years.

The Payback Calculation, Done Honestly

Take the fully burdened labor cost of the operator hours the machine removes, including benefits, payroll taxes, supervision, turnover, and overtime, which in most US manufacturing lands between $28 and $45 per hour rather than the wage rate. Multiply by the hours per year the machine actually runs, not by capacity.

Then subtract what nobody puts in the spreadsheet: maintenance and spares at roughly 5 to 8 percent of capital cost per year, the technician time the machine now requires, changeover time, and a realistic availability figure. A new machine running at 85 percent availability in its first year is normal, not a defect.

Add the benefits that are not labor. Scrap reduction is frequently larger than the labor saving and it is measurable, and so is consistency: a machine that holds a process within a tight band lets you demonstrate capability to a customer, which is where statistical process control starts paying off. A payback under two years is easy to approve; over three, the product had better be stable.

Write the Specification Before Anyone Draws a Machine

The document that governs a machine project is a user requirement specification, and it is written by the buyer, not the builder. It states the parts to be handled with their real dimensional variation, the required cycle time and how it will be measured, the uptime and yield the machine must demonstrate, the utilities available, the footprint and access constraints, changeover requirements, and the acceptance criteria in numbers. The structure is close to what any hardware project needs, described in how to write a product requirements document.

Two clauses save the most money later. First, specify that the cycle time is measured on your worst-case material, and supply that material for the run-off. Second, specify who owns the controller program and the electrical drawings. A machine whose PLC code you cannot access is a machine only its builder can ever modify.

Motion, Controls, and Safety

Most machine design decisions collapse to motion choices. Pneumatics are cheap and fine for two-position moves; anything needing programmable position or force wants electric actuation, per how to choose a linear actuator and stepper vs servo vs brushed DC. Size drives against real load including friction and acceleration, and leave 30 percent margin, because the first question after commissioning is whether it can run faster.

On controls, a PLC remains the right default for anything a plant technician must service; reserve industrial PCs for vision and complex sequencing. Build the panel to recognized industrial wiring and listing practice from the start, since retrofitting one to pass inspection after installation is slow and expensive.

Safety is a design input, not a shipment checklist. A formal risk assessment belongs at concept stage, not before shipment. It drives guarding, light curtains, interlocks, emergency stop architecture, and the required reliability level of the safety circuit, all of which change the machine's layout. Collaborative applications need a validated force and pressure assessment for every contact scenario, which regularly forces a speed reduction that changes the cycle time you promised. Treat guarding and electrical safety as hard constraints, the same way you would treat the certification requirements in product safety testing requirements.

Run-Off, Commissioning, and Handover

Accept the machine in two stages. A factory acceptance test at the builder's site, using your parts, proves function before anything is crated. A site acceptance test after installation proves it in your environment with your operators and your utilities, and should include a continuous run of a defined length at a defined availability, not a demonstration of ten good cycles.

Handover is a deliverable list: electrical schematics, pneumatic diagram, spares with lead times, maintenance schedule, operator training, and the controller source. Structure the closeout per acceptance testing and project handover, and hold retention against it.

Scoping a Machine for Your Line

Projects House specifies and develops custom production machines: process study, automation tier selection, payback modeling, mechanical and controls design, and acceptance planning. Describe the operation you want to automate, your volumes, and your cycle time target in the contact form.