Systems integration: is automation worth it?

Integrating an automation system means getting machines, sensors, robots and programs to work as a whole. But before the how comes a question many technicians never ask: is it worth it for the company? A badly chosen automation project can cost a lot of money and improve nothing. Being able to analyse it with numbers is what turns a good technician into someone whose opinion is asked before anything is bought.

First, measure: you do not automate by eye

Before proposing anything, you need to know how the process works today. The minimum data:

• Real output per shift, not what the machine’s brochure says.
• Cycle time of each station: how long each operation takes.
• Stoppages: how many, how long they last and why (breakdowns, lack of material, changeovers).
• Rejects: how many parts come out bad and at which station.
• People per station and per shift, and the real cost of each hour.
• Demand: how many parts must be made, today and foreseeably tomorrow.

From the demand you calculate the takt time, the pace at which you must produce to meet it:

takt time = available time / parts demanded

Example: a shift with 7.5 productive hours has 27,000 seconds. If 900 parts must be made per shift, the takt time is 27,000 / 900 = 30 s. Any station that takes more than 30 s per part makes it impossible to meet demand.

The bottleneck

A line with several stations in a row produces at the pace of its slowest station. That station is the bottleneck, and it sets the pace.

Two very practical conclusions follow from this:

• Automating a station that is not the bottleneck does not increase output. If station 1 goes from 20 s to 10 s, the line still makes one part every 35 s. It may have other advantages, such as less physical effort or better quality, but not more parts.
• When you improve the bottleneck, another one appears. If station 2 goes down to 25 s, the new limit is station 3, at 25 s. In a 27,000 s shift, output goes from 771 parts (27,000 / 35) to 1,080 (27,000 / 25): 40 % more, and there is no point taking station 2 below 25 s unless station 3 comes down too.

OEE: how much a machine really delivers

OEE (overall equipment effectiveness) sums up in one number how much of what a machine could produce it really produces. It is calculated by multiplying three factors:

• Availability = run time / planned production time. What is lost to stoppages: breakdowns, changeovers, waiting for material.
• Performance = (total parts × ideal cycle time) / run time. What is lost by running slower than planned and by minor stops.
• Quality = good parts / total parts. What is lost to bad parts.

OEE = availability × performance × quality

A complete example

An 8-hour shift (480 min) with 30 minutes of planned breaks leaves 450 min of planned production time. There were 45 min of stoppages, so the machine ran for 405 min. Its ideal cycle is 30 s per part: in 405 min (24,300 s) it could have made 810 parts, but it made 680. Of those, 660 were good.

• Availability = 405 / 450 = 90 %
• Performance = 680 / 810 ≈ 84 %
• Quality = 660 / 680 ≈ 97 %
• OEE = 0.90 × 0.84 × 0.97 ≈ 73 %

Quick check: the 660 good parts, at 30 s each, are worth 330 min of perfect work. 330 / 450 ≈ 73 %. Same result.

85 % is often quoted as world class, and many factories are well below it. But the usefulness of OEE is not comparing yourself with others: it is knowing where the losses are. In the example, the biggest loss is in performance: minor stops and slow running. It might be fixed with a badly adjusted sensor or a recurring jam, without buying anything. If the loss were in availability because of long changeovers, the solution might be to organise those changeovers better, not a robot.

When automation pays off

• Repetitive, high-volume tasks with a stable product.
• Dangerous or hard tasks: heavy loads, awkward postures, toxic or hot environments. Here the benefit is also measured in injuries and sick leave avoided.
• Quality that depends on each operator’s hand and varies from shift to shift.
• The bottleneck, when there is demand to produce more.
• Labour shortages for jobs nobody wants to do.
• Traceability required by the customer: automatically recording what was done, when and with which parameters.

When it does not pay off, or not yet

• The process is not stable. Automating a process that works badly only makes it fail faster and at greater cost. Stabilise first, then automate.
• Low volume and high variety. If the product changes every week, the cost of adapting the machine eats up the saving. Cobots and flexible tooling have improved this, but it is still the hardest case.
• The product is about to change. A machine designed for a product that will disappear within a year rarely pays for itself.
• Nobody can maintain it. If nobody in the plant can troubleshoot the machine, every breakdown will be a long stop waiting for an outside technician.
• The problem is something else. Lack of material, poor organisation or poor maintenance are not fixed by buying a machine.

The numbers: investment and return

The calculation looks simple, but it is often done badly because costs are forgotten and savings are inflated.

What it really costs

Not just the machine: engineering, installation, safety guarding (fencing, light curtains, interlocks), programming, commissioning, staff training, spare parts, annual maintenance and the production lost while it is being installed.

What it really saves

Labour hours (at their full cost to the company, not take-home pay), fewer rejects, extra capacity if there is demand, and fewer injuries. An important point: labour savings are only real if those hours stop being paid or are put to other useful work.

The payback period

payback period = total investment / net annual saving

An example with made-up round figures: a palletising cell with a total investment of €120,000. It frees one operator on each of two shifts, at €30,000 a year each in cost to the company: €60,000. It cuts rejects from knocks by €6,000 a year. Its maintenance costs €6,000 a year. Net saving: 60,000 + 6,000 − 6,000 = €60,000 a year. Payback period: 120,000 / 60,000 = 2 years.

Industrial companies often ask for the investment to be recovered in two or three years, although it depends on the company. A proposal with honest numbers, including the costs others forget, inspires far more confidence than a spectacular saving.

Intermediate options

Between doing nothing and a fully automatic line there are many steps, often more profitable:

• Error-proofing (poka-yoke): guides, stops and sensors that make it impossible to assemble a part wrongly.
• Handling aids: manipulators, lift tables, vacuum lifters.
• Semi-automation: the operator loads and unloads, the machine does the operation.
• Cobots: collaborative robots that are easier to reprogram. Careful: they also need their own risk assessment; “collaborative” does not mean “risk-free”.
• Retrofit: replacing an obsolete PLC, adding drives or upgrading the safety of a machine that is mechanically sound.
• Measure first: a few sensors, the PLC and an HMI that record stoppages and cycles. It costs little and gives you the data to decide on everything else.

The technical side: getting machines to understand each other

When machines from different manufacturers are joined together, the critical point is the interface: which signals are exchanged and what each one means. A typical example between a machine that delivers parts and a robot that picks them up:

This exchange of “request, response and confirmation” is called a handshake. For it to work well:

• It is documented in a signal list agreed in writing: name, address, direction and meaning.
• Every wait has a time limit: if the confirmation does not arrive, a clear alarm is raised instead of the machine sitting there with no explanation.
• You decide whether the signals are hardwired or over a network (for example PROFINET between two PLCs) and what happens if communication drops.
• Safety functions, such as an emergency stop shared by several machines or a door that must stop the robot, are not handled with normal PLC signals: they are designed with safety components and circuits.

Before the system is signed off, acceptance tests are carried out: the FAT (factory acceptance test), at the builder’s workshop, and the SAT (site acceptance test), once installed at the customer’s plant.

Safety and CE marking: the most forgotten risk

In the European Union, every machine must carry the CE marking, and its manufacturer is responsible for assessing its risks, preparing the technical file, drawing up the declaration of conformity and writing the manual. The reference is the Machinery Directive 2006/42/EC, which Regulation (EU) 2023/1230 replaces from 20 January 2027. Outside the EU the rules differ: Great Britain, for example, has its own UKCA marking, and in the United States machine safety is governed by OSHA regulations and standards such as NFPA 79.

• Assemblies of machinery. Joining several machines, even if each one is CE marked, so that they work as a whole can make the integrator the manufacturer of the assembly, with all the obligations that brings.
• Substantial modification. Changing an existing machine so that new risks appear or existing ones increase (more speed, new functions, removed guards) may require a new conformity assessment. The new regulation defines this concept explicitly.
• Equipment already in use. In the EU, Directive 2009/104/EC sets the minimum safety requirements for the use of work equipment.
• Supporting technical standards: EN ISO 12100 for risk assessment, and EN ISO 13849-1 and EN IEC 62061 for the parts of the control system that perform safety functions.

This section is guidance, not legal advice: on each project, a safety manager or a CE marking expert must confirm which obligations apply.

Checklist before proposing an automation project

1. Do I have real data on output, stoppages, rejects and cycle times?
2. Do I know which station is the bottleneck and whether there is demand to produce more?
3. What is the current OEE and where are the biggest losses: availability, performance or quality?
4. Is the process stable, or does it need fixing first?
5. Is there a cheaper intermediate option that solves most of the problem?
6. Have I included every cost: engineering, safety, commissioning, training and maintenance?
7. Are the savings real and calculated at full cost to the company?
8. What is the payback period, and is it acceptable to the company?
9. Is it clear who will be the manufacturer for CE marking purposes?
10. Are there people who can maintain and troubleshoot the system?

Summary

• Before automating, measure: output, stoppages, rejects, cycle times and demand.
• The line runs at the pace of its slowest station. Automating another station does not give more parts.
• OEE = availability × performance × quality, and its value lies in showing where the losses are.
• Payback period = total investment / net annual saving, with every cost included.
• An unstable process is not automated: it is fixed first.
• The interface between machines is documented, and safety is designed with safety components.
• Integrating or modifying machines can bring CE marking obligations.