Industrial Automation Product Development

Start from the part and the act, not from a catalogue of hardware. What must be true before the act is allowed, what confirms the act happened, and what the station does when the confirmation does not come. Special-purpose machines follow the same rule: the machine exists to perform one class of operation, so the operation is the specification and the frame is how you hold that operation still. Stage questions for the build are set out in Hardware Product Development Process Explained. The millisecond budget below is invented teaching arithmetic, not a BrahmWorks cycle time, price, yield or test result.

Write states a technician can point at. Part present, part seated, guard closed, act permitted, act complete, act failed with the energy removed in a way you can explain. A timer standing in for a sensor is acceptable on a bench only if the result says the sensor was absent.

Recovery is part of the sequence. After a failure, what is the mechanical position, what energy is stored, and what may a person touch. If only the programmer can reset the station, you have a demonstration. Watch a trained person recover while you stay quiet.

The surrounding line is an input: pitch, how parts arrive, and the variation in those parts. Look at a handful of real parts before you freeze a nest. That is not a capability study. It is a refusal to automate a perfect sample you do not have.

The fixture decides whether the sensor can see. A flag that works only when a part is pressed home by hand will pass in the lab and fail when the actuator lets go. Design the seated condition, then the sensor, then the input. The reverse produces a program that matches a wish.

A command is not a confirmation. A valve told to move is not a cylinder that moved. Motion also brings pinch and stored energy. Guarding and isolation are part of the product, scoped with someone qualified for that site. This article states no safety rating and reports no test.

Weekly access is a requirement. Filters, sensor faces and wear parts have to be reachable in the clothes people wear. A cover that needs the cell dismantled for a routine clean will be left off.

A special-purpose machine is justified when a standard machine cannot hold the operation, the part or the rate you wrote down. It is not justified by a wish to own a unique frame. The unique parts should be the ones that touch the operation: nests, tools, paths. Everything else that can be a bought axis, guard or control should stay bought, so that spares and skills exist when you are gone.

The prototype of a special-purpose machine should answer whether the operation works on real part variation, with the handling you mean to ship. A beautiful frame around an unproven operation freezes the wrong thing. Build the operation in a crude, safe rig first if that is the open question. Promote the rig to a machine only when the operation is the thing you are no longer guessing.

Documentation for a plant is part of delivery: inputs, outputs, the sequence, the recovery, the sensors that may be adjusted, and what must not be adjusted. A controller file without that narrative cannot be owned by the site. Agree who owns the file and who may change it before the machine arrives.

The teaching product must see a bracket seated before a press is allowed to cycle. It is not a BrahmWorks project, not a plant acceptance and not a tested machine. The question is whether three real brackets, including one slightly bent, are clearly seated or clearly refused, and whether a person can recover without opening the program.

In the build: the nest, two sensing methods, a permission signal rather than the production press, and a written recovery. Out of the build: the press itself, a guard design, a line-rate claim and any compliance statement. If the bent bracket looks seated to both sensors, the nest has failed. That is a useful prototype. Do not drop the bent part from the sample because it was awkward.

These times are assumed for the arithmetic. They are not a measured cycle, not a line rate and not a BrahmWorks result.

Assume the cell will give this station 400 milliseconds. Assume sense 40, decide 15, valve response 25, cylinder travel 180 and confirm 30, in milliseconds, and assume they happen in series.

40 plus 15 is 55. 55 plus 25 is 80. 80 plus 180 is 260. 260 plus 30 is 290 milliseconds. The margin against 400 is 110 milliseconds.

Change only the travel, because the stroke was copied from a shorter cylinder. At 250 milliseconds of travel the sum is 290 minus 180 plus 250, which is 360 milliseconds. The margin is 40 milliseconds. Nothing else in the controller changed. A pitch that looked comfortable becomes a pitch that fails if any assumption grows, and this sum has no queue, no retry and no part variation in it. Do not publish 290 milliseconds as the station time. Replace every term with a measurement or a supplier figure you are willing to file, and keep the series assumption visible. If two motions overlap, draw that overlap. Do not subtract it in your head.

• The sequence, including failed acts and recovery, is written in states a technician can see.
• Each state has a real input, or the stand-in is labelled on the result.
• Part variation has been looked at before the nest was frozen.
• Command and confirmation are separate.
• Guarding and energy isolation are a separate, qualified scope, not a sketch.
• Weekly access has been tried in the clothing the plant wears.
• Unique machine parts are limited to what the operation needs.
• The site can obtain the narrative, not only the controller file.

It is one part of the product. The nest, the sensors, the actuators, the recovery and the documents are the rest. A program that passes on switches can still be the wrong product. Review the station, not the repository.

Before the nest and the recovery are frozen. The people who reset faults and change shifts know facts a lab will invent. A late demonstration that surprises them is how a machine arrives that they will bypass. Bypassing is then your design, whether you intended it or not.

A rig may prove the operation if it is safe and the limits are written down. A station also has to be run, recovered and maintained by the site, at the pitch you agreed. Promote the rig only when the operation is no longer the guess. A welded frame does not perform that promotion by itself.

No. Time is the sum of mechanical motions, sensing, retries and the parts you were given, under the overlap you can draw. The program's scan is rarely the term that hurts. The labelled budget above is a way to see assumptions. It is not a promise.

Bring the state list, the awkward real parts, and the recovery a stranger must perform. BrahmWorks can say whether the next build is still the operation, or whether you are about to freeze a frame around an open question.

Review the next stage with BrahmWorks