Special machines: hand-tuned commissioning, know-how in one head

Every machine is tuned by hand at commissioning from the predecessor's values

What happens. The machine is built, the drives are connected and then the tuning starts: gains, filters, feedforward. The parameters come from the last machine, with changes until it runs. One or two people who know how do this on every machine, usually under time pressure at the customer.

What it costs. Days of commissioning per machine, travel and a delivery date that depends on the availability of the experts. Faults appear at the first full run with the product, when changes are expensive.

Why the usual fix fails. Copying the last machine works until the mechanics differ. A written tuning procedure helps only the people who already know the reasons for it. Training a new commissioning engineer takes years.

How a fixed-price remote diagnosis checks it. From recordings of drive tests and the mechanical data of two or three machines I identify the dynamics and check how far a model can set the parameters before the machine is at the customer. The fixed-price remote diagnosis answers whether a model-based tuning can shorten commissioning and who in your team could then do it.

Background: Replacing hand-tuned parameter tables by a model

The control concept of the machine lives in the head of one senior engineer

What happens. In many medium-sized machine builders, one senior engineer designed the control and knows why every parameter is set as it is. He is close to retirement. Documentation exists in pieces, and the younger colleagues use the settings without knowing the reasons.

What it costs. A risk to every order that touches the control, delays when he is absent and a loss of know-how that cannot be bought back once he has left.

Why the usual fix fails. A handover talk of a few weeks captures only what he thinks of. Asking him to write documentation competes with orders. Hiring a successor with the same experience is not possible in the time left.

How a fixed-price remote diagnosis checks it. From the code, parameter lists, drawings and a few test recordings I rebuild the model of the machine and write down what the control does and why, in a form your colleagues can use. The fixed-price remote diagnosis answers how much of the know-how can be saved in a defined piece of work, before it leaves.

Small firms have controller programmers but no control engineers

What happens. A small machine builder has people who program the controller well: sequences, safety, communication. A control loop that is unstable, a temperature that oscillates or a drive that vibrates is another discipline. The problem stays on the list and the machine is shipped with a workaround.

What it costs. Customers with a machine that runs below its quality, service visits for the same fault and a team that has learnt to live with it.

Why the usual fix fails. Trying settings until the loop looks better works for the day and not for the next product. Training a programmer in control engineering takes long, and the daily work does not stop. A permanent control engineer is not affordable for a small firm.

How a fixed-price remote diagnosis checks it. From the loop recordings, the programme and a description of the machine I check why the loop does not hold and what a model-based fix would be. The fixed-price remote diagnosis answers what an external package with a full handover would contain, so that your own people can maintain it.

Customers call service to retune machines

What happens. A machine is delivered with good settings. After a year the product, the material or the wear is different, and the machine does not hit the quality any more. The customer calls, and a service engineer travels to retune it on site.

What it costs. Service visits for settings, downtime at the customer and a service staff that is busy with things the machine could do itself.

Why the usual fix fails. A longer manual does not help an operator who does not know control. A remote login lets the engineer retune without travel and still needs the engineer for each case. Wide margins in the delivered settings cost performance from the first day.

How a fixed-price remote diagnosis checks it. From service reports and recordings of machines before and after retuning I check what changes in the machines and whether a self-tuning function inside the product could do the adjustment. The fixed-price remote diagnosis answers which cases such a function can cover and where service remains needed.

Background: Control that adapts to drift and wear

New operators need a long time to run complex machines

What happens. A complex machine has many parameters, and the best settings depend on the product. An experienced operator feels what to change. A new operator needs months before the output and the quality match. During that time the machine runs worse than it could.

What it costs. Scrap during the learning phase, lower output and a customer who blames the machine builder for the result.

Why the usual fix fails. Training courses show the controls and not the reasons. A thick settings table is followed blindly and fails at the first new product. The best operator cannot be everywhere.

How a fixed-price remote diagnosis checks it. From machine data of good and bad runs with the settings I check which parameters decide the result and whether a model could suggest a setting or show the effect of a change before it is made. The fixed-price remote diagnosis answers whether a simple assistant or a training simulator for your machine is realistic.

Nobody understands the old control code any more

What happens. The machine control was written years ago, extended many times and documented little. The author has left. A change request means finding the place in the code, guessing the effect and testing on the machine at the customer. The team avoids touching it.

What it costs. Customer requests that are refused or delayed, risk for each change and a machine that cannot get the new functions the market asks for.

Why the usual fix fails. Adding a patch is the cheapest step now and makes the next one harder. A conversion of the code line by line to a new controller carries the old structure along. Waiting for the controller to fail makes the rebuild an emergency.

How a fixed-price remote diagnosis checks it. From the code, the recorded behaviour of the machine and a list of what customers want changed, I check what the machine really does and what a rebuilt concept would look like. The fixed-price remote diagnosis answers whether a rebuild with a model, a new concept and documentation is better than further patches.

Weld seam quality is inspected after the fact

What happens. The seam is welded, then it is checked: by eye, with ultrasound or with a destructive test on samples. A bad seam is found minutes or hours later, when the part has moved on. The welding system itself records current, voltage, speed and wire feed, but nobody judges the seam by these signals.

What it costs. Rework or scrap of finished parts, inspection effort and the risk of a bad seam that is not found and goes to the customer.

Why the usual fix fails. More inspection stations add cost and delay. A limit on one signal, such as current, flags too many good seams and misses subtle defects. Selling a camera system on top adds a second system to maintain.

How a fixed-price remote diagnosis checks it. From process signals of welds with known quality I check how well a computed quality value follows the results of the inspection and which defects it would show early. The fixed-price remote diagnosis answers whether quality computed from the process can warn during the weld and what would be needed to rely on it.

Control architecture, testing and commissioning of complex equipment is slow

What happens. A contract manufacturer of equipment builds machines to the design of its customers, with their controls. The architecture of the control has to be understood, tested and commissioned for each new type. The knowledge for it sits with a few people, and each type starts almost again from the beginning.

What it costs. Long commissioning, test benches that are occupied for weeks and a delivery that depends on the person who knows the architecture.

Why the usual fix fails. A detailed specification from the customer is rarely complete. More testing at the end finds the faults when they are expensive. Copying the previous project transports its errors.

How a fixed-price remote diagnosis checks it. From the architecture drawings, the interfaces and the logs of earlier commissioning I check which parts of the tuning can be prepared by a model before the equipment is complete. The fixed-price remote diagnosis answers where the pre-tuning saves test time and what has to be checked on the real equipment.

New machinery rules ask for bounded and documented adaptive controllers

What happens. The new European Machinery Regulation also covers machines whose control logic changes by itself or uses learned functions, and it sets requirements for cybersecurity and allows digital documentation (IHK Nürnberg). Builders who want to sell a machine with an adaptive function have to explain what it can do, and what it cannot.

What it costs. Extra effort in the safety argument for each such machine, delays at the conformity assessment and the risk that a learned function is rejected because its behaviour cannot be bounded.

Why the usual fix fails. A black-box learned controller is hard to bound. Switching off the adaptive function saves the approval and loses the benefit. A safety layer added afterwards does not make the controller behave better.

How a fixed-price remote diagnosis checks it. From the control concept and the operating limits of a machine I check whether a model-based controller with explicit constraints can give the same adaptation in a form that is easier to document. The fixed-price remote diagnosis answers what the safety argument would contain and what proof of the limits can be written down.

Virtual commissioning is used rarely and faults still show at the first test

What happens. Virtual commissioning tests the control against a model of the machine before the hardware exists. Many builders do not use it, and those who do use it rarely. In most projects the faults of the control appear at the first function tests on the finished machine, when the delivery date is near.

What it costs. Weeks on site, travel and fixes made under pressure, with parts of the control changed at the last minute.

Why the usual fix fails. A model made for the design is not detailed enough for the control. A model built for the control costs time that the project plan did not include. Testing only the logic does not show the dynamics of drives and processes.

How a fixed-price remote diagnosis checks it. From drawings, drive data and the control code of a finished machine I check which model detail is needed to find the typical faults earlier and how well the model can be matched to the real machine. The fixed-price remote diagnosis answers what a virtual test would have found on your last machine and whether controller pre-tuning is within reach.