Practical instrumentation skills matter more than ever and your ability to troubleshoot is critical.
The scaremongering has reached a crescendo; with the assertion that AI tools will replace knowledge-based professionals, including engineers. I don’t deny that AI is becoming increasingly sophisticated – including the ubiquitous ChatGPT – but the reality is more nuanced.
To save time, engineering personnel are using AI to construct snippets of PLC code, make design suggestions, summarise manuals, generate ideas for loop tuning, and describe process optimisation. But when SCADA screens alert process operators to a plant spinning out of control, nobody calls a chatbot. They call the troubleshooting expert.
AI tools are based on probability, suggesting the next word in a sentence, for instance. A bit like Google on steroids. AI can find information efficiently and provide advice based on a given prompt. Despite presenting these results with confidence, even conviction, AI is deeply flawed. Users need to be hyper-vigilant; everything it produces needs a beady-eyed expert.
Theory – the feeding ground for AI – is tested by reality. Consider these briefly sketched scenarios. A pressure transmitter is installed ‘as per spec’, but the impulse lines are partially blocked. The status is ‘healthy’, but in wet gas a badly ranged DP transmitter kills turndown – low-flow DP vanishes and the numbers lie. The temperature appears normal, but the associated control valve is either faulty or the product entering the plant has changed grade. This is the world of instrumentation and automation professionals: a place where measurement is never just a number, and control is never just code.
AI can be a useful adviser – a ‘chum on the side’. What it can’t smell is hot insulation, hear pump cavitation, spot the subtle change to the vibration in an actuator, or feel the increasing heat on a terminal strip that’s about to become tomorrow’s incident report. It cannot walk the line, check an instrument air filter, or link that ‘mystery fault’ with a washdown cycle and a poorly sealed junction box. It cannot spot a poorly trained or over-tired operator, and it is not responsible when an oversight becomes a trip, a spill, or a near-miss. Humans are.
In our industry troubleshooting is the career moat. AI can recite the theory of pressure, flow and temperature measurement, but it cannot mimic experience and diagnose failure modes under pressure – calmly, methodically, with discipline.
Consider two common scenarios where hard-earned experience is essential: Firstly, the ‘perfect’ PID loop that still hunts. The tuning is textbook, but the loop oscillates because the valve is sticky or the actuator is undersized. Or the PLC that lies by omission, an intermittent trip that disappears when you watch it. The culprit is often noise on inputs, earthing/shielding errors, or a vibrating 24V rail.
Budding instrumentation professionals can be trained using EIT’s online learning platform – while working. The teaching and learning sessions are live and interactive, covering job-aligned modules and presented by real, grizzled instrumentation veterans (not AI-bots or even humans with PhDs). There are troubleshooting exercises using realistic scenarios, and assessments that reward diagnosis and decision-making, not memorisation.
A country’s competitiveness depends on its ability to proficiently troubleshoot in the industries which underpin its economy: energy, mining, processing, oil and gas, food and beverage manufacturing, and others. AI will be there as a sounding-board, but people and their skills build the national capability.