The mechanical and industrial engineer on moving from a heavy-fuel power station in Bangladesh to reliability work in American manufacturing, why failure leaves clues, and why the quietest year can be the most important one.

Interviewed by Musammad M Tamanna

Tanvir Anjum Anick has spent much of his professional life listening to machines before they become loud. In a power plant near Chattogram, on the Bay of Bengal, that meant watching the small movements of steam turbines, internal-combustion generators, waste-heat-recovery boilers and flue-gas systems that could not simply be allowed to stop. Later, in the American Midwest, it meant applying the same instinct to a high-speed food-processing line where downtime is measured not only in money, but in disruption.

The work sounds technical, and it is. Yet Anick talks about reliability in human terms. A bearing temperature, he says, is not just a reading on a screen. It is the electricity in a neighbourhood, the refrigerator in a family home, the hospital down the road, the production line that keeps food moving. That is why he became interested less in fixing breakdowns and more in predicting them.

“A schedule does not know whether a machine is healthy. The machine knows. It is sending signals the whole time.”

“I started beside machines that could not be allowed to fail,” he says. “You learn quickly when the lights of an area depend on your shift. Maintenance stops being a checklist and becomes a way of reading the machine before it tells everyone else there is a problem.”

That movement from instinct to evidence has shaped the rest of his career. After completing his mechanical engineering degree in Dhaka, Anick worked for more than six years in operations at a heavy-fuel-oil power plant. He then came to the United States for a Master of Engineering in industrial engineering at Lamar University, completed in 2024. The graduate training, he says, gave formal structure to what the plant floor had already taught him: machines fail for reasons, and those reasons often appear in the data before they appear as an emergency.

Research that does not stay inside one company

Anick’s argument for independent research is simple: the biggest benefits of reliability engineering come when methods are shared. A model locked inside one employer may help one operation. A validated framework published and adopted across industries can help many operators make better decisions.

“The value is in the spread,” he says. “If a predictive-maintenance framework helps one company, that is good. If it helps many companies run safer and more efficiently, that is the contribution I care about.”

Over the next five years, he wants to expand collaborations across equipment classes, align research with federal priorities in energy and resilience, and mentor younger engineers who are entering a world where mechanical systems and data systems increasingly depend on one another. He describes the goal as closing the distance between the paper and the pipe: making sure research does not sit unread while the machines it could help continue to fail in familiar ways.

The conversation also turned to the details behind Anick’s work: how engineers decide what to measure, why implementation is often harder than modelling, and how research can move from a journal article to a plant floor.

Q: What is the first sign, in your experience, that a machine is moving toward failure?

A : Usually it is not one dramatic sign. It is a pattern. A vibration reading starts to drift, a temperature trend behaves differently under the same load, or a motor begins drawing current in a way that does not match its normal profile. The mistake is waiting for one obvious alarm. By the time the alarm is obvious, the system may already be close to failure. Good reliability work is about noticing small changes early and asking whether they mean something.

Q: How do you decide which data matters and which data is just noise?

A : You begin with engineering judgment. Data is powerful, but it is not automatically useful. I ask: what physical behaviour does this sensor represent, and what failure mode could it help explain? If a signal has no connection to a real mechanism, it may make a model look sophisticated without making maintenance better. The best systems combine plant-floor knowledge with statistical discipline.

Q: A lot of companies collect equipment data now. Why do many still struggle to use it well?

A : Because collection is easier than interpretation. Many plants have historian data, PLC data, SCADA screens and maintenance records, but those systems often sit in separate places. The real challenge is turning scattered information into decisions: when to intervene, what to inspect, what risk to accept, and how to explain that decision to operations. Predictive maintenance is not only a technology project. It is a decision-making project.

Q: Where does artificial intelligence genuinely help in reliability engineering?

A : AI helps when the pattern is too complex for a simple threshold. A fixed alarm may tell you when temperature crosses a limit, but it may not understand how temperature, vibration, load and operating speed behave together. Machine learning can learn the normal behaviour of a system and identify combinations that look abnormal. But the model still needs engineering validation. I do not believe in black-box reliability decisions for critical equipment.

“The model should support judgment, not replace it.”

Q: What is the danger of relying too heavily on a model?

A : A model can be confident and wrong. That is why reliability engineers need humility. The model should support judgment, not replace it. If a recommendation does not make physical sense, we have to question it. The goal is not to make engineers disappear. The goal is to give engineers better evidence before they make a decision.

Q: How do you make research useful to someone working a night shift in a plant?

A : You have to translate it into something operational. A plant operator does not need a long mathematical explanation at 2am. They need to know what changed, why it matters, how urgent it is, and what action is recommended. I always think about that final mile: can the research become a procedure, a dashboard, a maintenance trigger, or a training tool? If not, it is incomplete.

Q: You have worked in both power generation and manufacturing. What connects those worlds?

A : The equipment is different, but the reliability logic is similar. Both environments depend on uptime, safety, energy efficiency and disciplined maintenance. In a power plant, failure affects the grid. In manufacturing, failure affects production, supply chains and sometimes food availability. In both places, small technical decisions can have consequences beyond the machine itself.

Q: What was the hardest adjustment after moving from Bangladesh to the United States for graduate study and work?
A The scale and pace were different, but the fundamentals were familiar. Machines still fail for physical reasons. Teams still need clear communication. What changed was the expectation that technical work should connect quickly to measurable outcomes: downtime, cost, safety, energy use, compliance. That pushed me to become more quantitative and more precise in how I explain value.

Q :What kind of engineers do you think industry needs more of?

A :Engineers who are comfortable crossing boundaries. Someone who can understand a pump curve, read a control-system trend, talk to technicians respectfully, and also build a data model has a real advantage. Industry does not need theory alone or experience alone. It needs people who can connect both.

Q : How do you want your work to be judged in the long run?

A : I would like it to be judged by adoption. Citations matter because they show that other researchers are using the work, but the deeper goal is practical change. Did the framework help plants reduce downtime? Did it help engineers catch failures earlier? Did it reduce waste or improve safety? Those are the questions that matter to me.

Q: What would you say to a young engineer who wants to enter reliability or industrial analytics?

A : Spend time with real equipment. Do not learn analytics only from clean datasets. Go to the floor, listen to technicians, understand how sensors are installed, how maintenance logs are written, and how decisions are made under pressure. Then learn the mathematics and programming. The strongest engineers are the ones who understand both the messy reality and the analytical method.

Q: Is there one principle that guides your work?

A : Do not wait for failure to become visible. In engineering, the most expensive problems often begin quietly. If we can learn to hear those early signals, we can protect people, equipment, energy and production before the damage happens.

K. M. Tanvir Anjum Anick is a mechanical and industrial engineer specializing in reliability engineering, predictive maintenance, industrial automation, and energy optimization. His work focuses on reducing equipment failure, improving plant safety, and advancing data-driven maintenance solutions for modern industrial systems.

Email: tanvir.anjum.anick1@gmail.com