Published on 31-Jul-2026
What Is Predictive Maintenance? The Role of NDT in Critical Infrastructure
Table of Contents
- Session Overview
- Meet the Panel
- What Is Predictive Maintenance? Defining the Term
- Preventive vs Predictive Maintenance: The Critical Shift
- Which NDT Technique Matters Most for Predictive Maintenance?
- Data Interpretation and Decision-Making
- Cost, Risk and Asset Life Extension
- The Next Five Years: AI, IoT and Continuous Monitoring
- Audience Q&A Highlights
- Key Takeaways
- Frequently Asked Questions
Session Overview
Infrastructure assets are ageing while operating conditions become more demanding. Downtime is costly, failures are unpredictable, and safety is non-negotiable. Against this backdrop, predictive maintenance has emerged as a key strategy, and at the heart of it lies non-destructive testing, enabling condition-based decisions rather than reactive actions.
This session of NDT Talks brought together leaders and practitioners who work closely with asset integrity, reliability, corrosion and maintenance optimisation in the real world. The discussion explored what predictive maintenance actually involves across sectors like oil and gas, energy, ports, transportation and heavy industry, and where NDT genuinely fits into the framework, beyond buzzwords.
As the moderator framed it at the outset: predictive maintenance has become a strategic priority for organisations managing critical infrastructure, but it only works when decisions are backed by reliable data and engineering judgement. This article captures the full conversation on how NDT helps asset owners move from time-based to condition-based maintenance.
Watch the full discussion on the role of NDT in predictive maintenance for critical infrastructure on YouTube.
Meet the Panel
The session was moderated by Mr. Vinay, Quality, Project and NDT Manager and ASNT Level III, with deep expertise in NDT cost optimisation, continuous improvement and project execution.
What Is Predictive Maintenance? Defining the Term
The panel was asked to open by defining predictive maintenance in the context of critical infrastructure, and to identify where NDT plays its most important role today. Both panelists arrived at essentially the same predictive maintenance meaning, from different operational angles.
A Reliability Engineer's Definition
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“In my experience working as an RBI engineer back at ExxonMobil for petrochemical plants, predictive maintenance has been the key, because it gives safe sleep to the management, knowing that the plant is reliable, with predictable outcomes. I have seen cases where advanced ultrasonics, TFM and FMC have given early detection of HTHA and other important damage mechanisms. Without predictive maintenance you are going to be counting the days, or you are just going to be blind, not knowing when your unplanned shutdown is going to come up.”
A Simple, Operational Definition
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“When you speak about predictive maintenance, especially when it comes to critical infrastructure, the meaning for me has always been the shift that happens from the state where we take action after the failure, into the state where we predict the failure before it happens. This is the main value we are looking for. Rather than waiting to operate to fail, we do our own groundwork using the NDT and inspection methods available in the market, so we can expect the failure before it happens.”
In practical terms, predictive maintenance involves collecting measurable condition data from an asset while it is in service, using that data to estimate how quickly the asset is degrading, and then scheduling intervention based on that actual condition rather than on a fixed calendar. Predictive maintenance is used to determine remaining useful life, corrosion rate, crack growth rate and the appropriate timing of repair or replacement, all before a failure occurs.
Panel consensus: NDT is the backbone of predictive maintenance. As Mr. Taufik put it, without it you cannot really stand and have a case, or else it would be just guesswork rather than a reliable predictive maintenance procedure.
Preventive vs Predictive Maintenance: The Critical Shift
One of the most valuable segments of the discussion addressed a question many asset owners still wrestle with: what is the difference between preventive and predictive maintenance, and why does the distinction matter so much for critical infrastructure?
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“To shift from preventive maintenance into predictive maintenance really changes the concept of time-based inspection into condition-based inspection. We have been following time-based inspection all along, internal inspection every 5 years or 10 years, external inspection every so many years, which is not actually accurate enough to build a reliable maintenance plan. Using NDT you can get measurable data that you can use for whatever assessment you want: to estimate corrosion rate, to estimate crack growth, to know exactly the rate at which this equipment or asset is degrading, so you can plan better.”
Preventive vs Predictive Maintenance at a Glance
Where Organisations Struggle Most
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“Time-based or schedule-based is very simple in terms of planning, but when you want to transition to predictive, which is risk-based, you need a lot of inputs from the mechanical, corrosion, operational and process aspects. You need to know the operating parameters, whether they are within limits from a corrosion perspective or a cracking mechanism perspective. It takes a lot of effort and joint hands from inspection, materials, mechanical and chemical engineers. But once you do it, you see the effect in the long term.”
Mr. Taufik emphasised that the transition should never be applied blindly. The correct approach is to first study the plant, understand the risk categories present, and then deploy NDT according to the risk and consequence of failure of each individual piece of equipment. This is the essence of risk-based inspection, or RBI: spending inspection effort where the risk actually exists.
Which NDT Technique Matters Most for Predictive Maintenance?
When asked which NDT technique provides the most value for predictive maintenance, both panelists resisted naming a single winner, and their reasoning is instructive for anyone building an inspection strategy.
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“There is no one answer to this question. It really depends on the case. What is the damage mechanism? If you are looking for pitting, it would not be smart to say I am going to use long range. If you are looking for corrosion under insulation, that does not mean you can use ultrasonics directly without removing the insulation. You have to get the case first, and based on the damage mechanism, the material, accessibility and operating conditions, all these together help you find the right technique.”
The Screening-Then-Quantifying Approach
Both panelists independently described the same two-stage methodology, which emerged as one of the strongest practical recommendations of the session.
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“I always prefer a combination of more than one method rather than using one method. I recommend we use a method we sometimes call screening, that can have larger coverage, volumetric inspection, so I get an idea about the entire structure. I prefer long range UT or acoustic emission, techniques that can inspect a large structure and pinpoint where exactly the area of concern is. Then I go to that location and do more quantitative methods. For me this is the most effective approach, rather than doing a quantitative technique at very small limited areas that might not necessarily represent the entire structure.”
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“There is no one best technique for everything. I started off very strongly in ultrasonics, but after real operative experience in the field as a fixed equipment engineer within a petrochemical plant, I realised each technique has its own advantage. If you are looking at stainless steel chloride stress corrosion cracking, sometimes it may be easier to just go and do a quick PT rather than trying to get complicated ultrasonics. For tube inspection, IRIS gives you very minute detail, but it takes time and effort, so you go for RFT first to screen, then nail down on a couple of tubes for detailed quantitative data.”
If You Had to Pick One
Pressed to single out one method, Mr. Taufik was clear: ultrasonics in general, and advanced ultrasonics in particular, plays the most important role in predictive maintenance today. Most of the data used for integrity assessment is either corrosion data or crack growth information, and both of those damage mechanisms are most reliably detected using ultrasonic methods. Advanced techniques such as phased array, TFM and FMC now extend that capability significantly further.
Practical rule of thumb from the panel: screen wide with a qualitative or long-range technique to find the areas of concern, then apply a quantitative technique at those specific locations to generate the data your predictive maintenance model actually needs.
Data Interpretation and Decision-Making
Predictive maintenance does not depend on inspection alone. It depends on interpretation and action. The panel was asked how organisations can ensure NDT data is effectively converted into actionable maintenance decisions.
Interpretation Is Getting Harder, Not Easier
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“As we get more advanced, with phased array, TFM, FMC and acoustic-based techniques, interpretation becomes a critical factor, and it gets harder with the advancement of technology. We are doing ultrasonics 3,000 metres below sea level, where we get a lot of noisy data and electrical disturbances. The skill set of the person analysing the advanced NDT or the complicated geometries needs to be top-notch. It is very easy to make false calls, to call something acceptable a rejectable defect, or to miss something. In our cases, checking whether risers are flooded three kilometres below sea level, even minute indications are critical.”
The Three-Part Framework
Mr. Taufik offered what became one of the clearest structural takeaways of the session: a three-part framework for converting NDT data into predictive maintenance action.
- Data quality and consistency, driven primarily by the competence of the person acquiring the data, supported by qualified procedures and calibrated equipment.
- System integration, where the data must feed into a system such as an RBI platform or predictive maintenance software that converts raw measurements into assessments and remaining-life predictions.
- Ownership, because someone must believe in the data and take action on it. Without a named owner empowered to act, the process produces documentation, not a maintenance plan.
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“NDT data becomes valuable when it changes or provides a decision, or changes a situation. If we just do NDT and keep the data in the files, it does not mean we are using it effectively for a predictive maintenance plan. We just have documentation, not plans.”
Where Predictive Maintenance Programmes Break Down
The moderator raised an issue familiar to almost every inspection professional: when things are running, operations teams are reluctant to interrupt production, and NDT, typically sitting under quality, can find itself in tension with operational priorities.
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“This is exactly the issue. Predictive maintenance always fails at this point of hand-over between inspection and action. You did the inspection, your role is over, and that is where it stops. The point is that you have to action this to get a reliable predictive maintenance plan.”
Who Owns the Decision?
The panel drew a clear line on responsibility. The NDT function's role is to deliver accurate data and interpret findings within its own technical limits, declaring an indication acceptable or not acceptable against the applicable standard. From there, the asset integrity department assesses whether the plant can continue running under current conditions, whether re-rating is appropriate, or whether repair or shutdown is required. Where a decision materially affects plant operation, the asset owner or an authorised decision-maker must be involved.
Cost, Risk and Asset Life Extension
One of the strongest promises of predictive maintenance is cost optimisation without compromising safety. The panel shared concrete predictive maintenance examples from petrochemical and offshore operations.
Case Study: High Temperature Hydrogen Attack
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“One of the critical factors which was a big nightmare was high temperature hydrogen attack for some low alloy materials, something that was not predicted when the materials were designed, but which started appearing in the 2010s for equipment established in the 1970s and 80s. We had to go through critical qualifications to see which technique was appropriate for such a damage mechanism, then apply it. After the application, we were able to push the replacement of critical equipment, reactors and pressure vessels, for almost 70% of the replacements, by about 10 to 20 years.”
The Two Ways Cost Bites You
Mr. Dhiraj laid out the economics with unusual clarity. Without a predictive maintenance programme, an organisation faces one of two outcomes. Either you run to failure, in which case you pay for an unplanned shutdown and then pay a premium replacement schedule to get the plant running again. Or you go conservative and replace equipment that is not yet in need of replacement, upgrading metallurgy across the board. In either case the cost eventually lands, and it lands harder than a targeted inspection programme would have.
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“It is a very tough decision to justify the inspection to the management team and operational team, explaining to them that if you spend some amount on this, do not think of it as a cost. It is actually a saving that is going to come back 10 to 15 times, or maybe even 50 times higher, in a three to five year time frame. It is always a tug of war: operations trying to reduce the inspection budget, inspection people trying to expand it. But it is a balance.”
Spend Where the Risk Is
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“To balance between cost and risk we need to make calculated trade-offs. We do the inspection according to the risk, which we call risk-based inspection. Instead of inspecting every small thing, we come up with a plan: what exactly is degrading faster, what can bring more risk, and what would be the consequence of that risk. High-risk components get more intensive NDT and coverage, medium risk gets less, and low-risk components can be deferred to another time with lower frequency. In short: spend the money where the risk actually exists.”
The Next Five Years: AI, IoT and Continuous Monitoring
Asked how the role of NDT within predictive maintenance frameworks will evolve over the next three to five years, the panel identified three converging trends: digitalisation, artificial intelligence, and continuous condition monitoring.
Digitalisation and Digital Twins
Mr. Taufik Muhammad — Chief Operating Officer, Arise Global
“Digitalisation is the thing. Everyone is asking for digital twins, results will be on the cloud all the time, accessible to everyone. And definitely AI as well. But it will take time. It is machine learning, the more you input, the more reliable it becomes. Whatever you can do in a day, AI can do in minutes. You just need to make sure you are using the right tool, a reliable tool, to deliver accurate results.”
IoT Predictive Maintenance and Continuous Monitoring
Mr. Taufik highlighted a trend he is already seeing across the industry: companies are moving toward continuous monitoring wherever it is applicable. IoT predictive maintenance approaches, using permanently installed ultrasonic thickness sensors, humidity sensors and temperature sensors feeding condition data continuously, give asset owners live indication of the most critical areas of an asset rather than a snapshot taken once every few years. He expects this approach to grow substantially in the coming years, particularly for the most critical assets.
AI Will Assist, Not Replace
Mr. Dhiraj — Asset Integrity Leader, Oceaneering
“AI is not going to engulf NDT within the next three or five years. It is going to become an integral part, a tool we use to make our life easier. It is not going to be a dominating factor. We have been developing AI models to aid human interpreters, to ease the initial screening and to give them a second, cold-eye review. But the burden on the Level IIIs and SMEs is going to increase, because the more you use these models, the more the senior technicians need to vet them properly. Give good data, you will get good results. Give bad data, your results are going to be bad.”
The Data Problem Behind AI
Mr. Taufik cautioned against underestimating what AI implementation actually requires in an NDT context. Having a data file containing crack indications does not mean a model will find every crack that may exist. Crack orientation, crack length, base material and joint configuration all vary, and a training dataset must cover those conditions across different materials and configurations before a model can be considered ready for effective deployment. This is fundamentally different from applying AI to general statistical data.
Panel consensus on AI: it will make interpretation faster and more consistent, but it raises rather than lowers the competence bar for Level III professionals and SMEs, who must validate model outputs against the specific geometry, material and damage mechanism in front of them.
Audience Q&A Highlights
Can AI be used to scrutinise good and bad data?
The panel agreed it can. Mr. Taufik explained that there is always a technical reason why data is poor, such as excessive noise in an ultrasonic dataset, or insufficient image quality in a radiograph. An AI system can reasonably be trained to flag that a dataset does not meet the quality threshold required for further processing and analysis, before a human wastes time interpreting it.
Should only Level III personnel analyse this data, or can Level II?
It depends entirely on the data. Some datasets can be interpreted competently at Level II; others cannot be reliably analysed even by a Level III if the acquisition quality is inadequate. The panel was firm on one point: this is highly technical, niche data that requires qualified, certified and experienced NDT personnel. It is not something management or operations staff should be interpreting.
NDT is important for predictive maintenance, but without proper RBI it will not be effective.
The panel strongly agreed. Data quality comes first, but the data must then be integrated into a system, RBI or equivalent, that processes it into assessments. Mr. Dhiraj added that linking the right damage mechanism to the right RBI approach and then selecting the correct NDT method is essential. If you apply radiography where ultrasonics is required for a given mechanism, the inspection will not be fruitful regardless of how well it is executed.
Can you share your experience with High Order Mode Cluster guided waves?
Mr. Dhiraj has used HOMC extensively for corrosion under pipe supports. It is particularly useful because that corrosion is effectively invisible to many techniques, and HOMC delivers good accuracy in sizing the corrosion. At Oceaneering this has been complemented with robotic X-ray systems and, more experimentally, computed tomography systems. Ultrasonics remains the core family of techniques supporting these long-range and pipe support applications.
Key Takeaways
- Predictive maintenance means shifting from acting after failure to predicting failure before it happens, and NDT is the mechanism that makes that prediction possible.
- The core difference between preventive and predictive maintenance is time-based versus condition-based: fixed intervals versus measured degradation rates.
- There is no single best NDT technique. The damage mechanism, material, accessibility and operating conditions determine the right method.
- The most effective field approach is screening first with a wide-coverage technique, then applying quantitative methods at the flagged locations.
- Advanced ultrasonics remains the single most dominant technique for predictive maintenance, since most integrity assessments rest on corrosion or crack growth data.
- Converting NDT data into action requires three things: data quality and consistency, integration into an assessment system, and clear ownership of the resulting decision.
- Predictive maintenance programmes most commonly fail at the hand-over point between inspection and action, not at the inspection itself.
- Risk-based inspection is the mechanism for balancing cost and safety: concentrate NDT spend where consequence and probability of failure are highest.
- Real-world predictive NDT deferred approximately 70% of critical equipment replacements by 10 to 20 years in one petrochemical case.
- AI and IoT predictive maintenance are coming, but they will assist interpretation rather than replace it, and they increase the responsibility placed on Level III professionals.
Frequently Asked Questions
What is predictive maintenance?
Predictive maintenance is a maintenance strategy in which the actual condition of an asset is measured while it remains in service, and intervention is scheduled based on that measured condition rather than on a fixed calendar. As the panel defined it, it is the shift from taking action after a failure to predicting the failure before it happens. In industrial settings, this condition data is generated primarily through non-destructive testing.
What is the difference between preventive and predictive maintenance?
Preventive maintenance is time-based: inspections and interventions occur at fixed intervals regardless of the asset's actual state. Predictive maintenance is condition-based: NDT and monitoring data are used to measure how quickly an asset is degrading, and intervention is timed accordingly. Preventive maintenance tends to over-inspect healthy equipment while potentially missing rapidly degrading components. Predictive maintenance directs effort where degradation is actually occurring.
What does predictive maintenance involve in practice?
It involves selecting the correct NDT technique for the specific damage mechanism present, acquiring reliable and consistent data using qualified procedures and calibrated equipment, interpreting that data against applicable standards, integrating the findings into a risk-based inspection or predictive maintenance software system, and assigning clear ownership for acting on the resulting recommendations.
What is predictive maintenance used to determine?
It is used to determine remaining useful life, corrosion rate, crack growth rate, the probability and consequence of failure for individual components, and the optimal timing for repair, re-rating or replacement, all before an unplanned failure occurs.
Which NDT method is best for predictive maintenance?
There is no universally best method. The correct technique is determined by the damage mechanism, material, geometry, accessibility and operating conditions. That said, advanced ultrasonics, including phased array, TFM and FMC, is the most dominant family of techniques for predictive maintenance, because most integrity assessments depend on corrosion or crack growth data that ultrasonic methods detect and quantify reliably.
How does IoT predictive maintenance fit into NDT?
IoT approaches use permanently installed sensors, such as ultrasonic thickness sensors and temperature and humidity sensors, to feed continuous condition data rather than periodic snapshots. The panel identified continuous monitoring as a fast-growing trend, particularly for the most critical assets where the cost of an unplanned failure is highest.
Ads
