Reliable Measurement, Reliable Process | AML TPR

Introduction

Ask a plant what it lost last year and you get a number. Ask how that number was arrived at and the room goes quiet. That silence is worth millions.

Somebody mentions a spreadsheet. Somebody else mentions that the flow meter on train four has been reading high since the last turnaround. Within two minutes, the conversation has stopped being about the plant and started being about the number.

And there it dies. Nobody acts on a figure they are still arguing about. The loss report gets filed, and the improvement list gets written from opinion instead, usually by whoever in the room is most senior or most recently annoyed.

At Hint, we built AML-TPR (Total Plant Reliability) to close that gap: one system that first makes the number true, and then turns it into a place to send an engineer. The result is not just reliable measurement or a reliable process, but an available plant.

Why Two Kinds of Reliability Matter

In most plants, “reliability” means two different things, and they are managed by different teams, in different tools.

Measurement Reliability: Is the instrument telling the truth? A transmitter that has drifted two percent, or a human error such as a wrong calibration or manual entry, does not alarm. It quietly poisons every calculation downstream until somebody checks.
Process Reliability: Did the plant do what it was designed to do, and where did the shortfall go: into planned work, into something that tripped, into a human error in operation, or into a supplier who did not deliver?
Why They Belong Together: A plant with process reliability but without measurement reliability may have a very confident answer that might be wrong. A plant with measurement reliability but without process reliability may have a well-kept instrument estate, but no clear understanding of its impact on plant performance. The market sells them separately, but they do not work separately.

One System, Two Halves: How AML Fits Together

That is the shape of AML: two halves, each useful on its own, neither worth what the pair is worth together.

Table 1. The two halves of AML

AML-TPR is not a separate tool. It runs inside the AML platform, on the same core as every other AML module: one instrument register, one plant equipment tree, one set of validated readings. Each module starts from numbers that are already true, and everything it learns flows back into the core.

What You Gain: Train-by-Train Visibility

To show how AML-TPR works, we use a simulated LNG plant, built to behave like production. The same approach applies to refineries, upstream facilities, hydrogen plants and (petro)chemical sites – there the split is made per unit instead of per train.

Consider a six-train LNG plant of 9.5 MTPA (million tonnes per annum). It runs at 94.96% of what it could have made. The shortfall is USD 240 million of product at the plant’s own price, of which USD 158 million was unplanned, which is to say nobody chose it. AML-TPR splits that loss by train, and each train again by the unit that caused it.

In the example, AML-TPR surfaces:

71 instruments behind the production figures, every one of them scored
37% of unplanned loss inside the trains, concentrated in one train
$27 million lost in that one train, in that one year

The split is weighted, because one percent of a 1.6 MTPA train and one percent of a 9.5 MTPA site are not the same loss. Split each train again, and the same unit keeps appearing:

Figure 1: Unplanned loss inside each train, by the unit that caused it

Train 4 accounts for 37% of the total, and every dollar of it is liquefaction. Liquefaction appears in five of the six trains and carries 62% of the money.

This insight would normally be difficult to see on Monday morning: not because the data is missing, but because it is distributed across multiple systems.

From a Number to a Repair: The Loss-to-Cause Loop

Knowing which train or unit costs the most is the beginning. What actually failed is the question underneath it, and that is where measurement and process reliability close into a loop. A plant that accurately records the resolution of failures gets faster at finding and fixing them. This loop makes that happen – six steps, and the sixth feeds the first:

Figure 2: Six steps, and the sixth feeds the first

Five of those six steps can run today. The sixth, accumulating history and generating learning, should be developed with use cases from your asset.

Step 5 is the one no supplier can deliver. It requires the plant’s own maintenance and engineering knowledge, investigation and precise input. “Valve replaced” teaches the plant nothing; “Seat erosion on HV-6142, second failure in 14 months” teaches it a lot.

The Hours Nobody Measures

Diagnosis Is the Hidden Cost: A repair has three parts: working out what is wrong, waiting for the part, and doing the work. Every plant measures the last two and manages them hard. Almost none measures the first – which is why the hours spent working out what to fix are the ones still lying on the table.
Recurring Causes Become Fundable Questions: A cause that turns up nine times in fourteen is no longer an isolated repair. It is a question about design, duty or procedure, with a count and a dollar figure behind it.

Beyond Dashboards: A System That Tells You What It Does Not Know

In the LNG example, twenty-four losses are still provisional. A flow measurement can show that production dropped; it cannot show whether the drop was planned. Until somebody decides, the system says so on every screen, in amber, with the count outstanding.

Rather than forcing an uncertain event into a predefined category, AML-TPR keeps it explicitly marked as unresolved until sufficient evidence is available.

A system that cannot say “not yet decided” will eventually say something false with exactly the same confidence.

What It Changes for Greenfield and Brownfield Operators

Figure 3: Greenfield and Brownfield Operators

When to Start

Before the Plant Runs: Specify AML-TPR in the project scope and set it up during engineering, and it is collecting at first gas or first feed. Commissioning and start-up produce the richest failure data a plant will ever have – and it is normally recorded nowhere.
Already Running AML-CORE? Then this is a switch, not a project. The instrument register, plant tree and readings are in place and need extension to the whole plant. AML-TPR is activated on top and configured with units, capacities, turnaround cycle and process flow diagrams.
In Operation, No System Yet: Initiate an AML proof of concept. Knowing what was lost and where to send the first engineer pays before any history exists; the compounding starts with the first record closed properly.

Conclusion

Reliable measurement and reliable process are two halves of the same question: what did the plant lose, and what should we fix first? Managed separately, each gives an answer you cannot fully trust or cannot act on. With AML-CORE and AML-TPR working as one system, the number is true before anyone quotes it, every lost hour points to the part of the plant it came from, and each repair makes the next one shorter.

That is what an available plant looks like, not something you report at the end of the quarter, but something you manage during it.

Thanks to Harold for writing this technical blog about AML-TPR & AML-CORE. Want to know more about Hint’s software team or the developments Harold mentioned? Feel free to get in touch.

Hint Global