Jan-2008

Identifying very small 
equipment leaks

Very small leaks in feed/effluent exchangers can be identified using chemical tracing. This online technique can detect leaks as low as 5.0 ppm, avoiding 
off-specification effluent and reducing the length of costly shutdowns.

Dave Ferguson, Tracerco
Troy Witherill

Article Summary

High sulphur content in the effluent of a hydrotreater stream is a signal that the unit has a problem with either the reactor catalyst bed or the feed/effluent exchangers. Plant resources can often isolate the offending vessel by analysing samples online or pressure testing exchangers offline. These approaches take time and can be expensive when they involve a shutdown. Also, conventional efforts often fail to locate the offending vessel. An online technique using radioisotopes tracers has been used for many years to leak test exchangers, but some leaks are too small to be measured using this technology. A new service using chemical tracers has therefore been employed to find very small leaks.

Identify the problem
A Midwest refinery in the US operates a naphtha hydrotreater with a bank of four feed/effluent exchangers (Figure 1). The flow through the hydrotreater varies from 5000–10 000 bpd. The effluent from the reactor is cooled by the exchangers and then sent as feed to the catalytic reforming unit. Results of the catalytic reformer feed sample analysis showed that sulphur levels, which had been 0.0–2.0 ppm, were now fluctuating between 2.0 and 10 ppm. Figure 2 shows a distinct trend towards higher sulphur levels. Plant personnel had tried to determine which exchanger was leaking by sample analysis, but had been unsuccessful. They contacted Tracerco to gain assistance in identifying the leaking exchanger before taking an outage to fix the problem.

Leak study methods
Tracerco has a long history of performing online heat exchanger leak tests using radioactive materials and helium as tracers. The standard technique is to mount radiation detectors on the feed and effluent lines of each exchanger and inject a radioisotope tracer that is compatible with the feed stream. As the tracer flows through the exchangers with the higher-pressure feed stream, the detectors on the feed piping will respond to the passage of the tracer, producing bell-shaped Gaussian response curves on the computer. If one of the exchangers has a leak, the leak to the effluent side will be at the same concentration of tracer as its percentage in the feed. Although diluted by the effluent, the detectors on the effluent exit lines will respond, producing a response curve on the computer. By comparing the area under the leak detector response curve to the area under the feed detector response curve, the percentage leak can be approximated.

The limitation of this technique is in its sensitivity. It is the easiest and quickest way to perform an online leak test if 0.5% or more of the feed is leaking into the effluent. Sometimes, the sensitivity can be improved to 0.2% or even 0.1%, but the more typical limit is 0.5%.

Smaller leaks can be found with radioisotope tracers by using a sampling technique. The tracer material and injection procedure are the same as before, but instead of mounting detectors on the feed and the effluent lines, sample points on the effluent lines have to be established. Sample points generally consist of a sample cooler, where hot effluent is condensed/cooled and collected in sample containers. The samples are counted with a highly sensitive radiation detector over several minutes. Compared to the online tracer technique, where the tracer material flashes past the detector in a few seconds, much smaller amounts of radiotracer can be measured in the samples through the accumulation of counts over time. This technique has been used to find leaks as low as 100 ppm. Unfortunately, finding smaller leaks than this requires the injection of so much radiotracer that its disposal becomes an issue, as does the safety of the crew performing the leak test. Helium has also been used as a tracer for finding small leaks online, but the leak size cannot be quantified.

New tracer approach
To find leaks smaller than 100 ppm, chemical tracer technologies were researched. A range of compounds were identified, some of which are compatible with organic and others with aqueous streams that are chemically and thermally stable and can be readily detected in samples by specialised gas chromato-graphy. These chemicals can be found in hydrocarbon samples at concentrations as low as 1.0 part per billion (ppb).
In the case of the Midwest refinery, the four feed/effluent exchangers each had a sample point, which meant the tracers could be injected into a single injection point (Figure 1). Each exchanger was tested with a separate tracer. This allowed a two-man crew to test the four exchangers. One tracer could have been used, but all four exchangers would have to have been sampled at the same time, requiring a five-man crew. It was more cost-effective to use two people and four tracers.

Since the effluent stream contained hydrogen sulphide (H2S), the samples were collected in stainless steel sample cylinders instead of using a sample cooler. Hydrogen and H2S would have been continuously released to the atmosphere while collecting the samples with a sample cooler. Ten sample cylinders were mounted to a manifold at the first sample point. As the tracer was injected, the first sample cylinder was opened and closed in ten seconds. The other nine sample cylinders were opened and closed at ten-second intervals. The sample cylinders were then removed from the manifold and immersed in an ice bath to condense the naphtha and tracer. Each cylinder was depressurised through a tube containing activated charcoal, and the naphtha in the cylinder was collected in plastic bottles.

The activated charcoal tubes and the sample bottles were sent to the Tracerco lab for analysis. Three of the sets of samples were negative. The set of samples for the E-1B exchanger showed tracer in samples 5 through 10 (Figure 3).

Unfortunately, the tracer was slower coming through the exchangers than had been expected. The ninth sample had the highest reading and the tenth was the first on the trailing edge of the response curve. If another ten samples had been caught, there would have been tracer in samples 11 through 13, at least, and probably in a few more, as the trailing edge of a response curve is usually more drawn out than the leading edge.

For this analysis, however, the conservative approach was applied and a short trailing edge was extrapolated. The average concentration for the samples was determined and the grams of chemical tracer that passed the sample point while the samples were being collected was calculated. When this result was divided by the total grams of chemical tracer that was injected, the resulting percentage leak was revealed. The results indicated that the rate of leakage of the feed into the effluent was 0.0044% or 44 ppm at the time of the test.

At a convenient time for the refinery, the unit was brought down and the E-1B exchanger was pressure tested with water. One tube had an obvious leak and four other tubes had small leaks. The tubes were plugged and the unit was brought back online. The sulphur levels returned to the 0.0–2.0 ppm level.
    
Conclusion
The plant personnel were able to avoid lost production from a longer shutdown and the maintenance costs of opening and hydrotesting the other three exchangers. This project showed the capabilities of the new approach to leak testing exchangers. ULSD hydrotreaters operate at such low sulphur specifications that very small leaks in the feed/effluent exchangers can result in off-specification effluent. With the new chemical tracers, leaks as low as 5.0 ppm can be detected.