Smart Leak Detection and Repair LDAR for Control of Fugitive Emissions Regulatory Analysis & Scientific Affairs June 2004 Copyright American Petroleum Institute... Smart Leak Detectio
Laboratory Testing of Fiber Laser
Introduction
This report outlines an advanced Leak Detection and Control (LDAR) method designed for refineries and petrochemical plants, emphasizing the Smart LDAR approach that efficiently identifies and repairs the most significant leaking components A 1997 study revealed that over 90% of controllable fugitive emissions originate from just 0.1% of refinery components, indicating that targeting these high “leakers” can significantly enhance environmental performance The integration of emerging optical imaging technologies facilitates the implementation of Smart LDAR by enabling remote sensing and real-time detection of leaks By illuminating areas with infrared light, operators can instantly identify significant leaks, allowing for prompt repairs and optimal resource utilization.
This report is organized in two sections:
• Section I - Studies Investigating an Alternative to Current Method 21, and
• Section II – Laboratory and Field Testing of optical imaging technologies
Section I – Studies Investigating an Alternative to Current Method 21
This Section describes studies and analyses undertaken to investigate alternative approaches to existing Leak Detection and Repair (LDAR) programs.
A Study of Refinery LDAR Data
The current LDAR procedure, known as EPA Reference Method 21, involves using an instrument probe to measure VOC concentrations at the surface of component seals A correlation exists between the mass rate of VOC leakage and the maximum concentration detected by the instrument The EPA and various state agencies have set specific VOC concentration thresholds that define a leak If the measured concentration exceeds these thresholds, which can range from 100 ppmv to 10,000 ppmv based on the component type and regulations, the component must be repaired or replaced to ensure compliance and reduce emissions.
U.S refineries are required to implement Leak
Detection and Repair (LDAR) programs for processes and streams described in the National
Emission Standards for Hazardous Air Pollutants from Petroleum Refineries (40 CFR 63 subpart CC) known commonly as the “Refinery MACT Rule”
(MACT is an acronym for Maximum Available
Control Technology) The procedures outlined by the current rule are labor and resource intensive, and time consuming
In 1997, the American Petroleum Institute (API) conducted a study to provide guidance for conducting
LDAR programs in a more cost-effective manner
The study evaluated data collected under the LDAR program by 7 Los Angeles, California refineries 5 in the South Coast Air Quality Management District (SCAQMD) The data were examined to help
5 Screening data were obtained for ARCO, Chevron, Mobil, Shell, Texaco, Ultramar and Unocal refineries
Reproduced by IHS under license with API
````,`-`-`,,`,,`,`,,` - determine: (1) the design and operational characteristics that influence fugitive emissions and (2) whether a focused LDAR program could be more effective than the current method of monitoring.
Technical Approach
The SCAQMD mandates that refineries conduct quarterly screenings of all accessible components, such as valves and flanges, defining a leak as any measurement equal to or exceeding 1,000 ppmv An extensive API Study evaluated 11.5 million monitoring values from LDAR programs over a span of 5 years The analysis aimed to identify whether specific component designs or applications, including gate versus globe valves, various process units, or differing actuation frequencies, lead to a higher incidence of significant leaks (≥ 10,000 ppmv) or recurring leaks (≥ 1,000 ppmv on multiple occasions within a four-quarter timeframe).
• Repeat Leakers –by quarter, for components leaking 2, 3 or 4 times in the preceding four quarters
(i.e chronic leakers) for “leak” definitions of 500, 1,000, 10,000, 50,000 and 100,000 ppmv
• High Leakers – by quarter for components, screening >10,000 ppmv
• Process-by-Process Variations – average for all quarters, comparing repeat (≥ 1,000 ppmv) and high
(≥ 10,000 ppmv) leakers for valves, connectors, pumps and an aggregate of all components
• Mean Time between Failures – a failure was defined as screening >500 ppmv
Study Results
A study documented in “Analysis of Refinery Screening Data” revealed that 84 percent of estimated refinery emissions originated from high leakers, which represent only 0.13 percent of the total components These high leakers had an average emission rate approximately 1,000 times greater than the overall average for all components Additionally, 9.5 percent of the remaining 16 percent of emissions came from non-leakers, who make up 99 percent of the components, with their emissions estimated using an EPA specified “zero default” value Consequently, high leakers are responsible for about 92 percent of the controllable emissions.
6 The overall percentage of high leakers (screening >10,000) in any of the seven refineries was less than 0.2 percent
Table 2-1 Distribution of Component Count and Estimated Emissions by Screening Range
Figure 2-1 Distribution of Component Count and Estimated Emissions by Screening Range
% of Total Count % of Total Emissions
Only 5.4 percent of all emissions were from repeat leakers The high leakers were found to occur randomly No systematic explanation for their occurrence was apparent
Valves accounted for roughly 68% of the total emissions (Table 2-2) Connectors accounted for about a third of total valve emissions even though there were nearly three times more connectors than valves
Reproduced by IHS under license with API
Table 2-2 Distribution of Emissions by Type of Component
Category, lb./hr Percent of Total
Study Conclusions
The refinery screening study showed that:
1 About 0.13% of components contribute greater than 90 percent of controllable fugitive emissions;
2 This small population of large leaks are random over time, type of component, and process unit; and
3 Typically, 10,000 components have to be screened to find about 10 significant 7 leaks
The Study concluded that a more cost effective LDAR program would be one that emphasizes the location and repair of high leakers.
Optical Imaging Technologies
Two types of optical imagers suitable for detecting hydrocarbon emissions at refineries and chemical plants are:
The Gas Vue is a commercially available CO2 laser imager produced by Laser Imaging Systems (LIS) This advanced instrument has been successfully tested at two chemical plants, demonstrating its effectiveness in various applications Throughout this report, it will be referred to as a CO2 laser imager.
The "fiber" laser imager, developed by Sandia National Laboratory's Lawrence Livermore facility, employs a patented backscatter technique from LIS This innovative instrument features an optical fiber laser amplifier and has been successfully tested at two refineries and a chemical plant.
Lasers are precisely tuned to emit specific wavelengths of infrared light, enabling the detection of particular compounds or compound types The CO₂ laser offers discreet tunability within the 8-10 micron spectral range, while the fiber laser provides continuous tunability in the 3 micron spectral region.
Table 3-1 summarizes some of the primary features of the optical imaging units
Table 3-1 Features of the CO 2 – and Fiber Lasers
Features Fiber Laser CO 2 Laser
Developer/Manufacturer Developed by Sandia National
Manufactured and marketed under the brand name Gas Vue® by Laser Imaging Systems
Laser Source Lithium Niobate fiber amplified laser (3-
Plant Chemicals Applicable for aliphatic and olefinic hydrocarbons Applicable for olefinic hydrocarbons
Chemical Plants Technology tested at chemical plants and refineries Technology tested at ethylene facilities
Commercial Availability Prototype Discussions regarding commercialization underway Available commercially Used to detect
SF6 leaks at electric power plants
Backscatter Absorption Gas Imaging (BAGI)
The principle of operation of the CO 2 and fiber lasers is Backscatter Absorption Gas Imaging (BAGI) In
BAGI technology generates video images by illuminating a designated area with infrared laser light A camera sensitive to this light captures the backscattered laser reflections When the selected laser wavelength is significantly absorbed by a specific gas, it creates a dark image that reveals the presence of that gas LIS holds a patent for the BAGI principle (U.S Patent #4,555,627).
Figure 3-1 Schematic Description of BAGI Process
A video camera-type scanner both sends out the laser beam and picks up the backscattered infrared light
The camera transforms backscattered infrared light into an electronic signal, presenting a real-time black and white image on the viewfinder and video monitor This image remains consistent regardless of whether scanning occurs in daylight or at night, as the scanner is solely responsive to illumination from the infrared light source, unaffected by sunlight.
The imager offers the capability to toggle between visible and infrared views, which is crucial for operators to distinguish between steam and gas plumes Additionally, it allows for video recording in both visible and infrared light.
2 and 3-3 show leaking components detected by the CO 2 and fiber lasers in visible light and infrared In
Source: As Adapted from McRae, Tom, GasVue: A Rapid Leak Location Technology or Large VOC Fugitive Emissions (Presentation at the CSI Petroleum Refining Sector Equipment Leaks Group, Washington, DC, Sept 9,
The gas plume does not need to be in direct contact with the background material; it only needs to be positioned between the background and the infrared camera for effective detection.
Reproduced by IHS under license with API
Current advancements in BAGI technology enable the visualization of vapor clouds from specific hydrocarbons and chemical leaks that are otherwise invisible However, this technology has yet to quantify the mass emission rate of these leak clouds A comprehensive technical overview of BAGI technology is available.
[3] “Backscatter Absorption Gas Imaging: a New Technique for Gas Visualization.”
Figure 3-2 CO 2 Laser Views of a Leaking Connector in Visible and Infrared Light ice ethylene leak tag connector
Figure 3-3 Fiber laser Views of a Leaking Flange in Visible and Infrared Light Visible light view of leaking flange Infrared view of leaking flange hydrocarbon plume flange flange
Three parameters that influence the performance of the two BAGI lasers are background, laser wavelength and atmospheric window
To effectively detect a gas leak, a reflective surface must be present behind the leak, as visualizing a gas plume against the sky or a distant background is not feasible In many cases, leaks can be imaged against the component itself, with distant backgrounds appearing dark Operators can identify when the imager is beyond detection range by noting the absence of images of the inspected components; the farther the component or background, the darker the image becomes Additionally, operators can switch between infrared and visible light viewing to assess whether there is a suitable background surface behind the leak point being inspected.
The sensitivity of gas leak detection through optical imaging is significantly influenced by the alignment of the laser wavelength with the absorption wavelengths of the target gas For effective imaging of a hydrocarbon gas cloud, it is essential that the gas can absorb the specific laser wavelength used.
Detection sensitivity is influenced by factors such as wind speed, optical resolution, gas plume motion, and viewing angle Higher wind speeds lead to faster dispersion of gas from the leak source, making it less visible to optical imagers However, some movement of the gas plume aids in leak detection, as the human eye is more attuned to detecting motion In contrast, a stationary gas cloud is challenging to identify against a varied background.
An atmospheric window refers to a spectrum region with minimal light absorption by common atmospheric gases such as oxygen, nitrogen, carbon dioxide, and water vapor The primary atmospheric windows in the infrared spectrum are located between 3 to 4.2 microns and 8 to 13 microns Laser beams traveling through the atmosphere at these wavelengths encounter minimal attenuation, making these regions significant for various applications.
Laser light in IR windows can be attenuated by airborne particulates, such as water droplets found in fog and steam These particulates manifest as dark clouds in BAGI images, similar to fugitive gases that absorb laser light However, unlike fugitive gases, these particulates are visible to the naked eye, allowing BAGI operators to easily differentiate between the two types of cloud displays.
If a cloud is visible to the naked eye as well as through the BAGI camera, it is likely composed of particulates or steam Conversely, if the cloud is only detectable with the BAGI camera, it indicates the presence of hydrocarbon vapor.
Description and Operations of the CO 2 and Fiber Lasers
CO2 and fiber lasers, weighing between 9 kg and 14 kg and comparable in size to a TV camera, consist of a camera unit and a power/control unit The CO2 laser's camera connects to a mobile power pack that converts 110V AC to DC, while Sandia National Lab's fiber laser camera is linked to a backpack-mounted power/control unit powered by a 28V lithium-ion battery The fiber laser operates for 1 to 1.5 hours on a single charge, with a twelve-hour recharge time, and allows for battery replacement without device shutdown Additionally, SNL’s Imager can be powered via a 110V AC outlet using the power unit’s 28-volt DC converter.
Both laser cameras can be used in either a shoulder-mounted or tripod-mounted position, featuring a swivel fitting for easy movement while scanning for leaks A zoom lens enables the operator to adjust the focal distance for an improved view The cameras allow switching between infrared and visible light views, with the capability to record video from both perspectives They also have straightforward startup procedures, requiring only 10 to 15 minutes after being powered on.
Reproduced by IHS under license with API
Figure 3-4 Laser Imagers shown Shoulder- and Tripod-Mounted Operations
3.2.1 Description of the CO 2 Laser Components
The CO₂ wave guided laser operates at 5W in the 9-11 micron spectral region, utilizing a unique optical arrangement that allows for synchronous scanning of the laser beam and the instantaneous field-of-view (IFOV) of an infrared detector This system employs a small cooled IR detector and a collimating lens, with two scan mirrors facilitating raster-like motion across the target area, ensuring perfect synchronization between the detector IFOV and the laser beam In long-range applications exceeding 10 meters, a beam expander is implemented to minimize laser beam divergence, thereby reducing power requirements.
See “Evaluation of the GRI Gas Imaging Leak Survey
System,” [4], and “GasVue VOC and SF6 Leak
Location Field Test Results,” [5] for additional details about the CO2 laser
Sandia National Laboratories (SNL) has created an innovative infrared laser source designed to effectively detect VOC emissions in industrial facilities This advanced device is an optical parametric oscillator (OPO) that utilizes a periodically-poled lithium niobate (PPLN) crystal as its active medium, functioning similarly to a laser.
Figure 3-5 CO 2 Laser’s Synchro-Scan
The optical parametric oscillator (OPO) operates within an optical cavity that houses a nonlinear crystal, specifically periodically poled lithium niobate (PPLN) A separate pump laser beam is focused into the PPLN crystal, where it is transformed into two new beams known as the signal and idler, with their frequencies summing to that of the pump laser The essential components of the PPLN-based imaging system include the pump laser, the OPO, and a raster scanner, as illustrated in Figure 3-6.
Figure 3-6 The basic elements of the PPLN-based imager
Pump laser OPO Scanned imager
(1) The pump laser creates the initial laser radiation
(2) The OPO converts the pump light to the infrared
(3) The scanned imager creates the laser-illuminated image
The current prototype SNL fiber laser was preceded by several generations including a van-mounted and table-mounted miniature version.
Determining Equivalent Leak Definitions for Alternative Work Practices to Method 21
Current U.S EPA regulations do not allow the use of optical imagers or the Smart LDAR concept for controlling fugitive emissions However, stakeholders can petition the Agency to recognize alternative controls or practices that offer equal or better environmental protection To facilitate this, the U.S EPA has established a demonstration protocol that outlines an approval process involving laboratory testing, field testing, and mathematical analysis This protocol aims to quantify the performance of alternative technologies and assess their ability to achieve equivalent fugitive emissions control compared to Method 21 monitoring.
The U.S EPA has developed a Monte Carlo simulation software using SAS to evaluate new technologies or practices proposed as alternatives in Leak Detection and Repair (LDAR) programs This software conducts random statistical simulations to compare the current Method 21 work practice with alternative control technologies, calculating predicted emission reductions for identified leaky equipment components The environmental benefit of using either technology is quantified, demonstrating equivalence when the emission reductions from the alternative technology meet or exceed those of the current practice.
8 Washington DC, United States Environmental Protection Agency, Code of Federal Regulations 1990b: Title 40, Part 60 Subpart GGG,
“Standards of Performance for Equipment Leaks of VOC in Petroleum Refineries,” Government Printing Office
Reproduced by IHS under license with API
API undertook a Monte Carlo Analysis for valves to determine a leak definition for an Alternative Work Practice (e.g optical imaging) that would result in equivalent environmental protection as Method 21 monitoring
The Monte Carlo analysis established an equivalent mass leak definition for the Alternative Work Practice (AWP) based on the Method 21 leak definition in ppmv This equivalence allows the AWP mass leak rates to be relevant across various monitoring technologies, including optical imaging devices, independent of the current Method 21 standards.
Summaries of the technical approach, results and conclusions are presented below Detailed discussions are presented in
“Equivalent Leak Levels & Monitoring Frequencies for Smart LDAR” [6]
Monte Carlo Analysis Focused on Mass
Current LDAR monitoring technologies report screening values in ppmv, but field bagging studies indicate that components with high ppmv values can have low mass leak rates A "Smart" LDAR approach prioritizes identifying significant mass leakers, leading to a Monte Carlo analysis aimed at correlating mass leak rates (Kg/hr) with the control levels established by existing ppmv leak definitions.
Technical Approach to Monte Carlo Simulations
One thousand simulations were performed using actual valve fugitive emissions data from an
“uncontrolled” plant for each combination of optical imaging and Method 21 leak definitions and monitoring frequencies Five monitoring frequencies were simulated for evaluating the control effectiveness of optical imaging technology:
• Bi-monthly (once every two months)
• Semi-Quarterly (twice per quarter)
• Monthly (three times per quarter)
• Semi-Monthly (six times per quarter)
The study determined equivalent mass leak rate definitions (g/hr) for optical imaging (or other AWP) using these monitoring frequencies and achieving equivalent emissions control for three typical Method
21 leak definitions used in the established state and federal regulatory LDAR programs that require quarterly monitoring:
The Monte Carlo approach evaluates emission reductions by comparing two leak detection methods Specifically, it assesses the effectiveness of the optical imaging method (AWP) against the emission reductions achieved by the U.S EPA Reference Method 21, which is currently mandated.
Results and Conclusions
Monte Carlo analyses revealed that the mass emission rate corresponding to a specific optical imaging or other AWP monitoring frequency, aligned with a defined Method 21 leak standard, was notably accurate and clear.
A recent study indicates that optical imaging and other advanced monitoring techniques, when employed bi-monthly, offer superior environmental protection for valves compared to the existing Method 21, which monitors quarterly for leak definitions of 500 ppmv, 1,000 ppmv, and 10,000 ppmv Additionally, Table 4-1 presents the corresponding AWP leak definitions across five different monitoring frequencies in relation to three CWP leak definitions under quarterly monitoring.
Table 4-1 AWP Leak Definitions at Different Monitoring Frequencies for Valves Equivalent to
Three Method 21 Leak Definitions at Quarterly Monitoring
Equivalent AWP Leak Definition for Specified Monitoring
(once per quarter) Bi-Monthly
500 ppmv 0.00023kg/hr 0.060 kg/hr 0.085kg/hr 0.10 kg/hr 0.17 kg/hr
1,000 ppmv 0.00041 kg/hr 0.061 kg/hr 0.085 kg/hr 0.11 kg/hr 0.17 kg/hr
10,000 ppmv 0.0049 kg/hr 0.069 kg/hr 0.090 kg/hr 0.13 kg/hr 0.18 kg/hr
Alternative Work Practice and Smart LDAR overcome Variability in Method 21
Equivalent total emissions reductions when using the proposed AWP are achieved by identifying all of the highest rate leakers This is illustrated in Figures 5-1 and 5-2
Figure 5-1 illustrates a significant scatter in the data correlating screening values with mass emission rates While the screening value indicates the leak rate in concentration (ppmv), the mass emission rate (kg/hr) provides a direct measurement of the leak There is considerable variability between the Method 21 screening rate and the actual mass emission rate, with the Method 21 screening value potentially varying across several orders of magnitude for any given mass emission rate Consequently, this variability raises the possibility of both false positives and false negatives when utilizing Method 21 to detect leaking components.
Method 21 measurements can lead to false positives, resulting in unnecessary repairs for components that, while flagged as leaking, have mass emission rates below the defined threshold Consequently, these repairs often yield minimal, if any, emission credits.
False negatives from Method 21 can lead to significant emissions, as some components may exceed the mass emission rate requiring repair without being detected by Method 21 screening Consequently, these undetected leaks persist under a Method 21 program In contrast, optical imaging technology, such as AWP, effectively identifies these leakers, highlighting a key advantage of this method in reducing emissions that Method 21 overlooks The Smart LDAR approach utilizing optical imaging enables the detection and repair of missed leaks, allowing for a higher mass leak definition compared to Method 21.
Reproduced by IHS under license with API
Figure 5-1: Variability of Method 21 Results for Equivalent Mass Emission Rates
A box plot visually represents the distribution of data by enclosing the middle 50% of values between the 25th and 75th percentiles It features a horizontal line at the median (50th percentile) within the box, while vertical lines, known as whiskers, extend to the 5th percentile below and the 95th percentile above Additionally, dots are used to indicate values that fall below the 5th percentile or exceed the 95th percentile.
Source: Monte Carlo Simulation Evaluation of Gas Imaging Technology
This illustration presents box plots from the 1993-94 Petroleum Industry bagging dataset, sourced from the American Petroleum Institute It highlights the reported screening value ranges (ppmv) for various levels of mass emission rates, categorized by mass magnitude bins For instance, a mass emission rate magnitude of "1E-8" refers to rates measured in units of 10^{-8} kg/hr, encompassing values of 1E-9 kg/hr or greater, but less than 1E-8 kg/hr.
8 kg/hr, because the integer portion of the base-10 logarithm for values between these bounds is
“-8”). the current fugitive emission control program required by U.S.EPA regulation, monthly or quarterly
Monitoring for leaks is essential, with repair thresholds set at 500, 1,000, or 10,000 ppmv Less frequent monitoring is permissible if the percentage of leaking components stays below a designated level over a specified number of periods Some U.S state regulations impose even lower leak definitions, contradicting findings from the API study illustrated in Figure 2-1 It is important to note that lower leak definitions for repairs do not necessarily enhance emissions control, as reducing the leak level results in only a few additional components being classified for repair, which minimally impacts overall mass emissions.
The Smart LDAR approach prioritizes the identification and repair of significant leakers, which are responsible for the majority of mass emissions By swiftly locating and addressing these large leaks, the method effectively offsets emissions from components with lower ppmv readings that may leak for extended periods Utilizing optical imaging technology offers a cost-effective solution for quickly detecting high leakers and addressing false negatives from the Method 21 approach, leading to substantial emissions control benefits with the Smart LDAR technique.
Figure 5-2 Equivalence From Quicker Repair of Highest
The diagram demonstrates that Smart LDAR identifies large leaks more quickly on average, while smaller leaks take longer to detect Although both methods achieve the same total reduction in leaks, Smart LDAR is more cost-effective.
Section II – Laboratory and Field Testing of Optical Imaging Technologies
Numerous studies have evaluated the effectiveness of CO2 and fiber lasers in detecting fugitive emissions, including two conducted at refineries, two at ethylene production facilities, and two in controlled laboratory settings, along with several additional demonstrations This section provides a summary of the objectives and results from these laboratory and field tests.
Refinery Demonstration of a Van-Mounted Fiber Laser
In April 1999, a demonstration of a van-mounted fiber laser prototype, developed by Laser Imaging Systems and Sandia National Laboratory, was conducted by the U.S EPA, API, SNL, and LIS at a Texas refinery The four-day event aimed to showcase the capabilities and effectiveness of the technology.
• Demonstrate that the fiber laser could work reliably for an extended period of time in a refinery setting, and
• Determine whether this technology performed well enough compared with existing Method 21 devices to warrant continued development.
Methodology
Two teams, working independently, collected data from seven process areas (see Text Box)
• The Method 21 Team primarily monitored tagged components subject to current LDAR programs
Reproduced by IHS under license with API
The population of components selected for Method 21 monitoring was defined by two parameters: Process Areas Monitored
1 Components had to be within reach of the OVA operator standing on the ground, and
2 Components had to be within the line-of-sight and range of the minivan carrying the fiber laser
A small percentage of components in each of the seven process areas were monitored, and the team observed the time taken to complete this monitoring.
The Fiber Laser Team effectively monitored all components within the laser's line of sight, promptly quantifying any detected leaks in ppmv using a TVA detector and assigning identification numbers as needed This advanced technology allows for the detection of leaks from various sources, regardless of specific LDAR regulatory requirements Consequently, the monitored components encompassed those tracked by the Method 21 Team, along with additional components within the laser's range The Coordination Team, overseeing the demonstration's design and management, estimated the total number of components monitored, while both infrared and visible-light video footage of the monitored areas was recorded.
Findings & Conclusions
The demonstration showed that the van-mounted fiber laser could successfully detect fugitive emissions The Test Team reached three conclusions from analysis of the collected data:
• The fiber laser could be used to locate high leaking components in a refinery setting
Most VOC emissions at refineries are attributed to a small number of components that experience significant leaks Utilizing optical imaging technology can enhance the detection of these leaking components more effectively.
• The demonstration results justified continued development of an operator-portable prototype of the fiber laser.
Laboratory Testing of Primary Components of an Operator-Portable Fiber Laser
In 2000, Sandia National Laboratories (SNL) performed a laboratory test on the unassembled components of an operator-portable fiber laser, utilizing the OPO from a van-mounted fiber laser in conjunction with a newly designed scanner by LIS.
The objective of the test was to assess how various factors, including the source of the leak, viewing distance, wind speed, reflective background, type of effluent gas, and mass flow rate, affect the leak detection threshold of fiber lasers in a laboratory setting.
Test Methodology
Background material Sandpaper Mass flow rate 1 0.2*LDL to 5*LDL, continuously varied m = meters LDL = lower detection limit
A leak occurred from a valve assembly located inside a wind tunnel during testing The fiber laser was positioned on a cart in front of the wind tunnel's view port, allowing the operator to have a clear line of sight to the generated plumes The images captured through the viewfinder were recorded on videotape.
The test parameters and matrix are shown in Tables
A run was performed by gradually increasing the mass flow until the plume became visible Continuous video was recorded across a range of mass flows from one-fifth to five times the lower detection limit (LDL), as established by the operators during the recording process.
Table 7-2 Operator-Portable Fiber Laser Laboratory Test Matrix
Fifteen baseline tests and six parameter variation tests were conducted, with each test involving continuous gas flow adjustments from 0.2*LDL to 5*LDL A panel of five gas imaging video observers analyzed the data, with each reviewing a third of the video segments—five for baseline runs and two for parameter variations The mass flow rate at which gas flow became visible for each observer was documented, allowing for the calculation of mean leak detection levels (LDL) and variability across all runs and observers.
Test Results and Analysis
Under specific test conditions, the gas flow was first calibrated to determine the lowest detectable level (LDL) as perceived by the test operators Subsequently, efforts were made to capture imagery at the corresponding mass flow rate.
Reproduced by IHS under license with API
````,`-`-`,,`,,`,`,,` - rates spanning the range between 0.2*LDL and 5*LDL The recorded data were assessed in two different ways:
• First, a single observer viewed the tapes and determined the point at which the visibility threshold occurred
• Second, a panel of observers viewed the data and indicated the intensity of leak that was observed (if any)
The primary results indicate that the detection limit tends to be higher when analyzing tapes rather than observing through the instrument in real time.
Table 7-3 summarizes the single-observer results, presenting LDL values as mass flow in grams per hour (g/hr) alongside the corresponding screening value (SV) in parts per million by volume (ppmv), calculated using EPA correlation equations The flange correlation equation was applied to assess the bonnet and flange leak.
The mass flow LDL values are presented in two contexts: at the leak point and within the downstream plume area In windy conditions, the LDLs are higher in the downstream area due to the wind's ability to quickly disperse the plume Conversely, the leak point thresholds are slightly lower since the wind has a reduced impact at this location Overall, detection limits ranged from 2.2 to 3 g/hr for a wind speed of 1 meter per second (m/s) across all viewing distances against sandpaper.
Wind speed significantly influenced the results of the single observer, with LDL values at both the leak point and plume area showing a marked increase as wind speed rose from 2.5 m/s to 10 m/s, while maintaining constant baseline conditions.
The results were significantly influenced by the background, as the LDL increased to 6 g/hr against a Styrofoam backdrop while maintaining constant baseline conditions The brightness of the laser return was so intense that it overwhelmed the camera, necessitating a reduction in laser power to between 20-60 mW.
LDL (grams/hr) at Leak Point
LDL (grams/hr) in Plume Area
Baseline Bonnet 3 Propane 1 Sandpaper* 2.8 Same 9,100
3 Bonnet 6.1 Propane 1 Sandpaper* 3.0 Same 10,050 3b Bonnet 7.6 Propane 1 Sandpaper* 3.0 Same 10,050
4 Bonnet 9.1 Propane 1 Sandpaper* 3.0 Same 10,050 9a Bonnet 3 Propane 2.5 Sandpaper* 3.0 4.0 10,050
*Laser power: 500 mW ** Laser power: 20 to 60 mW *** Screening Value
In the "panel of observers" test, five viewers with varying familiarity with gas imaging evaluated data after being trained on images of nonleakers, borderline leakers, and high leakers They then assessed randomly ordered image segments from a tape, which included different gas flow rates, background materials, and leak locations Each viewer graded the leak intensity on a scale from none to high (0 to 3) The segments reviewed were a selected subset from the total dataset, featuring the highest and lowest leaks, along with four near the transition region.
The lowest flow rates visible to the panel of viewers are listed in Table 7-5 The corresponding test conditions and the number of observers who detected the leak are presented
Table 7-5 Lowest Leak Detection Thresholds Detected by Panel of Observers
Lowest Detected Flow at Leak Point (grams/hr)
Number of Panelist who could See Leak
Reproduced by IHS under license with API
The results from the panel differed from those of individual observers, particularly at higher wind speeds At a wind speed of 1 m/s, both methods showed similar visibility onset values for sandpaper and styrofoam, approximately 2-3 g/hr and 4.8 vs 6, respectively However, as wind speeds increased, the panel's detection threshold became lower than that of the individual observer For instance, at 2.5 m/s, the panel recorded 1.5 g/hr compared to ~3 g/hr for the individual, and at 7.5 m/s, the values were 3 g/hr versus 13 g/hr This discrepancy is attributed to the individual observer being more conservative in defining the threshold, focusing on visibly obvious leaks, while the panel was instructed to assess all leaks and rank their intensity.
Laboratory Tests of SNL’s Portable Fiber Laser
In 2001/2002, Sandia National Laboratory conducted tests of its prototype portable fiber laser at its Lawrence Livermore facility The objectives of the laboratory tests were to:
To enhance the performance of portable fiber lasers, it is essential to analyze data that establishes relationships among various operating variables Key factors include operator influence, mass leak rate, wind speed, the type of reflective background, the reflectivity of that background, and the viewing distance from the leak Understanding these variables will enable the optimization of laser performance in different conditions.
The study aims to evaluate the effectiveness of the prototype fiber laser BAGI system in detecting hydrocarbon leaks within a controlled environment, utilizing simulated industrial components This assessment will focus on whether the system can identify leak rates that align with the defined values for alternative work practices, as established through Monte Carlo analysis.
To enhance the Leak Detection and Repair (LDAR) program, it is essential to collect comprehensive data and establish a statistical correlation between operating variables and the leak detection threshold This analysis will enable the EPA-OAQPS to formulate alternative work practice guidelines that incorporate optical imaging technology.
Test Methodology
Propane was used as the leak gas in the test, which was conducted in two phases:
In Phase 1, a controlled test was conducted using a wind tunnel to assess a leaking component, with an imager mounted on a cart for optimal viewing The study aimed to determine the mass rate detection threshold based on specific variables, including wind speed, viewing distance, and background reflectivity The results were instrumental in establishing a correlation between these factors, as detailed in Table 8-1.
Table 8-1 Wind Tunnel Test Variables
*The component itself acts as the reflective background
Phase 2, known as "The Roving Test," took place outdoors, featuring four leak stations equipped with various test components positioned in distinct locations and backgrounds A system operator utilized an imager to inspect the components during the test, which was conducted in a "blind" manner.
````,`-`-`,,`,,`,`,,` - fashion where the system operator had no knowledge of which component was leaking The test was repeated under different conditions, as the component that was leaking was varied
The roving test was based on wind tunnel testing results, with the Test Team monitoring wind speeds between 0 and 10 meters/s Once the appropriate wind conditions were established, tests were conducted to determine the mass detection threshold for specific wind speeds and distances (3 or 6.1 meters), which were expected to align with Phase I wind tunnel findings For instance, if an average wind speed of 1.5 m/s was recorded, the Phase I data indicated a mass detection threshold of 5 g/hr at a distance of 3 m, leading to the setup of a 5 g/hr leak at one of the test stations during the roving tests.
The laser operator began detecting a leak from a distance of 9.1 meters, gradually moving closer to identify the source If the operator mistakenly identified a station due to wind affecting the leak plume, it was recorded as a false positive, prompting further search for the correct station Upon successful identification of the leaking component, the threshold viewing distance and wind speed were documented Detailed information about the roving test stations can be found in Table 8-2, while Figures 8-1 and 8-2 provide illustrations and photos of the Roving Test setup.
Table 8-2 Roving Test Stations Station Description of Roving Test Stations
Station 1 Component mounted at eye level (1.2 meters) with sandpaper background
Station 2 Component mounted at eye level (1.2 meters) with painted curved sheet metal background (primed and painted with Rust-O-leum flat grey spray paint)
Station 3 Component 0.51 meters above the ground, no background At this height, the pavement under the component can serve as a background in a similar way as a concrete pad or a paved refinery process area would serve as a background
Station 4 Component 2 meters off of the ground with open sky in the background Since
Station 4 is above eye level, the component itself is the reflecting background
Reproduced by IHS under license with API
Figure 8-1 Roving Test Set-up to flow meters
Curved metal background, valve 1.2 meters high
Sandpaper background, valve 1.2 meters high No background, Component is reflective background
Operator 1.5, 3 or 6 meters from the leak point
Not shown: flow meter control panel that test administrator will use to control leaks
No background, Pavement is reflective surface
Figure 8-2 Test Stations for Roving Test
No Background/Below Eye Level
Laboratory Test Results
The mass rate detection threshold rises with greater distance and windspeed under a specific background Figure 8-3 illustrates the detection thresholds across all test conditions, revealing that, except for 10 cases, the mass rate detection threshold remained below 10 g/hr Notably, 73% of the instances reflected this trend.
(73%) of the mass detection thresholds determined in the lab tests were below 5 g/hr.
Statistical Analyses of Test Data
A statistical analysis of the laboratory test data was conducted to determine an overall model to predict results using the optical imager
8.3.1 Observed and predicted detection thresholds
The overall model fitted to the wind tunnel test data for a curved metal background reveals a linear relationship on a semi-log curve between leak detection threshold and windspeed, as illustrated in Figure 8-4 This curve predicts the leak values that can be detected 95% of the time under specified conditions, indicating that at a distance of 3 meters with a windspeed of 1 meter/s, the model provides significant insights into leak detection capabilities.
A leak detection threshold that ensures a 95% detection rate is approximately 5 g/hr, equivalent to a speed of about 2.2 miles per hour Additionally, a leak of 60 g/hr can be reliably detected from a distance of 3 meters when the wind speed is 10 m/s (22 mph).
Reproduced by IHS under license with API
Figure 8-3 Mass Rate Detection Thresholds for Wind Tunnel & Roving Tests Wi nd Tunnel Thr eshol ds and Rov ing Test Resul ts 0.2 0.35 0.35 0.43 0.48 0.6 0.7 0 7 0.75 1.05 1.17 1.2 1 2 1.25 1.3 1.4 1 4 1.6 1 6 1.7 1.75 1.75 1.75 1.75 2 222 2.25 3.2 3.25 3.35 3.35 4.3 4.5 4.75 9.2 9 2 10 11 17 21
160 sa ndp ap er, 3m, 0m/ s ndp sa er, ap 1m/ 3m, s rved me cu
, 3m, tal s 5m/ ndp sa er, ap 1m, 6.
0m/ s ing pa rov d gr ve nd ou , 3m 6m , 1 /s ing sa rov pa nd r, 3 pe
1m/ s ing sa rov pa nd r, 3 pe
79 m, m/s no ba grou ck , 3m, nd
0m/ s ndp sa er, ap 3m, 10m /s ndp sa er, ap 1m, 9.
0m/ s no ba grou ck , 6 nd 0m 1m, /s rved me cu
, 3m, tal m/s 10 no ba grou ck , 3m, nd
1m/ s ing rov y, 6 sk 4 1m, s 4m/ rved me cu
, 9 tal , 0m 1m /s no ba grou ck , 9 nd 0m 1m, /s ing, rov rved cu tal, 3 me
2.5 m, m/s ndp sa er, ap 3m, s 5m/ no ba grou ck , 3m, nd s 5m/ mi roa , sk ng 4.6m y,
, 2m /s ing, rov nd sa pe pa r, 3m 2m , 3.
, 6 tal , 0m 1m /s no ba grou ck , 6 nd 1m 1m, /s ing, rov y, 3m, 2 sk 1m/s ing rov av , p rou ed g , 3m, nd
55 1. m/s no ba grou ck , 9 nd 1m 1m, /s no ba grou ck , 9 nd 5m 1m, /s ing, rov rved cu tal, 3 me
, 6 tal , 1m 1m /s ing, rov rved cu tal, 3 me
53 m, m/s ndp sa er, ap 1m, 9. s 1m/ no ba grou ck , 6 nd 5m 1m, /s ing rov av , p ed g nd rou , 3m, 4m/ 2. s ndp sa er, ap 1m, 6.
5m/ s ing rov an , s pe dpa 1m r, 6 8m , 2.
, 30 tal ft, 1 m/s ing, rov y, 3m sk
/s ndp sa er, ap 1m, 6. m/s 10 ing rov av , p ed g nd rou 6m, , 7.
1m/ 2. s ndp sa er, ap 1m, 9. s 5m/ ndp sa er, ap
, 6 tal , 5m 1m /s ing, rov rved cu tal, 9 me
The statistical analysis assessed the likelihood of identifying a 60 g/hr leak based on factors such as distance, wind speed, and background conditions These findings are essential for evaluating the device's performance across different scenarios.
Figure 8-5, is for curved metal background at a very high wind speed of
10 meters/s Assume, for example, that the desired detection probability for curved metal background at a wind speed of 10 meters/s is 0.75 The dashed line at 0.75 intersects the probability curve,
The model estimates the maximum detection distance at 4.9 meters for a rate of 60 g/hr, while it intersects the lower bound curve at approximately 3 meters To ensure a 95% confidence level in detection probability, a more conservative distance of 3 meters is recommended.
0.75 Using the lower bound (3 m) rather than the point estimate (4.9 m) accounts for the uncertainties in the estimated model coefficients
9 Pr(detect 60 g/hr) = Φ({log(60) – [a1 × S + a2 × D×S + a3 × C + a4 × D×C + a5 × N + a6 × D×N + a7 × D×W + a8 × W×S+ a9 × W×C + a10 × W×N + a11 × P + a12 × D×P + a13 × W×P]} / sigma), where Φ is the standard normal cumulative distribution function
Reproduced by IHS under license with API
8.2.3 Predicted distance or wind speed to detect a 60 g/hr leak
The statistical model also estimates the distance at which a 60 g/hr leak 10 can be detected 95 % of the time, for a given wind speed
The maximum distance and wind speed values were determined by applying the equation for Pr(detect 60 g/hr) and adjusting the distance or wind speed to achieve a 95% probability.
Pr(detect 60 g/hr)) The equation can be solved uniquely because the model is linear and increasing in both wind speed and distance
To achieve a detection probability of 0.95 for curved metal at a wind speed of 5 meters/s (11 mph), the maximum detection distance is approximately 6.1 meters, while the lower bound detection distance is around 5.2 meters.
Therefore, the model’s best estimate of the maximum allowed distance is 6.1 m, but a more conservative value of 5.2 m would ensure with 95 % confidence that the detection probability is at least
0.95 Using the lower bound (5.2 m) rather than the point estimate
(6.1 m) accounts for the uncertainties in the estimated model coefficients
10 The Monte Carlo analysis equivalent leak rate
Refinery Test of Portable Fiber Laser
A field study at a refinery assessed the effectiveness of a prototype portable fiber laser for detecting fugitive emissions during normal operations The study aimed to evaluate the performance of this innovative technology in real-world conditions.
• Demonstrate that the prototype portable BAGI device detects and successfully images fugitive emissions
• Gather data that can be used to establish the mass-emission detection capabilities of the gas imaging technology
To assess the sensitivity of BAGI technology in refinery operations, it is essential to collect data on various influencing factors These factors encompass the distance from the scanned component, sight lines and angle-of-view, as well as the infrared backscatter and absorption characteristics of surrounding materials Additionally, weather conditions and the chemical composition of emissions must be considered to ensure accurate performance evaluation.
Study Methodology
Four process areas were monitored:
The mass rates of several leaks detected by the fiber laser were determined using bagging techniques
The distance from the fiber laser to the leaking component was measured, along with the local wind speed near the leak Below is a concise summary of the conclusions and findings.
Study Conclusions, Data Analysis and Results
The research demonstrated that the prototype fiber laser effectively identifies leaks of olefinic and aliphatic hydrocarbons from both LDAR and non-LDAR components, achieving detection rates of approximately 20 g/hr of total hydrocarbons under standard refinery operating conditions, particularly against common reflective surfaces in refineries.
The field study gathered data from 41 leak sources across four key process areas: the Alkylation Plant, Saturated Gas Plant, Unsaturated Gas Plant, and Isomerization Plant.
Thirty of the 41 leaks were detected in the Saturated Gas Plant (SGP); five (5) in the Unsaturated Gas
Plant (USGP); and 6 leaks at the Isomerization Plant (ISOM) No leaks were found at the Alkylation
Out of 41 identified leak sources, 28 were collected and analyzed for their mass leak rate and chemical composition The analysis revealed that the leaks comprised a mixture of hydrocarbons ranging from C1 to C6 and higher.
Reproduced by IHS under license with API
Performance of the Fiber Laser (Detected Mass Leak Rate)
Figure 9-1 demonstrates the fiber laser's effectiveness in detecting mass leak rates, with the Y-axis used solely for data separation The results show that the fiber laser successfully identified all bagged leaks exceeding approximately 20 g/hr However, it failed to detect 3 out of 12 leaks below this threshold, specifically one each at SGP and USGP, and another at ISOM Notably, the laser did detect two known leaks—a 16 g/hr leak at USGP and a 0.12 g/hr leak at SGP—when a Styrofoam background was positioned behind the component.
Figure 9-1 Performance of Fiber laser (Detected Mass Leak Rates)
Total Hydrocarbons Mass Leak Rate, (g/hr) seen w ith added background seen not seen
At the SGP, an enamel white sign obstructed the laser beam's reflection, complicating leak detection A slight adjustment in the operator's position enhanced the leak's visibility, but interference from the sign persisted Utilizing a Styrofoam background further improved the leak's visibility, leading to its documentation as observed with the background.
In both instances where leaks were identified using Styrofoam backgrounds behind the leaking components, the initial failure to detect these leaks was attributed to inadequate background conditions.
The analysis revealed a screening rate of approximately 35 components per minute, leading to an estimated total of around 27,000 components monitored, which includes valves, pumps, and connectors, as detailed in Table 9-1.
11 Time spent scanning process areas during the test approximated 13 hours
Table 9-1 Components Monitored During Test
Estimated % LDAR Components Scanned Components Scanned
Estimated Connectors a Totals Valves Pumps Valves Pumps
5,485 57 21,940 27,482 a 4 times valves per plant guidance.
Testing the CO 2 Laser for Ethylene Monitoring
In 2002, Houston Area Advanced Research Center (HARC) commissioned tests at two olefin plants in Texas The objectives of these tests were to:
A demonstration of a portable optical gas imaging device was conducted at two industrial sites, specifically ethylene and polyethylene producers, to assess the device's effectiveness in detecting fugitive emissions during standard chemical plant operations.
• Identify, if possible, leaking equipment detected with the portable optical gas imaging device but listed as non-leaking when monitored under Method 21 procedures;
• Gather data that could be used to establish the mass emission detection capability of the portable optical gas imaging device; and
To assess the sensitivity of portable optical gas imaging devices in chemical plants, it is essential to collect data on various influencing factors These include the distance from the components being scanned, sight lines and angle-of-view, infrared backscatter, and the absorption properties of background materials Additionally, weather conditions and the chemical composition of emissions must also be considered.
Study Methodology
The tests were conducted at two ethylene facilities in Texas 12 Four process areas were monitored:
Mass rates of various leaks identified by the CO2 laser were assessed using bagging techniques, with measurements taken of the distance from the laser to the leaking component and the local wind speed near the leak Below, we present a brief discussion of the conclusions and findings.
12 Referred to as Site A and Site B
Reproduced by IHS under license with API
Study Findings
The primary conclusions and supporting findings from this effort were as follows:
1 The CO 2 laser was able to identify leaking components while monitoring under normal petrochemical plant operating conditions and good weather conditions (light wind, clear sky, summer temperatures)
• All leaks above about 1g/hr were detected by the CO 2 laser (See Figure 10-1)
• Ethylene was the only species examined during testing at both sites
Figure 10-1 CO 2 Laser Performance at Ethylene Facilities
Site A - not seen Site A - seen Site B - not seen Site B - seen
Method 21 techniques can sometimes misidentify leaks, as components marked for repair may not actually be leaking Instead, a CO₂ laser analysis revealed that these components were merely in the path of a plume from another overlooked source When wind direction changed, these tagged components were no longer affected by the plume and were confirmed to be leak-free Plant personnel have reported instances where they detected and tagged components for repair, only to find no leaks upon further inspection.
2 The majority of components detected as leaking had screening vales above 10,000 ppmv This result is in keeping with trends seen in API’s 1997 study of refinery LDAR data
At Site A, around 97% of the 95 identified leaking components exhibited screening values exceeding 1,000 ppmv, while 63% surpassed 10,000 ppmv Similarly, at Site B, 83% of the 52 detected leaks had screening values over 1,000 ppmv, with 52% exceeding 10,000 ppmv.
[1] American Petroleum Institute, “Analysis of Refinery Screening Data,” Publication # 310, Washington,
[2] ICF Consulting, Compendium of Sensing Technologies to Detect and Measure VOCs and HAPs in
[3] McRae, T G and Kulp, T G., “Backscatter Absorption Gas Imaging: a New Technique for Gas Visualization” Applied Optics, 1993, Vol 32, 4037-4050
[4] Gas Research Institute (GRI), “Evaluation of the GRI Gas Imaging Leak Survey System,” GRI- 98/0014, February 1998
Tom McRae's study, presented at the SPIE Conference on Environmental Monitoring and Remediation Technologies II in September 1999, details the field test results of GasVue for locating VOC and SF6 leaks The findings highlight the effectiveness of GasVue technology in environmental monitoring applications.
[6] Epperson, David L., Siegell, Jeffery, H., “Equivalent Leak Levels & Monitoring Frequencies for
Smart LDAR,” Valve World 2002, November, 2002
Reproduced by IHS under license with API
Reproduced by IHS under license with API