SAE AE-2 Lightning Committee White Paper

Transcription

1 SAE AE-2 Lightning Committee White Paper Recommended Camera Calibration and Image Evaluation Methods for Detection of Ignition Sources Rev. NEW January

2 Table of Contents Executive Summary Introduction Sensor Standard Voltage Spark Energy Source Procedure Round Robin Test Results Recommendations Appendix Procedure used for the Round Robin Example procedure for sparker calibration/preparation Quantitative Output Examples Figure 1: Threshold Calibration Example (the energy measurement method was not specified explicitly) 9 Figure 2: Round Robin Test Results. The data sets are spaced apart for visualization purposes only Figure 3: Examples of Voltage Spark Images Figure 4: Electrode Geometry (material should be tungsten) Figure 5: Setup Diagram & Partial List of Setup and Settings to Record Figure 6: Example of an electrode assembly

3 Executive Summary The SAE AE-2 ARP 5416 working group and EUROCAE WG-31 were tasked with creating guidance for detecting ignition sources using digital cameras to supplement the existing guidance and enhance consistency in approaches across labs. Current guidance states that [t]he photographic method used must be capable of detecting 200 μj +0/-10% electrical sparks. The working group has proposed a camera calibration method using a quantitative threshold to determine a pass or fail. The SAE AE-2 ARP 5416 working group in close collaboration with the EUROCAE WG-31 performed a round robin test between eight different laboratories, analyzed the data using open source software, and reviewed the resulting quantitative image data. The threshold for each data is set at the 10% level of the 200 μj voltage sparks. Test results showed that 300 μj voltage sparks fall above the threshold for most electrode sets, which has been deemed by the committees to be a sufficiently conservative estimate. The tungsten electrodes 1.6 mm have been found the most reliable, repeatable and stable (limited scatter) making them the preferred electrode choice for a calibration procedure. In order to cover lower visibility sparks observed from other electrode systems, it has been proposed to add a margin on the calibrated threshold (tungsten 1.6 mm electrodes). This margin is defined from the results obtained by the laboratories who had contributed to the round robin test. The SAE AE-2 and EUROCAE WG-31 committees have also made recommendations for standardizing the voltage spark source system set up and defined the electrode design to increase the consistency of calibration procedure between laboratories. 3

4 1. Introduction SAE ARP5416A and EUROCAE ED-105 allow use of digital cameras for lightning spark detection because film for Polaroid photography, which is an acceptable lightning spark detection standard, has disappeared. A standardized camera calibration method is needed to produce consistent lightning spark detection between lightning test laboratories. This white paper describes a standard procedure for camera calibration. The use of digital cameras for lightning spark detection has introduced some new issues related to the use of a variety of different cameras, lenses, and settings (see Section 2). Round robin testing results have emphasized the need to standardize the reference voltage spark source system, recommendations for which are given in Section 3. The variety of cameras, lenses, and settings also creates a need for a consistent calibration method as well as a need for guidance for the pass/fail criteria for the test images (see Section 4). ARP 5416Aand ED-105 section states that [t]he photographic method used must be capable of detecting 200 μj +0/-10% electrical sparks. This creates a challenge when aiming to ensure consistency of results between different test labs. Consequently, the SAE AE-2 and EUROCAE WG-31 Lightning Committees defined a quantitative threshold determination specified as the allowable false-pass rate for the sensor when detecting 200 μj sparks. The committees recommend setting the allowable false pass rate at 10% and deriving a pass fail threshold (Tw) from multiple images of 200 μj sparks. The rationale for the 10% value is the correlation to the flammable gas mixture test method, which in section step 1 (k) requires the verification of the mixture calibration in nine out of ten successive tests using 200 μj sparks. The SAE AE2 and EUROCAE WG-31 Lightning Committees evaluated results of round robin tests to confirm that all 300 µj voltage spark fall above the selected 10% threshold (Tw - see Section 5) for the electrode system selected for the calibration process. In addition, 1.6 mm diameter tungsten electrodes have been found to be the most reliable, repeatable and stable (limited scatter) light source making them the preferred electrode choice for a calibration procedure. The 300 µj ignition sources created with different electrode systems or ignition sources created with heated filament are detected by the calibrated camera the vast majority of the time. However, lower visibility sparks could be observed from some electrode systems. To account for this variability, a margin was added to the calibrated threshold obtained from 1.6 mm tungsten electrodes. This margin is derived from the results obtained by the laboratories who had participated in the round robin test (see Section 5). Section 4 of this white paper gives the detailed procedures for calibrating cameras. The example procedures performed during the round robin tests are given in the appendix (Section 7). The voltage spark source is assumed to be the most difficult ignition mechanism producing visible light to detect using the photographic method relative to other ignition mechanisms created in direct lightning tests. 4

5 2. Camera In this white paper, the term camera includes light sensors, lenses, image data recording, and associated controls and settings. This white paper is focused on the use of digital cameras, and the observed peak pixel in the calibration and test image is the specified quantitative output value. Other non-imaging light sensors such as photo-multiplier tubes are not recommended but may also be used; however, they will require additional validation to ensure that ignition mechanisms are correctly identified. The camera, lens, and camera settings should be treated as a set of calibrated equipment which should be calibrated every 12 months. 3. Standard Voltage Spark Source Voltage Spark Source Definition: 200 µj voltage spark sources are widely used for gas and camera verification. These sources can be utilized for this procedure with the modifications defined here below. In general, a compact and simplified voltage spark source design will minimize stray capacitance. The following set of definitions do not include a complete specifications for a voltage spark source. The guidance provides better definition of certain aspects of the voltage spark source design that have shown to produce acceptable and consistent results. - Where it is necessary to measure capacitance of the circuit, a capacitance bridge should be used for most consistent and accurate results. - Dielectric materials used for support structure should be shaped to minimize E-field concentrations. Charge leakage across dielectric surfaces will concentrate and discharge as corona from dielectrics. - Sparker should be operated in a relatively dry air environment. The greatest source of poor sparker performance is operating in conditions that drive dielectric surface charge leakage (e.g. humid, oily). Electrodes: - Material: Tungsten (non-radioactive) - Shape: 1.6 mm (1/16 in) diameter rod with a 0.8 mm (1/32 in) radius hemispherical tip (see Figures 1 and 2). There should be at least 8 mm (5/16 in) of exposed rod on either side of the gap in order to avoid field distortions from any other features of the electrode assembly. - The rod electrodes should be set up so that the rod centers lie on the same line, with no visible off-axis displacement. - Gap distance: 1.5 mm +/- 0.1 mm - Both electrode butt and tip features should minimize E-field concentrations away from the tip - The machined surface roughness should be relatively smooth, the higher the roughness, the longer the conditioning process will be for new electrodes. - Electrodes should be conditioned such that the metal oxide formation on the spark location and the surface roughness reach an equilibrium state. The use of tungsten metal maximizes the ongoing consistency of a conditioned set of electrodes. Conditioning should consist of operating the electrodes for roughly 10,000 discharges do not sand the electrodes afterwards, however 5

6 electrodes should be cleaned with acetone and soft cloth or fabric. When conditioning the electrodes the voltage can be set higher than what is required to achieve 200 µj, i.e. 10 kv. - The use of threads in attaching electrodes to the assembly is a common method to enable the adjustment of the gap spacing however the implications to E-field concentrations should be considered. Figure 1: Electrode Geometry (material should be tungsten) Figure 2: Example of an electrode assembly 4. Procedure The following steps describe the procedure that should be used for determining thresholds: 1. Select and record the camera manufacturer, model number or name, and serial number. If the camera uses removable parts such as lenses, camera backs, etc., record the manufacturer, model number or name, and serial number for each of these items. Assemble the camera system. The calibration only applies to the recorded camera system. 2. Set up a conditioned 200 µj voltage spark source. 3. Set the voltage spark source in black-out-box in order to eliminate any extraneous light sources. 4. Set up and record camera and lens settings. The camera and lens settings should be manual and all internal image processing and filters disabled. The aperture setting (f-number) should be based on the required depth of field so that the distance range for expected test articles is in focus. 5. Set the front of the camera body at the same distance away from the gapped electrodes no more than 1.5 m. a. Position the camera so that it has a clear view of the gapped electrode and the camera s center line of sight is orthogonal to the gapped electrodes. 6. Take a test photo with the setup illuminated to ensure that the gapped electrodes are in focus and that the depth of field is sufficient to capture the range of the test article. Verify that the 1.5 mm electrode gap spans at least 5 pixels in the recorded photos. 7. In the black out box, take several photos of 200 µj +0/-10% sparks. Analyze the photos and determine the peak pixel value of the spark. Adjust and record the ISO and camera aperture settings so that the average of the measured peak pixel values are between 60 and 70% of the camera s dynamic range. a. Verify that the random noise in the complete image is well below the measured peak pixel values. 8. Set the aperture (f number) so that the distance range for expected test articles is in focus. Record the distance range for this camera calibration. Take at least 100 photos of 200 µj +0/-10% sparks. Verify that all discharges fall within the stored energy range. 6

7 9. Analyze the images and determine the peak pixel value of the spark photo. Use appropriate software (e.g. ImageJ, available from NIH: for peak pixel value determination. a. The same software used for the calibration must be used for the test itself. b. The region around the voltage spark gap can be chosen to be small enough to avoid camera noise, but should capture the entire spark channel. 10. After the processing is complete, copy the data into an appropriate file such as a spreadsheet. 11. Set the threshold level T W of the analysis method so that 10% of the photos have peak pixel values of the spark that fall below the threshold and the remaining 90% of the photos have peak pixel values of the spark above the threshold. 12. Define pass/fail threshold for subsequent lightning tests (T P/F) by multiplying the threshold T W by 0.8: T P/F = 0.8 T W 13. Record camera setup, settings, and pass/fail threshold T P/F. a. Record the distance between the front of camera body to electrode. b. Record the material of the electrodes (should be tungsten). c. Record the electrode gap (should be 1.5 mm +/- 0.1 mm). d. Record the breakdown voltage & energy. e. Record the humidity. f. Take a picture with the voltage spark source to ensure that the camera is focused. g. Save an unedited copy of the original photo data file. 14. Take at least 100 photos of 300 µj +0/-10% sparks. 15. Analyze the photos and determine the peak pixel value of the spark for each image 16. Verify that all peak pixel values in the photos of the sparks at 300 µj are above the threshold T W. The following steps describe the procedure that should be used for the performance of the test and the analysis of the test images: 1. Set up camera and test article. Use the same settings for the camera as were used in the calibration procedure; account for mirror reflection distances. a. Distance variations should be limited and if not identical, the distance to the object should be lower than the distance used for the calibration as the camera becomes more sensitive as the distance becomes shorter. Another option is to calibrate again further away than the test distance. 2. Perform test and capture test images. 3. Analyze the images and determine the peak pixel value of the image in the region of interest. 4. Evaluate the peak pixel value against the pass/fail threshold T P/F and identify any potential ignition sources. 5. Verify that the identified potential ignition sources can be correlated to the test article and eliminate possible camera defects or test artifacts. The use of two or three calibrated digital cameras will help eliminate possible camera defects that may otherwise be confused as ignition sources. 6. If the peak pixel value falls above the pass/fail threshold T P/F the test should be considered a fail. If all peak pixel values in regions of potential ignition sources fall below the pass/fail threshold T P/F the test should be considered a pass. Note 1: Cameras with interchangeable lenses 7

8 If the camera lenses are interchangeable, they shall be kept the same between calibration and test treat the camera and lens as a matched set. Note 2: Cap-on /background image subtraction Some cameras may have several dead pixels that show up in the image that do not correspond to the calibration source or an ignition source in the test. These pixels can be identified as artifacts by taking a photograph with the lens/view finder cap on while keeping all other settings the same. The cap on image can then be subtracted from the test image or calibration image in order to eliminate the dead pixels. 5. Round Robin Test Results The following section and appendix detail the results and procedure from the round robin testing. The detailed procedures are intended as suggestions with the overall recommendations of the working group captured in the next section. The SAE AE-2 ARP 5416 working group and EUROCAE WG-31 round robin testing included participants from NIAR, Subaru, NTS, DGA, Boeing, LCOE, Cobham, and Kyushu Institute of Technology. Figure 1 shows an example data set with images taken at 150, 200, and 300 µj. The data is analyzed using the ImageJ 1.48v software (from NIH: and the maximum grey scale value is extracted as the quantitative output. Other software may be used as long as it fulfills the same requirement. For this camera and its specific setup and settings, the threshold corresponding to the 10 th percentile of the max grey value TW is 145. When used in testing, any image producing a max gray scale value greater than or equal to the pass/fail threshold TP/F (145 x 0.8 = 116) will be considered a fail. This margin was derived from tests performed with electrodes of varying material and shape during the round robin. 8

9 Figure 1: Threshold Calibration Example (the energy measurement method for this particular test was not specified explicitly) Figure 2 shows the test results from all round robin data provided. Each data set s maximum grey scale value, which may be comprised of 150, 200, and 300 µj or 200 and 300 µj voltage spark image sets, has been normalized to the 10% value of the 200 µj results (TW). The resulting plot shows that the 10% allowable false-pass rate at 200 µj (TW) gives a reasonable non-zero threshold that will capture some discharges at 150 µj, but does not miss any discharges at 300 µj. 9

10 Figure 2: Round Robin Test Results. The data sets are spaced apart for visualization purposes only. Each data set has a nominal stored energy of the discharge of 150, 200, and 300 µj +0/-10% respectively. (Note that the green and red data sets use electrodes with a hemispherical tip with radius of 1/16 in all other used a radius of 1/32 in) For reference Figure 3 show a few examples of voltage sparks from the round robin. 6. Recommendations Figure 3: Examples of Voltage Spark Images The SAE AE-2 and EUROCAE WG-31 committees recommend the following camera calibration method for detecting ignition sources: 1. The camera should be calibrated using a 200 µj voltage spark source. A threshold should be set so that 10% of the photos have peak pixel values of the spark that fall below the threshold and the remaining 90% of the photos have peak pixel values of the spark that fall above the threshold. Record a minimum of 100 spark photos at 200 µj. 10

11 2. The threshold should be verified using a 300 µj voltage spark source. All photos recorded at 300 µj should have peak pixel values of the spark above the threshold. Record a minimum of 100 spark photos at 300 µj. 3. Multiply the threshold by 0.8 to set the pass/fail threshold for subsequent lighting tests, to account for variations observed with different electrode systems. 4. The same photo evaluation method should be used for both the calibration and subsequent lightning tests. 5. Tungsten (non-radioactive) electrodes (1.6 mm (1/16 in) diameter rod with a 0.8 mm (1/32 in) radius hemispherical tip) should be used. 6. The electrode gap should be set to a gap width of 1.5 mm +/- 0.1 mm. 7. The peak pixel value should be used for the quantitative image output. 8. Any combination of camera, lens, and settings may be used as long as the calibration method can be successfully performed. 11

12 7. Appendix Procedure used for the Round Robin The following procedure was used in the camera calibration round robin the results of which are shown in section Set and condition voltage spark source a. Use tungsten electrodes b. Electrode shape (see Figure 4) i. Hemispherical electrode with a radius of in (1/16 in diameter tungsten rod) c. Condition the electrodes with roughly 10,000 discharges do not sand the electrodes afterwards, however electrodes should be cleaned with acetone and soft cloth or fabric i. For conditioning the voltage can be set higher than what is required to achieve 200 µj, i.e. 10 kv. 2. Set voltage spark source to produce discharges of 200 µj +0/-10% stored energy. 3. Set the voltage spark source in black-out-box in order to eliminate any extraneous light sources. 4. Set the front of the camera body at a distance of the same distance away from the voltage spark as the test article in the final test, no more than 1.5 m and accounting for mirror reflection distances and mirror losses 5. Fix the camera and lens settings (set to manual control) and adjust the camera settings until the discharge is clearly visible in the image, but does not saturate the camera a. A suggested starting point is: 50 mm lens (max aperture f/1.8), aperture f/12, 5 second exposure time, ISO 1600 i. ISO 1600 (or equivalent gain for sensitivity of light sensor) is a starting value. The ISO (or gain) shall be set such that the measured intensity values are nominally halfway in the device s dynamic range b. Note all camera and lens settings. 6. Take a daylight/setup test photo to ensure that the voltage spark gap is in focus and that the depth of field is sufficient to capture the range of the test article. 7. Take a test image and verify that the light intensity follows the guidance in 5.a.i 8. Obtain 100 measurements (e.g., digital images) of the voltage spark with discharges of 200 µj +0/-10% stored energy. Check that all discharges fall within the stored energy range. 9. Set voltage spark source to produce discharges of 300 µj +0/-10% stored energy. 10. Obtain 100 measurements (e.g., digital images) of the voltage spark with discharges of 300 µj +0/-10% stored energy. 11. Analyze the images that fall within the stored energy range of stored energy 200 µj +0/-10% 12. Use ImageJ, available from NIH: After opening ImageJ, select Analyze Set Measurements - select all measurements 14. In Edit>Options>Conversions uncheck Weighted RGB Conversion 15. Select Process Batch Measure 16. Navigate to the folder where the images are located and click select. 17. After the processing is complete, copy the data into Excel 18. Sort the images by max value. 19. Set the threshold level of the analysis method such that 10% of the measurements fall below and the remaining 90% are above the threshold. (round down if necessary) 12

13 20. Analyze the images of discharges with a stored energy of 300 µj +0/-10% using same steps above to verify that the all of the 300 µj quantitative output values fall above the threshold. 21. Record keeping (see Figure 5) a. Note the distance between the front of camera body to electrode b. Note the material of the electrodes (should be tungsten) c. Note the electrode gap (shall be 1.5 mm) d. Note the capacitance e. Note the breakdown voltage & energy f. Note the humidity (should be less than 5%, the lower the better) g. Take a picture in daylight to ensure that the camera is focused h. Save an unedited copy of the original photo Figure 4: Electrode Geometry (material should be tungsten) Figure 5: Setup Diagram & Partial List of Setup and Settings to Record 13

14 Figure 6: Example of an electrode assembly Example procedure for sparker calibration/preparation - The operation of the reference sparker is sensitive to parasitic capacitance, so it should be installed and calibrated in the position it will be operated. An abbreviated set of steps to calibrate and ensure proper operation of the sparker once it has been mounted into its operating location are listed here. - Ensure the electrodes are affixed and aligned with the proper spark gap distance. The spark gap must be set to 1.50 mm ± 0.05 mm and use care not to scratch the electrodes. Set the electrode gap before conditioning the electrodes since scratching can contaminate the electrode surface and could require additional conditioning. - The entire sparker assembly must be cleaned with acetone to minimize charge leakage once the electrodes are affixed and aligned. Clean, non-powdered gloves (latex or nitrile) shall be worn at all times when handling the sparker during and at all times after cleaning. - Connect the high voltage and ground leads to the sparker. - Setup all supporting equipment and turn on devices that require a warm-up period. Dry the sparker and prepare the surrounding environment for the calibration procedure by flowing dry air ( 5% relative humidity) into the test enclosure. - Condition the sparker electrodes by activating the negative high voltage supply (nominally set at 8-9 kv) and allowing the sparker to naturally break down continuously. The indication that the sparker electrodes are sufficiently conditioned is observed when the natural consecutive breakdown frequency is consistent to within 1 sec. - Set the sparker voltage by adjusting the sparker high voltage supply until the consecutive breakdown frequency is nominally 10 to 15 sec. Then use the spark breakdown voltage to calculate the capacitance needed to produce the desired spark energy (e.g., 200 µj or 300 µj) using the equation E=1/2 CV 2. - Remove all electrical connection from the sparker and place the capacitance meter probe connectors near their respective connection locations without making contact. Using a calibrated capacitance bridge, measure the sparker capacitance across the spark gap by placing one of the capacitance meter probes on each of the sparker electrodes near the gap. Adjust the sparker adjustable capacitor until the desired capacitance is reached determined by the prior calculation. - Reconnect the high voltage and ground connections to the sparker. Keep the environment around the sparker dry throughout the entire sparker and camera calibration process by continually flowing dry air at a moderate rate. 14

15 Quantitative Output Examples Round robin used single pixel peak value, the maximum gray scale value, as the quantitative output from the calibration images. Capturing and evaluating the peak pixel value is a relatively simple process that analyzes the image on a pixel by pixel basis. Participants of the round robin were allowed to explore the use of mean value for both the full image and a region of interest around the voltage spark. However, this was rejected in favor of the use of the peak pixel value because it was found to be the most efficient and simple criteria. 15