Gas leak detection and event selection based on spatial concentration variability and other event properties
This work provides event selection in the context of gas leak pinpointing using mobile gas concentration and atmospheric measurements. The main idea of the present approach is to use a moving minimum to estimate background gas concentration, as opposed to the conventional use of a moving average for this background estimation.
1. A method for gas leak event detection and selection, the method comprising:
a) providing a series of gas concentration measurements c i indexed by an integer i; wherein the providing the series of gas concentration measurements c i comprises
i) performing raw gas concentration measurements at several times from a moving vehicle, wherein the raw gas concentration measurements are measurements of gas concentration at locations of the vehicle;
ii) determining the series of gas concentration measurements c i from the raw gas concentration measurements and from knowledge of position vs. time of the moving vehicle;
b) detecting peaks p i in the series of gas concentration measurements c i , wherein each peak p i has a corresponding amplitude a i ;
c) computing a moving minimum mc i given by the minimum of c j , wherein j is an index in a range i−n min +k 1 ≤j≤i+k 1 , wherein n min is a moving minimum window width, wherein k 1 is a first window offset, wherein k 1 is an integer≤n min , and wherein mc i =min([c i−n min +k 1 , c i−n min +1+k 1 , . . . , c i+k 1 ]);
d) computing a moving standard deviation σ i of mc j , wherein j is an index in a range i−n b +k 2 ≤j≤i+k 2 , wherein n b is a moving standard deviation window width, wherein k 2 is a second window offset, and wherein k 2 is an integer≤n b ;
e) performing gas leak event selection by computing a ratio R i of peak amplitude a i at peak p i to the moving standard deviation σ i of at peak p i and selecting gas leak events that have R i exceeding a predetermined threshold condition; and
f) providing selected gas leak events as an output;
wherein the performing gas leak event selection further comprises computing a width w i of the peak p i ;
wherein the predetermined threshold condition is a predetermined curve defined in a w i −R i plane.
2. The method of claim 1 , wherein the threshold curve is a line.
3. The method of claim 1 wherein the series of gas concentration measurements c i has its points substantially equally spaced apart in position.
4. The method of claim 1 wherein the width w i is normalized by a speed of the moving vehicle.
5. The method of claim 1 , wherein the detecting peaks p i comprises identifying data points having corresponding concentrations c i such that c i−1 ≤c i and c i+1 ≤c i .
6. The method of claim 1 , wherein the detecting peaks p i comprises using a peak finding algorithm to identify peaks p i .
7. The method of claim 6 , wherein the peak amplitude a i is provided by the peak finding algorithm.
8. The method of claim 1 , wherein the peak amplitude a i is determined according to a i =c i −mc i .
9. The method of claim 1 , wherein the moving minimum window width n min corresponds to a spatial width in a range from 20 m to 200 m.
10. The method of claim 1 , wherein the moving standard deviation window width n b corresponds to a spatial width in a range from 50 m to 500 m.