Systems, methods and apparatus for acousto-ultrasonic assessment of metallurgical vessels

1 . A method for inspecting a metallurgical vessel wall, the method comprising:

receiving a plurality of time domain signals from one or more sensors, each of the plurality of time domain signals generated by a reflected stress wave induced by an impactor impacting the metallurgical vessel wall at an impact point, wherein each of the plurality of signals is received at a location at a predetermined distance from the impact point;

converting the plurality of time domain signals into a plurality of frequency domain signals;

identifying peaks in each of the plurality of frequency domain signals;

generating a combined frequency domain signal based on the plurality of frequency domain signals by cancelling out any peak from one of the plurality of frequency domain signals not having a corresponding peak in at least one other of the plurality of frequency domain signals; and

determining at least one of a thickness of the metallurgical vessel wall or an estimated location of a defect in the metallurgical vessel wall based on the combined frequency domain signal.

2 . The method of claim 1 wherein receiving the plurality of time domain signals comprises receiving two or more signals substantially simultaneously from two or more sensors.

3 . The method of claim 1 wherein receiving the plurality of time domain signals comprises receiving a first signal from a first sensor at a first time and receiving a second signal from the first sensor at a second time.

4 . The method of claim 1 wherein cancelling out any peak from one of the plurality of frequency domain signals not having a corresponding peak in at least one other of the plurality of frequency domain signals comprises cross-correlating at least one of the plurality of frequency domain signals with every other frequency domain signal.

5 . The method of claim 1 comprising determining an estimated defect location based on differences in frequency and amplitude between corresponding peaks from different ones of the plurality of frequency domain signals.

6 . The method of claim 1 comprising utilizing a k-means clustering machine learning algorithm to identify one or more clusters of corresponding peaks in the plurality of frequency domain signals.

7 . A method for inspecting a metallurgical vessel wall, the method comprising:

a. receiving a plurality of time domain signals from one or more sensors, each of the plurality of time domain signals generated by a reflected stress wave induced by an impactor impacting the metallurgical vessel wall at an impact point, wherein each of the plurality of signals is received at a location at a predetermined distance from the impact point;

b. converting the plurality of time domain signals into a plurality of frequency domain signals;

c. identifying peaks in each of the plurality of frequency domain signals;

d. generating a combined frequency domain signal based on the plurality of frequency domain signals by cancelling out any peak from one of the plurality of frequency domain signals not having a corresponding peak in at least one other of the plurality of frequency domain signals; and

e. determining at least one of a thickness of the metallurgical vessel wall or an estimated location of a defect in the metallurgical vessel wall based on the combined frequency domain signal,

the method further comprising utilizing a k-means clustering machine learning algorithm to identify one or more clusters of corresponding peaks in the plurality of frequency domain signals and, for each cluster of corresponding peaks:

determining a prominence metric comprising a sum of prominences of the peaks in the cluster;

determining a variance metric comprising a sum of absolute values of frequency difference for each peak in the cluster from a center frequency of the cluster; and,

determining a ratio of the prominence metric to the variance metric,

wherein the ratio of the prominence metric to the variance metric is used to determine a confidence score for a thickness determined based on that cluster of corresponding peaks.

8 . The method of claim 7 further comprising receiving a plurality of time domain signals at each of a plurality of predetermined measurement points, repeating the steps of the method for each of the plurality of predetermined measurement points, and generating an overall wear profile for the metallurgical vessel wall based on the confidence score of the thickness determined for each of the plurality of predetermined measurement points.

9 . The method of claim 8 wherein each of the plurality of predetermined measurement points is at a different predetermined height along a sidewall of the metallurgical vessel.

10 . An apparatus for inspecting a metallurgical vessel wall, the apparatus comprising:

one or more sensors configured to detect reflected stress waves in a metallurgical vessel wall and generate a plurality of time domain signals, each of the plurality of time domain signals generated by a reflected stress wave induced by an impactor impacting the metallurgical vessel wall at an impact point, wherein each of the plurality of signals is received at a location at the same predetermined distance from the impact point; and

processing electronics configured to:

receive the plurality of time domain signals from the one or more of sensors;

convert the plurality of time domain signals into a plurality of frequency domain signals;

identify frequency peaks in each of the plurality of frequency domain signals;

generate a combined frequency domain signal based on the frequency domain signals by cancelling out any peak from one of the plurality of frequency domain signals not having a corresponding peak in at least one other of the plurality of frequency domain signals; and

determine at least one of a thickness of the metallurgical vessel wall or an estimated location of a defect in the metallurgical vessel wall based on the combined frequency domain signal.

11 . The apparatus of claim 10 , wherein each of the sensors comprises a piezoelectric sensor.

12 . The apparatus of claim 10 , wherein each of the sensors comprises a fiber optic sensor.

13 . The apparatus of claim 10 , wherein each of the sensors comprises a laser-Doppler sensor.

14 . The apparatus of claim 10 , wherein each of the sensors comprises an accelerometer.

15 . The apparatus of claim 10 , wherein the processing electronics are configured to cancel out any peak from one of the plurality of frequency domain signals not having a corresponding peak in at least one other of the plurality of frequency domain signals by cross-correlating at least one of the plurality of frequency domain signals with every other frequency domain signal.

16 . The apparatus of claim 10 , wherein the processing electronics are configured to determine estimated defect location based on differences in frequency and amplitude between corresponding peaks from different ones of the plurality of frequency domain signals.

17 . The apparatus of claim 10 , wherein the processing electronics are configured to utilize a k-means clustering machine learning algorithm to identify one or more clusters of corresponding peaks in the plurality of frequency domain signals.

18 . The apparatus of claim 17 , wherein the processing electronics are configured to, for each cluster of corresponding peaks:

determine a prominence metric comprising a sum of prominences of the peaks in the cluster;

determine a variance metric comprising a sum of absolute values of frequency difference for each peak in the cluster from a center frequency of the cluster; and,

determine a ratio of the prominence metric to the variance metric,

wherein the ratio of the prominence metric to the variance metric is used to determine a confidence score for a thickness determined based on that cluster of corresponding peaks.