[9A2] Analysis of the repeatability of the pencil lead break artificial sound source
Reemt Hinrichs, Nan Jiang, Thomas Krause, Alexander Lange and Jörn Ostermann
Institut für Informationsverarbeitung, Leibniz University Hannover, Germany
Acoustic emission testing is used to monitor the structural health by the recording and analysis
of sound signals due to crack formation. Usually, a set of sensors is attached at key
positions on a structure under test to record and analyze acoustic signals created inside of
the structure. For reliable measurements, the sensors have to be coupled properly to the
surface. A common tool to test the coupling and therefore the effective sensitivity of the
sensors is the pencil lead break (PLB) which induces sound waves in a structure through
sudden breakage of a lead mine placed on the component’s surface. The effective sensitivity
is understood as the total sensitivity of the sensor in conjunction with the coupling
mechanism. To reliably test the coupling of a sensor across frequency, the magnitude spectrum
(MS) of the signal induced by the PLB has to be known. Additionally, the deviations
of the MS between individual PLBs have to be considered. The larger these deviations, the
lower the repeatability and therefore the reliability of the PLB.
This work analyzes the repeatability of the PLB in the frequency range of about 10 Hz to 300 kHz through controlled breakage of lead mines on a concrete girder. 50 PLBs were performed and the relative standard deviation (RSD) across frequency of the MS was examined. For frequencies with high signal amplitudes of the PLB, a RSD of up to 0.14 was calculated. It was found that higher frequencies tended to have larger RSD of about 0.4. The length of the fractured lead mine was measured for all PLBs and a correlation analysis was performed. Moderate to high correlations of up to -0.62 between the fractured length and the observed MS were found, explaining a significant part of the observed large deviations of 10 dB and more. A reduction of the RSD by up to 62% was observed when only PLBs with similar length of the fractured lead mine within a range of 2.8 mm to 2.9 mm were considered.
Overestimation of a sensor’s effective sensitivity can occur, when the PLBs consecutively induce signals significantly larger than the mean value. To assess the likelihood of this, the probability distribution of the MS was investigated. As only limited data points existed to estimate large deviations from the mean, a probability distribution was fitted to the data to replace the observed empirical distribution of the MS. It was calculated that three consecutive PLBs have a probability of about 1 % to 10 % that all PLBs exhibit magnitudes of at least +1 dB above the mean magnitude for a given frequency. For a deviation of at least +3 dB, a probability of about 0.1% to 1% was calculated. For five consecutive PLBs, these probabilities shrink by a factor of about five or more. It can be concluded that controlling the fracture length of the PLB can significantly increase its repeatability. Additionally, more than three PLBs should be used to test a sensor’s coupling with high reliability.
This work analyzes the repeatability of the PLB in the frequency range of about 10 Hz to 300 kHz through controlled breakage of lead mines on a concrete girder. 50 PLBs were performed and the relative standard deviation (RSD) across frequency of the MS was examined. For frequencies with high signal amplitudes of the PLB, a RSD of up to 0.14 was calculated. It was found that higher frequencies tended to have larger RSD of about 0.4. The length of the fractured lead mine was measured for all PLBs and a correlation analysis was performed. Moderate to high correlations of up to -0.62 between the fractured length and the observed MS were found, explaining a significant part of the observed large deviations of 10 dB and more. A reduction of the RSD by up to 62% was observed when only PLBs with similar length of the fractured lead mine within a range of 2.8 mm to 2.9 mm were considered.
Overestimation of a sensor’s effective sensitivity can occur, when the PLBs consecutively induce signals significantly larger than the mean value. To assess the likelihood of this, the probability distribution of the MS was investigated. As only limited data points existed to estimate large deviations from the mean, a probability distribution was fitted to the data to replace the observed empirical distribution of the MS. It was calculated that three consecutive PLBs have a probability of about 1 % to 10 % that all PLBs exhibit magnitudes of at least +1 dB above the mean magnitude for a given frequency. For a deviation of at least +3 dB, a probability of about 0.1% to 1% was calculated. For five consecutive PLBs, these probabilities shrink by a factor of about five or more. It can be concluded that controlling the fracture length of the PLB can significantly increase its repeatability. Additionally, more than three PLBs should be used to test a sensor’s coupling with high reliability.
