2.3.3. Damage Classification Using Acoustic Emission

To provide a means for interpretation of the data, Intensity Analysis graphs were developed at each AE monitoring location. The method was first introduced by Fowler and others [26] for the evaluation of fiber reinforced polymer vessels and is based entirely on signal strength. Intensity Analysis is a graphical method that differs from many other forms of acoustic emission assessments in the sense that it is focused on trends in the AE data as opposed to individual events. Intensity Analysis uses two parameters, which are both based on signal strength: (a) historic index (plotted on the horizontal axis) and (b) severity (plotted on the vertical axis).

Historic index and severity can be calculated using Equations (1) and (2) where *N* is the number of hits up to time (*t*), *Soi* is the signal strength of the *i*-th event, and *K* is an empirically derived factor that varies with the number of hits. The value of K has been previously selected in one case as: (a) *N*/*A* if *N* ≤ 50, (b) *K* = *N* − 30 if 51 ≤ *N* ≤ 200, (c) *K* = 0.85*N* if 201 ≤ *N* ≤ 500, and (d) *K* = *N* − 75 if *N* ≥ 501 [4].

$$H(t) = \frac{N}{N-K} \frac{\sum\_{i=K+1}^{N} S\_{oi}}{\sum\_{i=1}^{N} S\_{oi}} \tag{1}$$

$$S\_r = \frac{1}{50} \sum\_{i=1}^{i=50} S\_{oi} \tag{2}$$

Historic index, *H*(*t*), is a form of trend analysis that incorporates historical data in the current measurement. It is sensitive to changes of slope in cumulative signal strength versus time and compares the signal strength of the most recent hits to a value of cumulative hits. Severity, *Sr*, is defined as the average signal strength for the 50 hits with the highest numerical value of signal strength. The intensity analysis method has been widely used for the assessment of structural systems during load testing including reinforced concrete systems [12,27,28] and has been extended to the case of corrosion damage in pre-stressed concrete specimens [2–4,22].

By plotting the maximum historic index and severity values obtained over the duration of the test, an Intensity Analysis plot is generated. Due to the relationship between AE signal strength and damage growth, points that plot upward and to the right are associated with higher levels of damage.

Because Intensity Analysis (IA) uses historical information, an initial point on the Intensity Analysis plot must be chosen. This may be approached through visual inspection, numerical modeling, electrochemical measurements (in the case of corrosion damage), coring and petrographic examination, and other methods including suitable nondestructive evaluation techniques. Only a visual inspection was practical for the monitored locations within 105-C. Therefore, the initial point was chosen based on visual inspection.

The values of historic index and severity for the initial point were considered to account for pre-existing damage such that the historic index value at any time cannot be less than that for the initial point. For the severity, the distribution of the highest 50 signal strength values collected during the monitoring period in terms of their scattering from the mean value was used to develop the other 50 signal strength values with the same distribution but have a mean value equal to the severity of the initial point. Then the highest 50 numerical values from the collective set of 100 signal strengths, 50 from the monitoring period, and 50 developed from the initial point are used to calculate an updated severity value that takes into account the pre-existing condition.

Figures 10 and 11 are plots of Intensity Analysis results from the period beginning 10 September 2014 and ending 25 August 2015 for data recorded from resonant and broadband sensors, respectively. For the majority of field applications, only resonant sensors would be utilized due to the increased sensitivity of this sensor type in comparison to broadband sensors. The use of resonant sensors, therefore, reduces the number of sensors needed for a given application. However, resonant sensors do not provide high fidelity representations of the frequency content in comparison to broadband sensors. One purpose of using the two different sensor types is to investigate the associated differences in the results. The limits of the Intensity Analysis chart were developed based on data from resonant sensors [2]. Thus, it is expected that data collected from broadband sensors may yield underestimated damage classification if the same limits are used.

**Figure 10.** Intensity Analysis results for resonant sensors: (**a**) roof to column interface, (**b**) horizontal beam location, and (**c**) control location.

**Figure 11.** Intensity Analysis results for broadband sensors: (**a**) roof to column interface, (**b**) horizontal beam location, and (**c**) control location.

The preliminary estimation of damage was based on visual inspection during the initial visit to 105-C and was located near the border between the 'no damage' and the 'depassivation' regions of the chart (severity = 300,000 and historic index = 2.0) for both the vertical column to roof interface location and the horizontal beam location. This assumed level of damage underestimated the actual condition of the structures since these areas are known to have ongoing corrosion damage. Ideally, this initial point would be established through a combination of methods including visual inspection and electrochemical methods. Electrochemical methods, however, were not collected during this part of the study. For the control location, no damage was assumed and, therefore, the initial point was located at the left corner in the 'no corrosion' region of the chart.

Acoustic emission activity during the monitoring period (approximately one year) at the vertical column to the roof interface location indicated a progression from the initial state to the severe damage state for both sensor types. It is noted that the results of IA after approximately 2 months of monitoring (1 November 2014) showed that corrosion is ongoing at this location. On 1 December 2014, Intensity Analysis results indicated that severe damage occurred. For monitoring over this relatively short duration, such a progression from the initial state to the 'severe' damage state is indicative of a relatively high level of ongoing damage growth in the monitored areas. For this plot, the term 'cracking' refers to micro-cracking that is generally non-visible while 'severe damage' refers to visible cracking that may be accompanied by spalling. This result is supported by the spalling that occurred at this location during the monitoring period.

Acoustic emission data from the resonant sensor and the broadband sensor at the horizontal beam location progressed from the initial state to the cracking state over the duration of the monitoring period. This result is also an indication of ongoing damage growth at this location when the relatively short monitoring period is considered. The broadband sensor results (Figure 11b) indicated less damage than the resonant sensor results (Figure 10b) especially during the first 3 months of monitoring. This can be attributed to the lower sensitivity of the broadband sensors.

In contrast to the roof interface location and the horizontal beam location, the intensity analysis results for the control location indicate no damage progression during the monitoring period and, therefore, the initial state and final state coincide (plot on top of one another) for the control location.

#### 2.3.4. Remote Monitoring at +48 Level

A cellular connection was used to remotely operate the wireless acoustic emission data acquisition system. Data from the wireless system was collected between 9 September, 2014 (commencement of test) and 13 November 2014. Due to the loss of power from the solar power/battery system, 10 days of data were lost starting from 11 September 2014. The power was reconnected and the system continued to monitor the connection until 15 October 2014 when a thunderstorm caused a power outage and data was lost for another 13 days. The system continued to collect data afterwards until the data acquisition laptop was damaged on 13 November 2014. The damage was most likely due to moisture and was not repairable.

As described for the wired system data, the raw data was analyzed and appropriate data filters were used to separate meaningful data from spurious emissions. The limits of the data filters are shown in Table 2. Figure 12 shows acoustic emission activity in terms of amplitude versus time (showing both rain and temperature data) collected between 9 September 2014 and 13 November 2014 from the wireless acoustic emission system. This data set contains a significant number of hits with amplitude exceeding 80 dB. These hits are of relatively high amplitude and may be correlated to ongoing damage.

**Figure 12.** (**a**) Amplitude and temperature versus time for four wireless sensors at the +48 level and (**b**) rain versus time.

One objective of monitoring this location was to assess whether the large vertical crack in the wall is still active. This vertical crack has a width between 0.125 and 0.25 inches with several small hairline cracks extending from it in the horizontal direction. Figure 13 plots the cumulative signal strength (units of pico-Volt seconds) versus time (days) for the collected signals over the monitoring period. An increasing trend in the acoustic emission activity is observed in the figures, which indicates that damage may be progressing at this location.

**Figure 13.** Cumulative signal strength (pVs) versus time (days) for four wireless sensors at the +48 level.

To further investigate the trends in this data set, triangulation algorithms were used to investigate if AE events were generated from crack growth. Figure 14 shows the source location results from filtered acoustic emission data. In this figure, each red dot indicates a located acoustic emission event, which means that all four sensors received data with a specified time increment. The time increment

was determined based on the characteristic wave speed of the structure, which was experimentally determined during the installation site visit, and the geometry of the sensor grid. Source location from raw data was inconclusive since it showed acoustic emission activity throughout the monitored area. The 6 acoustic emission events from the filtered data set were located in the vicinity of the vertical crack. These results imply that crack growth or friction between crack surfaces was ongoing in this area during the monitoring period.

**Figure 14.** Source location results at the +48 level. Red dots indicate located AE events.
