**3. Results**

### *3.1. Results of AE Signal Analysis*

### 3.1.1. Individual AE Signal Class Distribution Analysis

After dividing the recorded AE signals into four classes, using k-means algorithms, two AE parameters were analysed (as an illustration of ongoing processes)—signal strength and signal duration. The number of classes was input by the authors into the software used to analyse the AE signals. The value adopted corresponds to the nature of the material's work, and allows us to link individual classes with specific processes. Another criterion explaining the imposed number of classes was the level of individual signals matched to the appropriate classes, which in this case was over 90%. The following individual AE signal classes were assigned to processes occurring in the structure of the tested material:


The paper presents graphs of the descriptors over time, including their division into classes, for mock samples from each of the tested series. The results obtained for samples within the group were similar.

Analysing Figures 3–7, it can be seen that for samples in the air-dry state exposed to environmental factors (BS, BW and BM), the signals of 1–3 classes can be seen from the beginning of the external load action. Shortly after applying the force, Class 3 signals begin to appear, indicating that the fiber destruction process has begun. As the cracks in the tensile zone progress and the cracks deepens, Class 4 signals of gradual fiber breakage start to appear at the bottom of the sample. Exceeding the stress limits in the material results in a rapid increase in the recorded descriptors.

In the case of samples subjected to direct fire (BP) and high temperature (BC), two classes of signals were observed in the recorded time waveforms—Class 1 and Class 2. The authors related this fact to the possibility of the damage or degradation of the reinforcing fibers under the influence of temperature, which was associated with a change in the way the components are destroyed. It was found that dry samples and samples exposed to environmental factors, due to the presence of reinforcing fibers, were destroyed by exceedingly high bending stresses. On the other hand, the samples exposed to temperature cracked due to too much shear stress, likely because of the degradation of the fibers. To confirm these assumptions, a microscopic analysis of the fractures extracted from the tested components was conducted. During the analysis of the characteristics of the recorded descriptors, it was also found that during the bending of flamed and fired samples, the occurrence of most AE events was associated with a lower strength of signals than those found in the cases of samples from other groups. According to the authors, this fact confirms previous assumptions, which in their opinion validates the use of the AE method for assessing the degree of change of the mechanical parameters of the fiber–cement boards.

**Figure 3.** AE signal graphs for the mock BS sample: (**a**) signal strength distribution in time; (**b**) signal duration distribution in time, with plotted force increment curve.

**Figure 4.** *Cont.*

**Figure 4.** AE signal graphs for the mock BW sample: (**a**) signal strength distribution in time; (**b**) signal duration distribution in time, with plotted force increment curve.

**Figure 5.** AE signal graphs for the mock BM sample: (**a**) signal strength distribution in time; (**b**) signal duration distribution in time, with plotted force increment curve.

**Figure 6.** AE signal graphs for the mock BP sample: (**a**) signal strength distribution in time; (**b**) signal duration distribution in time, with plotted force increment curve.

**Figure 7.** *Cont.*

**Figure 7.** AE signal graphs for the mock BC sample: (**a**) signal strength distribution in time; (**b**) signal duration distribution in time, with plotted force increment curve.

Analysing the classes of AE signals occurring in the analysed patterns for samples BP and BC, it was found that the absence of Class 3 and 4 signals in the recorded events clearly indicates damage to the reinforcing fibers, or the possibility of delamination and voids in the material structure. It should also be noted that there is a close correlation between the type of AE signal classes recorded, and the destructive force value. This gives rise to the conclusion that the use of the AE method may allow the analysis of fiber–cement components in terms of the occurrence of material defects already formed at the production stage (e.g., discontinuities, uneven distribution of reinforcing fibers).

In summary, it was found that in the case of samples in the air-dry state, soaked in water and cyclically frozen and defrosted, the process of sample destruction was ductile. The result of the bending force was a deepening and widening of the crack in the element's tension zone. After reaching maximum strength, the samples were gradually unloaded. On the other hand, samples exposed to flames and high temperatures broke in a fragile manner. Along with reaching the maximum value of force, the element rapidly split into two parts. The differences in destruction mechanisms clearly influenced the types of recorded signal classes.

### 3.1.2. Analysis of Frequencies Accompanying Changes in Mechanical Parameters

Using the options available in the Vallen software, the frequencies accompanying the event emissions were extracted for the recorded data. The analysis results are presented in Figures 8–12. For a more readable representation of the frequency ranges emitted by the material, two frequency distribution graphs are provided for each of the mock sample series. The first one is used to present the frequencies from the whole tested run, and the second one details the frequencies of signals recorded before the destruction moment.

**Figure 8.** Frequency distribution graph during the test of the mock BS sample: (**a**) taking into account the entire frequency range; (**b**) specifying the frequency range before destruction.

**Figure 9.** Frequency distribution graph during the test of the mock BW sample: (**a**) taking into account the entire frequency range; (**b**) specifying the frequency range before destruction.

**Figure 10.** Frequency distribution graph during the test for the mock BM sample: (**a**) taking into account the entire frequency range; (**b**) specifying the frequency range before destruction.

**Figure 11.** Frequency distribution graph during the test of the mock BP sample: (**a**) taking into account the entire frequency range; (**b**) specifying the frequency range before destruction.

**Figure 12.** Frequency distribution graph during the test of the mock BC sample: (**a**) taking into account the entire frequency range; (**b**) specifying the frequency range before destruction.

Analysing the graphs presented above, it can be concluded that the results from subjecting the tested components to two groups of operating factors (environmental and exceptional) illustrated significant differences in the emitted frequency ranges. Changes in the mechanical parameters of the samples in the air-dry state, water-soaked and cyclically frozen then defrosted under an external load, are associated with low and high frequency signals. Most of the recorded frequencies exceed the

200 kHz threshold, and some events produce sounds of 300–500 kHz. The situation is different for flamed and fired samples. The bending of components exposed to high temperatures caused events at much lower frequencies, only some of which exceed 100 kHz.

### *3.2. Results of Microscopic Analyses*

Microstructure images of the extracted sample fractures are shown in Figure 13. All images show a 250-fold magnification of the surface.

(**a**)

(**E**) 

**Figure 13.** *Cont.*

**Figure 13.** Image of the mock microstructure: (**a**) of BS sample, (**b**) of BW sample, (**c**) of BM sample, (**d**) of BP sample, and (**e**) of BC sample. 144

(**H**)

Based on the analysis of the results shown in Figure 13, it should be concluded that the macrostructure of each of the analysed board samples before microscopic examination was determined, visually, as compact. Microscopic observations determined that the structure of the samples were fine-porous, with pore size up to 50 μm. Deep grooves of up to 500 μm in width were found on the fracture surfaces. A high density of irregularly distributed cellulose fibers was observed at the tested fractures (Figure 13a–c), except for the flamed sample and the fired sample. In the burned sample, it was observed that most of the fibers are fired or blended into the matrix. Flaming causes gradual burning of fibers, and the degradation of their structure, depending on the flame range (Figure 13d,e). Caverns and grooves from the torn fibers were observed. The matrix structure is defined as granular with numerous delaminations. No space was observed between the fibers and the matrix, indicating a strong bond between them. In the analysis of the flamed sample (Figure 13d) a locally altered matrix structure was observed.

According to the authors, the microscopic observations performed confirmed previous assumptions, which justify the necessity and legitimacy of the AE method for evaluating the degree of change in the mechanical parameters of fiber–cement composites.
