*6.2. Signal Processing*

The STFT spectrograms of both sensors for the three grinding conditions are shown in Figure 4. The color scale in the spectrograms represents the magnitude of a frequency in a given interval of time, the characteristic frequencies of the process, as well as the most influential frequencies, are shown in different shades of red colors. On the other hand, frequencies that are slightly sensitive to the grinding process are presented in shades of blue. The green, yellow, and orange colors, located in the middle of the color scale, represent frequencies sensitive to the process, but with a minor influence compared to the red ones.

In the spectrograms corresponding to the three slight grinding conditions (25 μm, 35 μm, and 50 μm), variations of magnitude can be observed over time. These variations were caused by the contact between the workpiece and the grinding wheel abrasive grains, which did not show a constant contact behavior, since many of the grains did not remove material from the workpiece surface, generating low levels of AE activity. Thus, little energy was released during the process, justifying the low harmonic content of the signals.

The two moderate grinding conditions (105 μm and 150 μm) showed higher magnitudes when compared to the slight conditions. An increase in red tones can be observed throughout the frequency spectrum, which is primarily caused by the grinding wheel wear and the process severity. The color pattern was shown to be more uniform due to the increased contact between the abrasive grains of the grinding wheel and the workpiece surface. The spectrograms representing the severe grinding processes (210 μm and 350 μm) presented uniform harmonic content over time. This behavior was caused by the continuous and severe contact between the grinding wheel grains and the ceramic component, generating more AE activity. In the three grinding conditions, a characteristic frequency was observed, with an approximate value of 35 kHz, present throughout the machining time. However, the more severe grinding condition presented higher harmonic content, assuming higher magnitude levels. The results obtained in Figure 4 are in agreement with the results of the direct measurements of Figure 3 and Table 2. As in Figure 3, it is possible to observe the color changes as the process severity increases, which is directly related to the surface roughness and surface quality of the ground workpiece. The attenuation of the PZT diaphragm response, near its maximum response region

(200 kHz), can be clearly observed in the spectrograms. There is a frequency component close to 200 kHz, which can be easily observed in the AE spectrograms, whereas in the PZT spectrograms it is difficult to observe, especially in slight cutting conditions, which generate lower acoustic levels. For this reason, it is not recommended to choose frequency bands near the maximum response region of the sensors.

The differences between the spectrograms of both sensors are related to the characteristics of the sensors, such as: The frequency response of the piezoelectric diaphragm (200 kHz) [18] and the AE sensor (300 kHz) [7,27,63], the construction of the sensors (AE sensor housing), and the signal unit. Through the spectrograms generated by the STFT, it is possible to differentiate the cutting conditions of the grinding process. In the results of Figure 4, it was possible to observe the changes in magnitude through the different cutting conditions and continuity through the grinding time. The two sensors presented similar results, showing a good sensitivity to the acoustic activity generated by the ceramic grinding process. The low-cost PZT diaphragm presented stronger color shades, which represents a greater sensitivity to the process stimuli. However, the AE sensor has a higher frequency response band and a signal unit, which filters and establishes gains for the raw signals, which makes the signal more reliable and accurate over time.

Frequency bands were chosen by means of the spectrograms shown in Figure 4. The selection criterion was based on the continuity of the frequency band over time and on the color difference between each grinding condition. As a result, two frequency bands were chosen, one for each sensor. Based on the selection criterion, the (142–147) kHz band was chosen for the piezoelectric diaphragm, while for the AE sensor, the (138–143) kHz band was selected. The frequency bands, chosen directly from the matrices containing the absolute values of each STFT, were used to calculate the ROP metric, which presented characteristics related to the process conditions.

The ROP results, applied in the selected frequency bands for both sensors, are shown in Figure 5. A higher ROP level can be clearly observed as the depth of cut increases. The results of the three slight cutting conditions presented low magnitudes, while moderate and severe conditions showed a greater magnitude and peak-to-valley distance. This behavior results from the accumulation of chip in the cutting region, which generates higher cutting force, AE activity, and surface roughness. Thus, the grinding of ceramic components at a depth of cut of 350 μm generated the highest levels of ROP, STFT, and Ra.

An increase in the ROP level at the beginning and at the end of the machining process can be clearly seen in Figure 5; all the cutting conditions showed the same characteristic. This is a typical behavior of the ceramic grinding process as it represents the first and last contact of the grinding wheel grains with the workpiece surface. The contact grinding area is smaller at the beginning and at the end of the machining process, regions in which the workpiece presents higher mechanical stress and a smaller area of energy propagation. This characteristic can be observed in Figure 2, from 0.0 to 0.2 and 1.0 to 1.2 on the *x*-axis, with color changes representing slight increases in surface roughness. In the middle region of the workpieces, the surface roughness presented more uniform values; these changes become clearer in the depths of cut that represent severe cutting conditions.
