*2.1. Passive Monitoring Techniques*

The grinding process is a complex manufacturing process influenced by many factors, such as the workpiece, the grinding machine, the grinding wheel and the process parameters [15]. Due to the complexity of the process, grinding monitoring is very important for its automation and optimization. The use of different sensors in the passive configuration, such as the acoustic emission sensor and the piezoelectric diaphragm along with digital signal processing techniques, allows the detection of events in the grinding process and other industrial processes. Some advantages of the piezoelectric diaphragm can be herein highlighted as follow. The cost of the sensor is one of the main factors that makes the piezoelectric diaphragm more attractive than the acoustic emission sensor. This transducer has a very

simple construction, consisting of a thin layer of piezoelectric material (active element) adhered to a circular metal plate (diaphragm). The ceramic is coated with a thin metallic film (usually silver) that acts as an electrode [12]. The PZT diaphragms can operate in both active and passive configurations and have an average cost of a few cents versus the high average cost of AE sensors, which range from hundreds to thousands of dollars [18]. In addition, they are compact, flexible, lightweight and simple acoustic components widely used in various electronic devices to produce sound (alarm, ringing and beep) [12,19]. Furthermore, the fastening of the PZT sensor generally requires only a thin layer of cyanoacrylate glue, while for some AE sensors it may be necessary to machine the holder in order to fit the sensor, which is difficult in some cases. In addition, the data acquisition system for the piezoelectric diaphragm has a lower cost compared to the system required for the AE sensor (this is mainly due to the lower frequency response of the PZT sensor). Finally, the PZT diaphragm is highly available as it can be easily bought by users. Although the sensitivity of the low-cost piezoelectric diaphragm is generally lower than the acoustic emission sensors, it can be considered as an alternative for monitoring several applications, as presented in Castro et al. [20], Viera et al. [21] and Ribeiro et al. [16].

The monitoring of the grinding wheel wear has been the subject of several studies since a worn grinding wheel can directly influence on the geometry and quality of the ground surfaces [18]. The most widely used sensors for grinding wheel wear monitoring (macro and micro effects) are acoustic emission, temperature, power and force sensors. The acoustic emission sensor is used for various purposes, such as monitoring the grinding wheel condition during the grinding and dressing process and detecting contact between the grinding wheel and the workpiece [22]. In this context, de Oliveira et al. [23] presented a monitoring system for the grinding process based on the analysis of the acoustic emission signal (RMS) and the construction of acoustic maps. This innovative system enables the evaluation of the dressing process, the topographic mapping of the grinding surface, as well as the interaction between the grinding wheel and the workpiece during the grinding operation.

The approach in which the PZT (buzzer) and the acoustic emission sensor are employed in conjunction with digital signal processing techniques can be found in many studies, where the objective is the monitoring of the grinding process and related processes (dressing). In order to detect the workpiece burning during the grinding process, Ribeiro et al. [16] collected the acoustic activity of the grinding process with a piezoelectric diaphragm and a commercial acoustic emission sensor. The results demonstrate that the selection of frequency bands optimized the use of metrics, such as the RMS and RMSD, in the detection of the burning phenomenon during grinding. In addition, it demonstrates that the piezoelectric diaphragm can achieve similar results to the AE sensor in detecting surface burns on ground workpieces.

A comparative study using a low-cost piezoelectric diaphragm and a commercial acoustic emission sensor, along with digital signal processing techniques in the evaluation of the surface quality of ground ceramic workpieces, was developed by Viera et al. [21,24]. The results showed that the proposed technique (statistic ROPSTFT) was able to detect roughness variations on the surfaces of the ground workpieces.

Castro et al. [20] employed the piezoelectric diaphragm as an alternative to the acoustic emission sensor in the detection of partial discharges in power transformers. The results showed that the low-cost sensor has low sensitivity when compared to the conventional acoustic emission sensor, however, for the application studied, the collected signals from both sensors had similar behavior, which characterizes its application in the detection of partial discharges in power transformers.

### *2.2. Active Monitoring with Emission-Reception Techniques*

The elastic waves, when propagated in a material that presents intrinsic changes, are dispersed in all directions. The active monitoring methods employ fixed transducers at specific locations of the structure to generate and collect elastic waves after propagating through the material [25]. The active monitoring techniques typically employ ultrasonic [25] and piezoelectric transducers [26]. The following sections describe the most important piezoelectric-active techniques applied to processes monitoring.

### 2.2.1. Electromechanical Impedance Method (EMI) and Frequency Response Function Method (FRF)

The electromechanical impedance method (EMI) has been widely studied for the development of structural health monitoring systems [27]. According to Castro et al. [28], the EMI technique is based on the piezoelectric effect, which allows the monitoring of structural conditions through the electrical impedance of the transducer connected to a structure, since the electrical impedance of the piezoelectric transducer is related to the mechanical properties of the structure. The EMI technique can be applied to different types of materials, such as metallic materials and composites. For example, highlighting the works of Na et al. [29] on composite structure, Annamdas et al. [30] on concrete structure and Zhu et al. [31] on steel structure, thus indicating the wide range of applications. Structural damage is typically characterized using damage indexes. According to Budoya et al. [32], the most common damage indexes are based on the comparison between two electrical impedance signatures, where one of them is obtained when the structure is considered healthy. The most common indices in SHM analysis are the root mean square deviation (RMSD) and the correlation coefficient deviation metric (CCDM) [33]. The RMSD is indicated to quantify the amplitude differences between two spectrums (baseline and damage) [34], while the CCDM is appropriate to measure the variations or displacement between the frequencies of the analyzed signals [35].

In order to assess the damage present in ground 1020 steel workpieces, Marchi et al. [15] obtained the electromechanical impedance levels for the workpiece without damage and after the grinding process, which made the comparison possible. To assess the damage caused by the grinding process, the results of the RMSD and CCDM indexes were compared with the microhardness and surface roughness of the workpieces.

de Oliveira Conceição Jr. et al. [18,36] assessed the structural changes caused by wear in single-point dressing tools during their lifetime employing EMI method to ensure the reliable monitoring of the tool condition in grinding operations. Representative damage indices, such as RMSD and CCDM, obtained from impedance signatures at different frequency bands were computed for diverse tool conditions. Moreover, an intelligent system, implemented on the basis of artificial neural networks, was able to select the most damage-sensitive features based on the optimal frequency band. The authors highlighted that the EMI method was capable of effectively detect damages in the relatively small diamond tool tips showing a general overall classification error lower than 2% for the best neural models.

In contrast to the conventional EMI method, which is known for using a single transducer, the frequency response function (FRF) method uses at least two transducers, each operating separately as an actuator and as a sensor [37]. Two transducers are employed in this method, one in the emitter mode and the other one in the receiver mode. A significant change caused by incipient damage (detachment, rust, cracks, etc.) results in a difference in the FRF response for an application of the electromechanical impedance technique in an electronic circuit equipped with piezoelectric elements [38]. Liang et al. [39] employed the FRF method in the study of the loosening monitoring of threaded plumbing structures. The use of this method along with the RMSD index allowed the correlation between the RMSD values and the severity of the loosening found in the threaded structures.

### 2.2.2. Transmitter-Receiver Arrangements for Ultrasonic Inspection

The ultrasonic inspection of a component can be performed through two methods—active sensing and passive sensing [40]. In active sensing, a transmitter and a receiver are attached to the structure of interest. In this configuration, the transmitter is responsible for transmitting the signal and the receiver is responsible for receiving the signal. The presence of damage in the region between the transmitter and the receiver causes changes in the ultrasound signal. The damaged region can be identified by means of the analysis of the received signal [41]. The three most common configurations used in ultrasound wave analysis are shown in Figure 1.

**Figure 1.** Three most common transmission-reception configurations: (**a**) pulse-echo; (**b**) pitch-catch; and (**c**) through-transmission [41].

Figure 1a,c show configurations based on bulk waves. This type of wave is used when analyzing defects in the internal region of the sample. Figure 1b shows the configuration in which surface waves are used, which detect anomalies on the sample surface. In this work, bulk waves are prevalent due to the transducer configuration, which is similar to the configuration of Figure 1c.

In the pulse-echo system (Figure 1a) the transducer is positioned on the surface of the workpiece. In this method, the same transducer functions as a transmitter and receiver of ultrasound waves [42]. The use of this system for the monitoring of the temperature distribution in the inner region of materials was presented by Ihara et al. [43,44]. The authors applied the proposed technique on heated steel discs at different temperatures. The results prove the feasibility of the technique, the values obtained were consistent with the values measured by thermocouples installed in the sample, thus indicating the effectiveness of the technique.

In the pitch-catch technique, the system has two transducers, the first emits the ultrasound pulses while the second transducer detects the pulses, as shown in Figure 1b. The transducers can be aligned in different ways, such as in normal incidence or oblique incidence, that is, with a certain inclination angle. The technique allows the monitoring of local damages through the emission of known input signals, adequate to diagnose a structure regarding the existence of damage [40]. A study was developed by Ihn et al. [40] employing the pitch-catch technique in the structural health monitoring of an airplane fuselage. The results showed that the use of a damage index allowed the evaluation of damage, similar to the results obtained by the Eddy current method or by the ultrasonic scanning.

The through-transmission technique, similar to the pitch-catch technique, also uses two transducers; however, the positioning of the transducers in the structure is different, as shown in Figure 1c. This technique uses two transducers positioned on opposite sides of the sample. A pulse generator is used to generate continuous electrical pulses with specific frequencies and amplitudes, which are converted into ultrasound waves by the first transducer (transmitter). The ultrasound waves, propagated through the sample, are detected by the second transducer (receiver) [42]. In Raišutis et al. [45], the through-transmission technique was used to estimate the phase velocity dispersion of ultrasound waves in plastic materials. The monitoring of lactic acid fermentation by means of the through-transmission technique was presented by Resa et al. [46]. Through the ultrasonic velocity, it was possible to detect the changes that occurred during the fermentation of carbohydrates. Thus, this non-invasive technique presents great potential in the monitoring of biological processes.

### **3. RMS and Counts in AE Signal Processing**

One of the most used statistics in the analysis of the acoustic emission signal is the root mean square (RMS). The application of the RMS statistic in the identification of failures in the grinding and dressing processes has been extensively studied in the specific literature [47–50]. According to Webster et al. [51], the best integration interval for calculating the RMS statistic in the monitoring of the grinding process is 1 ms. The RMS statistic is defined by Equation (1).

$$X\_{rms} = \sqrt{\frac{1}{N} \* \sum\_{i=1}^{N} x^2(i)}\tag{1}$$

where *x* is the raw signal and *N* is the number of discrete samples (*i*) considered in the calculation.

On the other hand, the Counts statistic was defined by Lopes et al. [9] as the number of times that the signal crosses a threshold per unit of time. The Counts statistic, calculated from the AE signal, has often been implemented in scientific works with different purposes, such as the location and measurement of acoustic emission events in snow blocks before their fracture [52] and the on-line monitoring of abrasive water cutting processes [53].

With regard to the signal analysis, the frequency domain is generally used in identifying events that are difficult to observe in the time domain. The most commonly used frequency domain techniques are the fast Fourier transform (FFT) [54], the short-time Fourier transform (STFT) [55] and the wavelet transform (WT) [56]. The FFT is an algorithm developed to improve the computational efficiency of the discrete Fourier transform (DFT), being a popular method used in spectral analysis and digital signal processing. Ahirrao et al. [57] used the FFT in the analysis of the vibration and dynamics of engines. Kang et al. [58] used the FFT in the tool condition analysis and monitoring of the micro-lens fabrication process.

The statistics presented in this section for the acoustic emission signals can also be used for signals of different natures acquired by other sensors, such as accelerometer and PZT. This work proposes the application of the RMS and Counts statistics in the signals sampled by a set of piezoelectric diaphragms in the emitter-receiver configuration for the monitoring of the material removal in the grinding process. It is worth mentioning that the implementation of the piezoelectric diaphragm in this configuration and for this purpose is unprecedented in the literature.

### **4. Bases of the Chirp-through-Transmission Ultrasound Technique**

Nowadays, the inspection of structures by means of ultrasound techniques is widely used. A continuous effort of researchers and industries has been made in order to improve and increase the applicability of non-destructive evaluations (NDE). Thus, a new inspection technique based on ultrasound waves and low-cost piezoelectric transducers is presented in this paper. Section 4.1 presents some basic concepts of ultrasound waves, while Section 4.2 presents some common classifications of these waves. Finally, Section 4.3 presents the method proposed in this paper.

### *4.1. Ultrasound Waves and Their Parameters*

Ultrasound signals are used to predict the behavior of materials (detecting internal defects) and to characterize them in a variety of structures. The interest of the scientific community for ultrasound techniques has increased in recent years due to its wide range of applications [41]. Immersion testing and ultrasound contact methods are very popular NDE techniques for characterizing material conditions [59]. However, this type of technique requires the removal of the structure, causing significant interruptions in the operation of the machines [60].

Ultrasound waves are composed of frequencies greater than the human's hearing range, which is typically 20 kHz [41,42,61]. The main characteristics of the ultrasound waves are length, velocity, pressure, frequency and period. The propagation of the waves in the structure results in two

phenomena—velocity alteration and wave attenuation, both caused by the mechanisms of absorption and dispersion [42]. As the ultrasound waves propagate through the material, their energy is diminished and reflections are generated on the surface. These reflections are used to determine the presence and location of discontinuities and defects [61].

The wave velocity is the product between the frequency and the wavelength, so high-frequency waves have shorter wavelengths while low-frequency waves have longer wavelengths [42]. The velocity of the ultrasound waves (*v*) is determined by the density (ρ) and elasticity (*E*) of the propagation medium according to the Newton-Laplace equation.

$$v = \sqrt{\frac{E}{\rho}}\tag{2}$$

Equation (2) implies that the velocity of an ultrasound wave in a solid material is greater than in a liquid [42].

There are other ultrasound parameters that correlate the physical-chemical properties of the materials, such as the attenuation coefficient and the acoustic impedance. The attenuation is a result of the energy reduction caused by the compression and decompression of the ultrasonic waves due to the absorptions and dispersions of the propagation medium [62]. Absorption is mainly associated with homogeneous materials while dispersion is related to heterogeneous materials. In addition, attenuation is affected by viscosity, compressibility, material contours and absorption and dispersion effects [42]. It is worth mentioning that the attenuation coefficient of a given material is highly dependent on the way in which the material was manufactured, being useful for product quality control.

Acoustic impedance is the product between the density and the velocity of the wave that passes through the contours of different materials. Materials with different densities have different acoustic impedances, resulting in reflections on the contours between two materials with different acoustic impedances. The attenuation and the acoustic impedance are expressed by Equations (3) and (4), respectively.

$$A = A\_{\mathcal{O}} \mathfrak{e}^{-\text{ax}} \tag{3}$$

$$R = \frac{A\_T}{A\_t} = \frac{z\_1 - z\_2}{z\_1 + z\_2} \tag{4}$$

where *Ao* is the initial amplitude of the wave, *x* is the distance traveled; *R* is the ratio of the amplitude of the reflected wave (*AT*) to the incident wave (*At*); *z*<sup>1</sup> and *z*<sup>2</sup> are the acoustic impedances of two materials.
