This paper analyses mechanical property prediction through Machine Learning for continuous fiber-reinforced polymer matrix composites printed using the novel Material Extrusion Additive Manufacturing technique. The composite is formed by a nylon-based matrix and continuous fiber (carbon, Kevlar, or fiberglass). From the literature, the elastic modulus and tensile strength were taken along with printing parameters like fiber content, fiber fill type, matrix lattice, matrix fill density, matrix deposition angle, and fiber deposition angle. Such data were fed to several supervised learning algorithms: Ridge Regression, Bayesian Ridge Regression, Lasso Regression, K-Nearest Neighbor Regression, CatBoost Regression, Decision Tree Regression, Random Forest Regression, and Support Vector Regression. The Machine Learning analysis confirmed that fiber content is the most influential parameter in elasticity (E) and strength ($\sigma$). The results show that the K-Nearest Neighbors and CatBoost provided the closest predictions for E and $\sigma$ compared to the other models, and the tree-based model presented the narrowest error distribution. The computational metrics point to a size versus prediction time tradeoff between these two best predictors, and adopting the prediction time as the most relevant criterion leads to the conclusion that the CatBoost model can be considered, when compared to the others tested, the most appropriate solution to work as a predictor in the task at hand.

The monitoring and maintenance of cracked structures are generally carried out using structural integrity assessments. The plastic zone (PZ) crack path (CP) criteria state that a crack grows in a specific direction when the radius of the plastic zone ahead of the crack tip reaches a minimum value. The PZ can be evaluated using stress intensity factors (SIFs). The SIFs under mixed-mode loading were extracted from the literature from three samples: two single edge notch tension (SENT) samples (E = 2.5 GPa, v = 0.38) made from polycarbonate and one modified compact test (C(T)) sample made from low-carbon steel (E = 200 GPa, v = 0.3). In addition, the CP angle was evaluated for the W and R criteria with experimental data, which included non-linear effects such as fatigue-induced plasticity, crack roughness, and debris. It was found that both can predict the CP for lateral cracks in both tested materials and monotonic and cyclic load when the mode mixity does not change considerably from one crack length to the next or goes beyond 0.2. Moreover, the R criterion exhibited an error as high as 1.7%, whereas the W criterion showed a 6% error on the last crack length for the low-carbon steel sample under cyclic load, which had a 100% increase in mode mixity. Finally, the applicability of LEFM was checked, while the CP was sought by finding the size of the PZ.

The loads acting on mechanical components are usually complex, and the more traditional fatigue analysis criteria may be non-conservative, especially in out-of-phase and non-proportional loadings. Many multiaxial fatigue models have been developed and need to be experimentally validated. This work describes the load control mechanism of a low-cost electromechanical multiaxial fatigue testing machine, which applies variable axial and torsional loads. A load cell interconnected to the motor control system was manufactured and instrumented with extensometers. For the axial load, a four-bar mechanism coupled to a power screw was designed to apply alternating axial loads, thus ensuring a higher load frequency and preservation of the motor, which will not need to alternate its direction of rotation. The machine’s control system, focused on out-phase loading, is based on the data sent by the load cell and pulse encoders per degree for each motor, ensuring good accuracy in applied loads. Tests were carried out to check the system’s response, in which it was possible to apply load curves of torque and axial force in different phase shift angle. Therefore, several existing multiaxial fatigue models can be tested with good reliability.

Monitoring and maintenance of cracked structures are generally done using structural integrity assessments. The plastic zone (PZ) crack path (CP) criteria state that a crack grows in the direction when the radius of the plastic zone ahead of the crack tip reaches a minimum value. The PZ can be evaluated using Stress intensity factors (SIF). The SIFs under mixed-mode loading were extracted from literature from three samples: two SENT samples made from polycarbonate and one modified C(T) sample made from low-carbon steel. In addition, the CP angle was evaluated for the W and R-criteria. It was found that both can predict the CP for lateral cracks in both tested materials and monotonic and cyclic load when the mode-mixity does not change considerably from one crack length to the next one or goes beyond 0.2. Moreover, the R-criterion exhibited an error as high as 1.7%, whereas the W-criterion showed a 6% error on the last crack length for the low carbon steel sample under cyclic load, which had a 100% increase in mode-mixity. Finally, the applicability of LEFM is checked, while the CP is sought by finding the size of the PZ.

Modeling cylindrical wave propagation is paramount in several applications, ranging from data transmission through curved barriers to studies in seismic analysis. To obtain solutions, numerical methods, such as the finite element method (FEM), are often employed. However, depending on the frequencies, they may become impractical since very refined meshes could be required, leading to high computational efforts. Within this context, analytical approaches based on transfer matrices emerge as prominent options, since they take advantage of matrix operations. Therefore, the propagation of cylindrical waves in elastic and piezoelectric materials is modeled in this paper using the two-port network method based on ABCD parameters. These parameters were validated by modeling the transmission of acoustic signals through cylindrical barriers using a pair of piezoelectric transducers. In the beginning analysis, the internal radius was varied. At high radii, slight changes were observed in the shape of the resultant waveform, in line with the intuition of a plane wave condition. Further analyses were carried out by comparisons with FEM simulations. Finally, an experimental setup was prepared, and the amplitude-increase behavior predicted in the multi-layer configuration models could be experimentally observed. Overall, good agreement between the results was obtained, validating the proposed lightweight method.

Piezoelectric transducers are part of many devices employed in several applications, such as energy harvesters and sensors. For modeling piezoelectric transducers, one can use numerical methods, such as the finite element method (FEM), or analytical approaches. Among the options for the latter, the two-port network using ABCD parameters emerged as a prominent option since it takes advantage of simple matrix operations. However, this approach is currently only available for transducers polarized in the same direction as their main displacement. Another common polarization is transversal to the direction of its main displacement, which can be found in many sensors for the aircraft industry, rock analysis, as well as biomedical research. Thus, in this paper, the formulation for two-port networks based on ABCD parameters is devised for transversely polarized transducers. New ABCD matrix was derived, allowing the modeling of the d31-mode of vibration. To validate the derived parameters, two different transducers, consisted of PIC144 (hard type) and PIC255 (soft type) materials, were modeled and further compared with finite element simulations and with experiments. Their impedance curves were computed by means of the proposed model and compared with those experimentally obtained with an impedance analyzer and obtained with numerical finite element modeling. Good agreement was observed in the comparisons. In addition, resonance and antiresonance frequency peaks could be correctly modeled for the d31-mode of vibration. Finally, the performance of the analytical model herein proposed was approximately 48.67% faster than FEM; therefore, providing a lightweight method for modeling the transversely polarized transducer.