An aircraft is a multilayer structure that is assembled by rivets. Under extreme working conditions, corrosion appears and quickly propagates at the rivet sites of the layers; thus, it threads the integrity and safety of the aircraft. Corrosion usually occurs at the hidden layer around the rivet, making it difficult to detect. This paper proposes a machine learning approach incorporating an electromagnetic testing system to detect the hidden corrosion at the riveting site effectively. Several machine learning methods will be investigated for the detection of different sizes and locations of corrosion. The training strategy of the machine-learning models on the small numbers of data will also be investigated. The result shows that the proposed approach could effectively detect 89.48% of the hidden corrosion having from 2.8 to 195.4 mm³ with only 20% of training data and could be increased to 99.0% with 60–80% of the training data.

Pressure reducing valves (PRVs) are commonly used for pressure control in water distribution systems (WDSs) by means of dissipating the pressure excess. The use of pumps as turbines (PATs) is an alternative and more favorable system since they not only control the system pressure to decrease water leakage, but also utilize the pressure excess to generate electrical energy. The optimal localization of PATs can be casted into a mixed-integer nonlinear program (MINLP) where binary variables are used to represent the presence of PATs on links. Most of the available MINLP models for optimal PAT localization adopted the optimization approaches for PRV localization without considering the bound constraints on flow rates and heads of PATs. As a result, such an optimization model may make PATs delivering a non-desired output. In this paper, we propose a new MINLP model for optimal PAT localization. Instead of using a constraint on the maximum number of PATs to be placed in a WDS, new constraints relating to the minimum power generated by PAT are introduced to find links having adequate flows and head drops for placing PATs. Moreover, constraints are used to restrict flows and heads of PATs to their feasible operating range, so that the problem can be efficiently solved. The proposed MINLP model is applied to the optimal localization of PATs for a WDS benchmark and a real-world WDS in Vietnam. The results demonstrate that the new MINLP model can efficiently identify optimal locations for PAT placement where the specified working range and minimum power generated by the PATs are ensured.

A single-layer design of non-uniform metasurface (MS) based circularly polarized (CP) antenna with wideband operation characteristic is proposed and investigated in this paper. The antenna is excited by a truncated corner squared patch as a primary radiating CP source. Then, a non-uniform MS is placed in the same layer with the driven patch. Besides increasing the impedance bandwidth, the non-uniform MS also generates two additional CP bands in the high frequency band, leading to significantly increase the antenna’s overall performances. The use of non-uniform MS distinguishes our design from the other CP MS based antennas in literature, in which the MS is formed by multiple uniform unit cells and placed in different layer with the radiating element. For validation, an antenna prototype, whose overall dimensions of 0.94 at the center operating frequency, is fabricated and experimentally tested. The measured operating bandwidth with |S dB and dB is 27.1% (5.1–6.7 GHz) and the broadside gain within this band is from 5.7 to 7.2 dBic. Compared to the other related works, the proposed antenna has advantage of very wideband operation with single-layer design.

This paper presents a modification of the deep Q-network (DQN) in deep reinforcement learning to control the angle of the inverted pendulum (IP). The original DQN method often uses two actions related to two force states like constant negative and positive force values which apply to the cart of IP to maintain the angle between the pendulum and the Y-axis. Due to the changing of too much value of force, the IP may make some oscillation which makes the performance system could be declined. Thus, a modified DQN algorithm is developed based on neural network structure to make a range of force selections for IP to improve the performance of IP. To prove our algorithm, the OpenAI/Gym and Keras libraries are used to develop DQN. All results showed that our proposed controller has higher performance than the original DQN and could be applied to a nonlinear system.