Applied Sciences

When machine learning supports decision-making in safety-critical systems, it is important to verify and understand the reasons why a particular output is produced. Although feature importance calculation approaches assist in interpretation, there is a lack of consensus regarding how features’ importance is quantified, which makes the explanations offered for the outcomes mostly unreliable. A possible solution to address the lack of agreement is to combine the results from multiple feature importance quantifiers to reduce the variance in estimates and to improve the quality of explanations. Our hypothesis is that this leads to more robust and trustworthy explanations of the contribution of each feature to machine learning predictions. To test this hypothesis, we propose an extensible model-agnostic framework divided in four main parts: (i) traditional data pre-processing and preparation for predictive machine learning models, (ii) predictive machine learning, (iii) feature importance quantification, and (iv) feature importance decision fusion using an ensemble strategy. Our approach is tested on synthetic data, where the ground truth is known. We compare different fusion approaches and their results for both training and test sets. We also investigate how different characteristics within the datasets affect the quality of the feature importance ensembles studied. The results show that, overall, our feature importance ensemble framework produces 15% less feature importance errors compared with existing methods. Additionally, the results reveal that different levels of noise in the datasets do not affect the feature importance ensembles’ ability to accurately quantify feature importance, whereas the feature importance quantification error increases with the number of features and number of orthogonal informative features. We also discuss the implications of our findings on the quality of explanations provided to safety-critical systems. Full article

Accurately determining the nutritional status of plants can prevent many diseases caused by fertilizer disorders. Leaf analysis is one of the most used methods for this purpose. However, in order to get a more accurate result, disorders must be identified before symptoms appear. Therefore, this study aims to identify leaves with excessive nitrogen using one-dimensional convolutional neural networks (1D-CNN) on a dataset of spectral data using the Keras library. Seeds of cucumber were planted in several pots and, after growing the plants, they were divided into different classes of control (without excess nitrogen), N_30% (excess application of nitrogen fertilizer by 30%), N_60% (60% overdose), and N_90% (90% overdose). Hyperspectral data of the samples in the 400–1100 nm range were captured using a hyperspectral camera. The actual amount of nitrogen for each leaf was measured using the Kjeldahl method. Since there were statistically significant differences between the classes, an individual prediction model was designed for each class based on the 1D-CNN algorithm. The main innovation of the present research resides in the application of separate prediction models for each class, and the design of the proposed 1D-CNN regression model. The results showed that the coefficient of determination and the mean squared error for the classes N_30%, N_60% and N_90% were 0.962, 0.0005; 0.968, 0.0003; and 0.967, 0.0007, respectively. Therefore, the proposed method can be effectively used to detect over-application of nitrogen fertilizers in plants. Full article

To minimize the impacts of large losses and optimize the emergency response when a large earthquake occurs, an accurate early warning of an earthquake or tsunami is crucial. One important parameter that can provide an accurate early warning is the earthquake’s magnitude. This study proposes a method for estimating the magnitude, and some of the source parameters, of an earthquake using genetic algorithms (GAs). In this study, GAs were used to perform an inversion of Okada’s model from earthquake displacement data. In the first stage of the experiment, the GA was used to inverse the displacement calculated from the forward calculation in Okada’s model. The best performance of the GA was obtained by tuning the hyperparameters to obtain the most functional configuration. In the second stage, the inversion method was tested on GPS time series data from the 2011 Tohoku Oki earthquake. The earthquake’s displacement was first estimated from GPS time series data using a detection and estimation formula from previous research to calculate the permanent displacement value. The proposed method can estimate an earthquake’s magnitude and four source parameters (i.e., length, width, rake, and slip) close to the real values with reasonable accuracy. Full article

Among people with diabetes mellitus (DM), the two common strategies for decreasing peak plantar pressure (PPP) to reduce diabetic foot ulcers (DFUs) risks are to modify walking speeds and to change insole materials. This study reviewed the PPP reduction based on various walking speeds and insole materials. The articles were retrieved from four major scientific databases and manual search. We identified 1585 articles, of which 27 articles were selected for full-text analysis. We found that in faster walking speeds, the forefoot PPP was higher (308 kPa) than midfoot (150 kPa) and rearfoot (251 kPa) PPP. The appropriate walking speed for reducing the forefoot PPP was about 6 km/h for non-DM and 4 km/h for DM people. The forefoot PPP in DM people was 185% higher than that of non-DM people. Ethylene–vinyl acetate (EVA) insole material was the most popular material used by experts (26%) in the forefoot and reduced 37% of PPP. In conclusion, the suitable walking speed for DM was slower than for non-DM people, and EVA was the most common insole material used to decrease the PPP under the forefoot. The clinicians might recommend DM people to walk at 4 km/h and wear EVA insole material to minimize the DFUs. Full article

Terahertz (THz) radiation sources are increasingly significant for many scientific frontiers, while the generation of THz radiation with high-power at wide-tunable frequencies is still a limitation for most existing methods. In this paper, a compact accelerator-based light source is proposed to produce coherent THz radiation with high pulse energy and tunable frequency from 0.1 THz to 60 THz. By using a frequency beating laser-modulated electron beam and undulator taper, intense coherent THz radiation can be generated through undulators. Theoretical analysis and numerical simulations demonstrate that the proposed technique can generate narrow-bandwidth THz radiation with a pulse energy up to 6.3 millijoule (mJ) and the three-dimensional effects of beam has limited influence on its performance. The proposed technique will open up new opportunities for THz spectroscopic and time-resolved experiments. Full article

This paper presents studies on the dynamic analysis of the ASPIRE robot, which was designed for the medical recovery of brachial monoparesis. It starts from the virtual model of the existing version of the ASPIRE robot, which is analysed kinematically and dynamically by numerical simulations using the MSC.ADAMS software. For this purpose, this paper presents theoretical aspects regarding the kinematics and dynamics of the markers attached to the flexible bodies built in a specifically developed MSC.ADAMS model. Three simulation hypotheses are considered: (a) rigid kinematic elements without friction in couplings, (b) rigid kinematic elements with friction in couplings, and (c) kinematic elements as deformable solids with friction in couplings. Experimental results obtained by using the physical prototype of ASPIRE are presented. Results such as the connecting forces in the kinematic joints and the torques necessary to operate the ASPIRE robot modules have been obtained by dynamic simulation in MSC.ADAMS and compared with those determined experimentally. The comparison shows that the allure of the variation curve of the moment obtained by simulation is similar to that obtained experimentally. The difference between the maximum experimental value and that obtained by simulation is less than 1%. A finite element analysis (FEA) of the structurally optimized flexion/extension robot module is performed. The results demonstrate the operational safety of the ASPIRE robot, which is structurally capable of supporting the stresses to which it is subjected. Full article

Clearances of cages in ball bearings, including pocket and guiding clearances, play a vital role in the stability and reliability of bearings. In this paper, experiments on the cage motion and wear were carried out to investigate the influence of clearances in ball bearings. Firstly, the cages with a series of pocket and guiding clearances were specially designed and tested for prescribed operating conditions on a bearing test rig in which the cage motions were measured, and corresponding wear was also observed. Then, the normalized trajectory, waveform, and spectra of cage motion were constructed and compared to illustrate the effects of clearances on the cage motion and then to establish the relationship between cage motion and wear. Results reveal that the cage motion and wear are both significantly affected by its clearances. The increment of cage guiding clearance makes the whirl trajectories of the cage regular and the motion frequency of cage motion significantly change. However, the increment of cage pocket clearance make the whirl trajectories change from well-defined patterns to complicated ones, and the frequency of cage motion apparently changes. Additionally, the bearing wear is closely related to the cage motion. If the inner ring frequency is of domination for the cage motion, the cage guiding surface will wear seriously. While cage motion is dominated by two times cage frequency in spectrum domain, the cage pocket will wear more seriously. Full article

Search and rescue missions using rotorcrafts need to be reliable all year long, even in winter conditions. In some cases of deployment prior to take off, the crew may need to manually remove accumulated contaminant from the critical surfaces using tools at their disposal. However, icy contaminant may be hard to remove since the rotorcrafts critical surfaces could be cooler than the environment, thus promoting adhesion. Currently, there exists several passive ice protection materials that could reduce the ice adhesion strength and assist the manual de-icing. The aim of this paper is to propose a detailed comparative procedure to assess the ability of materials to assist the manual de-icing of rotorcrafts. The proposed procedure consists of the characterization of materials using several laboratory tests in order to determine their characteristics pertaining to wettability, their icephobic behavior, and finally their assessment under a multi-tool analysis to evaluate if they can assist. The multi-tool analysis uses different mechanical tools, which are currently used during normal operation, to execute a gradual de-icing procedure, which begins with the softest to the hardest tool using a constant number of passes or strokes, under different types of simulated precipitation. Five different materials were used to evaluate the proposed procedure: Aluminum (used as a reference), two silicone-based coatings (Nusil and SurfEllent), an epoxy-based coating (Wearlon), and finally a commercial ski wax (Swix). All of the tested materials could assist the manual de-icing, within a certain limit, when compared to the bare aluminum. However, SurfEllent was the material that obtained the best overall results. This procedure could be easily adapted to different fields of application and could be used as a development tool for the optimization and the assessment of new materials aimed to reduce ice adhesion. Full article

Light entrains human circadian rhythms, but increased time spent indoors and decreased daylight exposure may disrupt human circadian regulation and cause health problems. Much research is focused on improving indoor lighting conditions to minimize the adverse circadian impact of electric lights, and few studies investigate the circadian impact of daylight during the incidental time that people spend outdoors. For instance, when people commute from home to work, they are exposed to daylight. The purpose of this study is to investigate daylight’s impact on commuters’ circadian rhythms. Measurements of the illuminance and the spectral irradiance distribution (SID) of daylight were taken for three modes of commuting: driving, riding on trains, and walking; and under different weather conditions, on different days, and at different locations throughout the summer and autumn in the Sydney metropolitan region in Australia. With the SID data, three metrics were calculated to estimate the circadian impacts: α-opic irradiance, circadian stimulus (CS), and equivalent melanopic lux (EML). The results suggest that driving or walking on sunny or cloudy days and riding trains on sunny days are beneficial for the commuters’ circadian synchronization. Full article

In the past few years, educational data mining (EDM) has attracted the attention of researchers to enhance the quality of education. Predicting student academic performance is crucial to improving the value of education. Some research studies have been conducted which mainly focused on prediction of students’ performance at higher education. However, research related to performance prediction at the secondary level is scarce, whereas the secondary level tends to be a benchmark to describe students’ learning progress at further educational levels. Students’ failure or poor grades at lower secondary negatively impact them at the higher secondary level. Therefore, early prediction of performance is vital to keep students on a progressive track. This research intended to determine the critical factors that affect the performance of students at the secondary level and to build an efficient classification model through the fusion of single and ensemble-based classifiers for the prediction of academic performance. Firstly, three single classifiers including a Multilayer Perceptron (MLP), J48, and PART were observed along with three well-established ensemble algorithms encompassing Bagging (BAG), MultiBoost (MB), and Voting (VT) independently. To further enhance the performance of the abovementioned classifiers, nine other models were developed by the fusion of single and ensemble-based classifiers. The evaluation results showed that MultiBoost with MLP outperformed the others by achieving 98.7% accuracy, 98.6% precision, recall, and F-score. The study implies that the proposed model could be useful in identifying the academic performance of secondary level students at an early stage to improve the learning outcomes. Full article

The objective of the present study was to differentiate and compare the proximate composition, minerals, flesh colour, phenolic composition, and antioxidant activities of varieties of white-fleshed sweet potato (WFSP) and potato (WFP) locally grown in Pakistan. The results showed that WFP presented higher moisture and crude fat content, while WFSP offered better ash, crude protein, and crude fibre contents. Colour analysis revealed that WFSP and WFP showed the highest L* (lightness) values and exhibited the maximum total phenolic content and total flavonoids content of 9.27 ± 0.88 mg GAE/g and 19.01 ± 0.66 mg QE/g. In vitro, results demonstrated that WFSP possessed better antioxidant activity with the highest ferric reducing antioxidant power of 58.67 ± 0.22 mM Fe²⁺/g and DPPH scavenging activity of 39.12 ± 0.33% compared to WFP. It is concluded that WFSP possesses a better proximate and mineral profile followed by higher antioxidant activity. Full article

The determination of a volatile fatty acid content (FOS) and total alkalinity (TAC) can be carried out using Nordmann’s FOS/TAC titration method developed in the 1970s. This two-point titration (pH = 5 and 4.4) can be simply implemented and is widely employed by both the academic and industrial worlds. However, the present study proves that Nordmann’s method is only valid in limited ranges, since the titration of one FOS and TAC has an impact on the determination of the other, especially in extreme conditions. The present work develops a numerical tool with Scilab simulating the acid–base equilibria of titration. The program is efficient in predicting the experimental equivalent volumes obtained from Nordmann’s method with different combinations of sodium acetate and sodium bicarbonate contents. The mean absolute percentage errors (MAPE) between the simulation and experiment are below 7%. Two new formulas are developed, considering both equivalent volumes at pH = 5 and 4.4 to calibrate FOS and TAC values. The proposed formulas show their good performance in predicting various combinations of FOS and TAC contents in an anaerobic digestate at TAC ranging from 0 to 20,000 mg CaCO₃·L⁻¹ and FOS ranging from 0 to 31,000 mg HAc·L⁻¹. Full article

Log-structured merge-tree (LSM-Tree)-based key–value stores are attracting attention for their high I/O (Input/Output) performance due to their sequential write characteristics. However, excessive writes caused by compaction shorten the lifespan of the Solid-state Drive (SSD). Therefore, there are several studies aimed at reducing garbage collection overhead by using Zoned Namespace ZNS; SSD in which the host can determine data placement. However, the existing studies have limitations in terms of performance improvement because the lifetime and hotness of key–value data are not considered. Therefore, in this paper, we propose a technique to minimize the space efficiency and garbage collection overhead of SSDs by arranging them according to the characteristics of key–value data. The proposed method was implemented by modifying ZenFS of RocksDB and, according to the result of the performance evaluation, the space efficiency could be improved by up to 75%. Full article

In recent movie recommendations, one of the most important issues is to predict the user’s sequential behavior to be able to suggest the next movie to watch. However, capturing such sequential behavior is not easy because each user’s short-term or long-term behavior must be taken into account. For this reason, many research results show that the performance of recommending a specific movie is not good in a sequential recommendation. In this paper, we propose a cluster-based method for classifying users with similar movie purchase patterns and a movie genre prediction algorithm rather than the movie itself considering their short-term and long-term behaviors. The movie genre prediction does not recommend a specific movie, but it predicts the genre for the next movie to watch in consideration of each user’s preference for the movie genre based on the genre included in the movie. Using this, it will be possible to provide appropriate guidelines for recommending movies including the genres to users who tend to prefer a specific genre. In particular, in this study, users with similar genre preferences are organized into clusters to recommend genres. For clusters that do not have relatively specific tendencies, genre prediction is performed by appropriately trimming genres that are not necessary for recommendation in order to improve performance. We evaluate our method on well-known movie data sets and qualitatively determine that it captures personalized dynamics and is able to make meaningful recommendations. Full article

Increasing power demand from passive distribution networks has led to deteriorated voltage profiles and increased line flows. This has increased the annual operations and installation costs due to unavoidable reinforcement equipment. This work proposes the reduction in annual costs by optimal placement of capacitors used to alleviate power loss in radial distribution networks (RDNs). The optimization objective function is formulated for the reduction in operation costs by (i) reducing the active and reactive power losses, and (ii) the cost and installation of capacitors, necessary to provide the reactive power support and maintain the voltage profile. Initially, the network buses are ranked according to two loss sensitivity indices ($LSIs$), i.e., active loss sensitivity with respect to node voltage ($LS{I}_1$) and reactive power injection ($LS{I}_2$). The sorted bus list is then fed to the particle swarm optimization (PSO) for solving the objective function. The efficacy of the proposed work is tested on different IEEE standard networks (34 and 85 nodes) for different use cases and load conditions. In use case 1, the values finalized by the algorithm are selected without considering their market availability, whereas in use case 2, market-available capacitor sizes close to the optimal solution are selected. Furthermore, the static and seasonal load profiles are considered. The results are compared with recent methods and have shown significant improvement in terms of annual cost, losses and line flows reduction, and voltage profile. Full article

Otosclerosis is a common middle ear disease that requires a combination of examinations for its diagnosis in routine. In a previous study, we showed that this disease could be potentially diagnosed by wideband tympanometry (WBT) coupled with a convolutional neural network (CNN) in a rapid and non-invasive manner. We showed that deep transfer learning with data augmentation could be applied successfully on such a task. However, the involved synthetic and realistic data have a significant discrepancy that impedes the performance of transfer learning. To address this issue, a Gaussian processes-guided domain adaptation (GPGDA) algorithm was developed. It leveraged both the loss about the distribution distance calculated by the Gaussian processes and the loss of conventional cross entropy during the transferring. On a WBT dataset including 80 otosclerosis and 55 control samples, it achieved an area-under-the-curve of $97.9\pm 1.1$ percent after receiver operating characteristic analysis and an F1-score of $95.7\pm 0.9$ percent that were superior to the baseline methods ($r=10$, $p<0.05$, ANOVA). To understand the algorithm’s behavior, the role of each component in the GPGDA was experimentally explored on the dataset. In conclusion, our GPGDA algorithm appears to be an effective tool to enhance CNN-based WBT classification in otosclerosis using just a limited number of realistic data samples. Full article

Singing voice detection is still a challenging task because the voice can be obscured by instruments having the same frequency band, and even the same timbre, produced by mimicking the mechanism of human singing. Because of the poor adaptability and complexity of feature engineering, there is a recent trend towards feature learning in which deep neural networks play the roles of feature extraction and classification. In this paper, we present two methods to explore the channel properties in the convolution neural network to improve the performance of singing voice detection by feature learning. First, channel attention learning is presented to measure the importance of a feature, in which two attention mechanisms are exploited, i.e., the scaled dot-product and squeeze-and-excitation. This method focuses on learning the importance of the feature map so that the neurons can place more attention on the more important feature maps. Second, the multi-scale representations are fed to the input channels, aiming at adding more information in terms of scale. Generally, different songs need different scales of a spectrogram to be represented, and multi-scale representations ensure the network can choose the best one for the task. In the experimental stage, we proved the effectiveness of the two methods based on three public datasets, with the accuracy performance increasing by up to 2.13 percent compared to its already high initial level. Full article

In order to study the shear behavior of the interface between sand and structure, a series of shear tests were carried out using an HJ-1 ring shear apparatus (Nanjing, China). First, through the monotonic shear tests, the loose sand and dense sand were sheared at the steel interface with different roughnesses. The results showed that when the interface was relatively smooth, the shear stress–shear displacement curves of loose sand and dense sand both exhibit strain hardening characteristics. When the interface was rough, the dense sand showed strain softening. The initial shear stiffness of the sand–steel interface increased with the increase in normal stress, interface roughness, or sand relative density. Then, considering the influence of initial shear stress, through the cyclic shear test, this work analyzed the shape of the loading and unloading curves and the development law of cumulative normal deformation, and discussed the change of loading and unloading shear stiffness under different stress level amplitudes and the residual deformation generated during the cycle. The research results showed that loose sand and dense sand generally shrunk in volume during the cycle. The initial loading process was similar to the case of static loading. In the later dynamic loading process, the shear shrinkage per cycle was relatively small and continued to develop. Additionally, it was found that the unloading stiffness of the sand–steel interface is always greater than the initial loading stiffness. As the number of cycles increases, the loading stiffness increases, and it may eventually approach the unloading stiffness. Full article

In this paper, the authors present the relationship between the assumptions of ergonomics in the work of a strawberry picker and quality of picked fruit. The body posture that a person adopts while working has a significant impact on their health, working comfort, and productivity, but also on the quality of the fruit that is harvested. This paper identifies three characteristic picker positions during strawberry harvesting. A synchronized surface electromyography (sEMG) instrument together with the Tekscan^® surface pressure measurement system allowed for the determination of the influence of working position on changes in the load of the picker’s musculoskeletal system and the surface pressure exerted on the fruit during manual strawberry picking, which are decisive factors for maintaining fruit quality. In addition, compression tests on whole strawberry fruit were carried out as a benchmark to evaluate and compare the maximum forces as well as the destructive pressures on the fruit. From the tests, we found that the most comfortable position of the worker’s body was determined along with the harvesting technique (position during work) that has the least negative effect on the quality of the harvested fruit. Consequently, the level of dynamic load on the worker was determined. Full article