Testing Phase Research Articles

Preparation and NH3 adsorption behavior of porous magnesium oxychloride cement-based SrCl2

In addressing challenges in Selective Catalytic Reduction (SCR) systems in vehicles, the development of novel SrCl2 composite materials with high ammonia adsorption and structural stability is crucial. In this context, we employed the porous magnesium oxychloride cement (PMOC) as a carrier material, successfully fabricating a mesoporous PMOC-SrCl2 material via solution impregnation technology. Simultaneously, we employed various characterization techniques such as XRD, full-pore analysis, TG, SEM, TEM, and EDS to examine the material. We conducted static adsorption assessments on numerous candidate materials at 0.1Mpa and 293.15 K. The synthesized C3 series material (Mass ratio of SrCl2 to PMOC = 10:3), solution impregnated for 1 h, exhibited exceptional ammonia adsorption capacity (up to 38.01 mmol·g−1). Through model fitting of experimental data, we discovered that the adsorption process of this material closely aligns with the pseudo-second-order kinetic model and the Langmuir model. Over an adsorption time span of 180 min, the adsorption rate of the composite material increased by 129 % compared to pure SrCl2. In the cycle performance testing phase, we observed that the material exhibited no significant degradation in performance after undergoing 10 cycles of adsorption/desorption. This structurally stable, cost-effective, and readily accessible high-efficiency ammonia adsorption material undoubtedly possesses immense potential to solve the application challenges of ammonia adsorption materials within automotive selective catalytic reduction systems.

Improving stability and performance of spiking neural networks through enhancing temporal consistency

Spiking neural networks have gained significant attention due to their brain-like information processing capabilities. The use of surrogate gradients has made it possible to train spiking neural networks with backpropagation, leading to impressive performance in various tasks. However, spiking neural networks trained with backpropagation typically approximate actual labels using the average output, often necessitating a larger simulation timestep to enhance the network’s performance. This delay constraint poses a challenge to the further advancement of spiking neural networks. Current training algorithms tend to overlook the differences in output distribution at various timesteps. Particularly for neuromorphic datasets, inputs at different timesteps can cause inconsistencies in output distribution, leading to a significant deviation from the optimal direction when combining optimization directions from different moments. To tackle this issue, we have designed a method to enhance the temporal consistency of outputs at different timesteps. We have conducted experiments on static datasets such as CIFAR10, CIFAR100, and ImageNet. The results demonstrate that our algorithm can achieve comparable performance to other optimal SNN algorithms. Notably, our algorithm has achieved state-of-the-art performance on neuromorphic datasets DVS-CIFAR10 and N-Caltech101, and can achieve superior performance in the test phase with timestep T = 1.

Robust Multimodal Representation under Uncertain Missing Modalities

Multimodal representation learning has gained significant attention across various fields, yet it faces challenges when dealing with missing modalities in real-world applications. Existing solutions are confined to specific scenarios, such as single-modality missing or missing modalities in test cases, thereby restricting their applicability. To address a more general scenario of uncertain missing modalities in both training and testing phases, we propose Robust Multimodal Representation under Uncertain Missing Modalities (RMRU). This framework projects each modality's representation into a shared subspace, enabling the reconstruction of any missing modalities within a unified model. We propose an interaction refinement module that utilizes cross-modal attention to enhance these reconstructions, particularly beneficial in scenarios with limited complete modality data. Furthermore, we introduce an iterative training strategy that alternately trains different modules to effectively utilize both complete and incomplete modality data. Experimental results on four benchmark datasets demonstrate the superiority of RMRU over existing baselines, particularly in scenarios with a high rate of missing modalities. Remarkably, our proposed RMRU can be broadly applied to diverse scenarios, regardless of modality types and quantities.

A smarter approach to liquefaction risk: harnessing dynamic cone penetration test data and machine learning for safer infrastructure

This paper presents a novel approach for assessing liquefaction potential by integrating Dynamic Cone Penetration Test (DCPT) data with advanced machine learning (ML) techniques. DCPT offers a cost-effective, rapid, and adaptable method for evaluating soil resistance, making it suitable for liquefaction assessment across diverse soil conditions. This study establishes a threshold criterion based on the ratio of the penetration rate to the dynamic resistance (e/qd), where values exceeding four indicate high liquefaction susceptibility. ML models, including Support Vector Machine (SVM) optimized with Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), Genetic Algorithm (GA), and Firefly Algorithm (FA), were employed to predict the e/qd ratio using key geotechnical parameters, such as fine content, peak ground acceleration, reduction factor, and penetration rate. The SVM-PSO model demonstrated superior performance, with high R2 values of 0.999 and 0.989 in the training and testing phases, respectively. The proposed methodology offers a sustainable and accurate approach for liquefaction assessment, reducing the environmental impact of geotechnical investigations, while ensuring reliable predictions. This study bridges the gap between field testing and advanced computational techniques, providing a powerful tool for geotechnical engineers to assess liquefaction risks and design resilient infrastructures.

Tooth numbering with polygonal segmentation on periapical radiographs: an artificial intelligence study.

Accurately identification and tooth numbering on radiographs is essential for any clinicians. The aim of the present study was to validate the hypothesis that Yolov5, a type of artificial intelligence model, can be trained to detect and number teeth in periapical radiographs. Six thousand four hundred forty six anonymized periapical radiographs without motion-related artifacts were randomly selected from the database. All periapical radiographs in which all boundaries of any tooth could be distinguished were included in the study. The radiographic images used were randomly divided into three groups: 80% training, 10% validation, and 10% testing. The confusion matrix was used to examine model success. During the test phase, 2578 labelings were performed on 644 periapical radiographs. The number of true positive was 2434 (94.4%), false positive was 115 (4.4%), and false negative was 29 (1.2%). The recall, precision, and F1 scores were 0.9882, 0.9548, and 0.9712, respectively. Moreover, the model yielded an area under curve (AUC) of 0.603 on the receiver operating characteristic curve (ROC). This study showed us that YOLOv5 is nearly perfect for numbering teeth on periapical radiography. Although high success rates were achieved as a result of the study, it should not be forgotten that artificial intelligence currently only can be guides dentists for accurate and rapid diagnosis. It is thought that dentists can accelerate the radiographic examination time and inexperienced dentists can reduce the error rate by using YOLOv5. Additionally, YOLOv5 can also be used in the education of dentistry students.

Psychometric Properties of the Hungarian Passion Scale-8 (PS-8-HU) in Physically Active and Inactive Adults.

The 8-item Passion Scale (PS-8) is a unidimensional instrument used in dual versus single passion research. The PS-8 was validated in Icelandic, Persian, and Turkish student samples. In this study, we translated the PS-8 and validated it with physically active and inactive adults from Hungary. We also evaluated measurement invariance (across gender, physical activity, exercise frequency, and age groups) of the Hungarian version of the Passion scale (PS-8-HU). Data were collected online between fall 2023 and winter 2024, including a test and retest phase. The sample comprised 729 Hungarian physically active and inactive adults (68% females, 32% males; ages 18-78 years), of whom 196 participated in the retest. We used confirmatory factor analysis (CFA), item analysis, correlations, t-test and ANOVA in checking the scale's validity and reliability, and we used a multigroup CFA to test the measurement invariance.The results supperted a univariate structure of the PS-8-HU. The instrument was invariant for gender (at the strict level), exercise frequency (metric), physical activity (metric), and age (scalar). The internal reliability of the PS-8-HU was .93 and test-retest reliability was .77. The PS-8-HU also exhibited good concurrent and construct validities, high homogeneity, and adequate discriminant validity.

A novel spatial-aware deep learning approach for exploring the environmental context of terrorist attacks and armed conflicts

The quantitative assessment of terrorist attacks and armed conflicts (TAACs) is a crucial component of global public safety research and is vital for societal stability and national security. This study addresses the spatial dependency of such events, i.e., the relationship between the outbreak of an event and its environment. Based on geographic big data and artificial intelligence (AI), we propose a spatial feature utilization pattern that takes into account the impact of the event environment, and established a deep learning (DL) framework of features within the joint event location and space neighborhood to improve the precision of the quantitative assessment. The results demonstrate that in scenarios under a combination of 14 social, natural, and geographic driving factors, models that incorporate spatial features outperform those that only use location features during both the training and testing phases. Furthermore, models that consider both location and spatial features outperform models using only a single feature across various evaluation metrics. Global attribution analysis further confirms the spatial dependency of events, manifested in the mutual influence on the likelihood of events occurring among adjacent cities and the correlation with various environmental factors, particularly elements related to human activities and living environments. We find that both prosperous urban centers and underdeveloped rural areas are hotspots for TAACs, and that such events more likely to occur in harsh climatic patterns characterized by high temperatures and low precipitation. This enhances our understanding and preparedness for managing and preventing such events.

Redesain UI UX Website Lemonder Indonesia Menggunakan Metode Design Thinking Untuk Meningkatkan Media Promosi

Lemonder Indonesia's website is designed to compete in the market for lemons, vegetables, and other high-quality fresh products. This website provides easy access to information about the company and its products for all users. The purpose of this research is to increase customer trust and improve the design of the Lemonder Indonesia website. The Design Thinking approach is used as a method for redesigning the website. This method includes five stages: Empathy, Define, Ideate, Prototype, and Testing. It focuses on consumer-centered innovation, emphasizing product and service designs that attract attention in the business world. During the testing phase, this research uses usability testing to evaluate the effectiveness of the website redesign. The System Usability Scale (SUS) involves using a questionnaire distributed via Google Forms. Before the redesign, the SUS score was 51.75%, which is a grade D with a "poor" indicator. However, after the redesign, the SUS score significantly increased to 85.08%, which is a grade A with an "excellent" indicator. This result demonstrates that the SUS testing in this research was successful and that the new design is well-received by users.

Developing Independent Living Support for Older Adults Using Internet of Things and AI-Based Systems: Co-Design Study.

The number of older people with unmet health care and support needs is increasing substantially due to the challenges facing health care systems worldwide. There are potentially great benefits to using the Internet of Things coupled with artificial intelligence to support independent living and the measurement of health risks, thus improving quality of life for the older adult population. Taking a co-design approach has the potential to ensure that these technological solutions are developed to address specific user needs and requirements. The aim of this study was to investigate stakeholders' perceptions of independent living and technology solutions, identify stakeholders' suggestions on how technology could assist older adults to live independently, and explore the acceptability and usefulness of a prototype Internet of Things solution called the NEX system to support independent living for an older adult population. The development of the NEX system was carried out in 3 key phases with a strong focus on diverse stakeholder involvement. The initial predesign exploratory phase recruited 17 stakeholders, including older adults and family caregivers, using fictitious personas and scenarios to explore initial perceptions of independent living and technology solutions. The subsequent co-design and testing phase expanded this to include a comprehensive web-based survey completed by 380 stakeholders, encompassing older adults, family caregivers, health care professionals, and home care support staff. This phase also included prototype testing at home by 7 older adults to assess technology needs, requirements, and the initial acceptability of the system. Finally, in the postdesign phase, workshops were held between academic and industry partners to analyze data collected from the earlier stages and to discuss recommendations for the future development of the system. The predesign phase revealed 3 broad themes: loneliness and technology, aging and technology, and adopting and using technology. The co-design phase highlighted key areas where technology could assist older adults to live independently: home security, falls and loneliness, remote monitoring by family members, and communication with clients. Prototype testing revealed that the acceptability aspects of the prototype varied across technology types. Ambient sensors and voice-activated assistants were described as the most acceptable technology by participants. Last, the postdesign analysis process highlighted that ambient sensors have the potential for automatic detection of activities of daily living, resulting in key recommendations for future developments and deployments in this area. This study demonstrates the significance of incorporating diverse stakeholder perspectives in developing solutions that support independent living. Additionally, it emphasizes the advantages of prototype testing in home environments, offering crucial insights into the real-world experiences of users interacting with technological solutions.

Attention and nocebo hyperalgesia: Testing a novel virtual reality attention bias modification paradigm

Nocebo effects in pain (nocebo hyperalgesia) have received significant attention recently, with negative expectancies and anxiety proposed to be explanatory factors. While both expectancy and anxiety can bias attention, attention has been rarely explored as a potential mechanism involved in nocebo hyperalgesia. The present study aimed to explore whether attention bias modification (ABM) using an immersive, ecologically valid VR paradigm successfully induced attention biases (AB) and subsequently influenced nocebo hyperalgesia. One-hundred and two healthy participants were randomised in a 2 (AB training: towards vs. away from pain) x 2 (nocebo condition: nocebo vs. control) design. Pain-related AB was successfully changed by the VR paradigm as measured by reaction time and gaze, with moderate to large effects. Participants then completed either a nocebo instruction and conditioning procedure (nocebo paradigm) or a matched control procedure. The primary outcome was self-reported pain intensity. Secondary outcomes were attention bias and self-reports of expectancy, anticipatory anxiety, and state anxiety. The nocebo paradigm induced significantly greater pain expectancy, anticipatory anxiety and pain intensity during the test phase for the nocebo group compared to control. Pain expectancy also fully mediated the effect of the nocebo group on nocebo hyperalgesia and anticipatory anxiety in separate models. ABM did not, however, affect nocebo hyperalgesia or pain expectancy, casting doubt on the potential for ABM to inoculate against nocebo hyperalgesia. Unexpected effects of ABM were observed for state anxiety and anticipatory anxiety, whereby training away from pain exacerbated each, which necessitates further exploration. PerspectiveThis article tests the efficacy of a novel attention bias modification paradigm, designed in virtual reality, for inducing pain-related biases, and whether these biases exacerbate or inoculate against nocebo hyperalgesia. While pain-related biases were successfully induced, there was no relationship with the strength of induced nocebo hyperalgesia.

Recycling waste tires into sustainable biofuel for replacing petroleum fuel in diesel engines using nanoparticles and biohydrogen.

In this paper, the production of pyrolyzed oil in lab scale plant and its possible use in variable speed diesel engine with nanoparticle and secondary fuel oxy-hydrogen gas is investigated. The pyrolyzed oil is produced from waste tires in lab scale plant, and after its distillation, it is blended with neat diesel in the proportion of 20:80 (TDF). The different testing fuels are prepared by mixing nanoparticle CeO2 (NP) at the concentrations 50ppm and 100ppm with blended fuel (named as TDF-NP50 and TDF-NP100). While in the testing phase, the HHO gas is supplied through specially designed venturi at fixed flow rate (fuel designated as TDF-NP50-HHO and TDF-NP100-HHO). The focus of current investigation is to investigate the combined impact of inducting HHO with nanoparticle mixed pyrolyzed blends on engine performance and emissions characteristics. The pyrolyzed blended fuel deteriorate the BTE, BSFC, emissions of HC and NOx while improves the CO emission. The improvement in all the characteristics can be achieved by introducing CeO2 nanoparticle at 50ppm and 100ppm which improves BTE by 15-24%, BSFC by 2.2-4.2%, HC by 8-18%, CO by 7-12%, and NOx by 7-16%. The further improvement can be achieved by inducting HHO gas, which are BTE 11%, BSFC 5-8%, HC 15%, and CO 13-16%. However, the NOx emission is increased by 14-17%. Hence, the waste tire-derived pyrolyzed oil along with green hydrogen and nanoparticles CeO2 can be used as a sustainable transportation fuel in existing CI engine.

Integrating automated dispensing cabinets into the medication dispensing process: feedback from the practice in European hospitals

ObjectivesAutomated dispensing cabinets (ADCs) offer improved medication safety, greater efficiency and return on investment. However, integrating ADCs into medication dispensing processes can be challenging in complex hospital environments. This study...

Sex differences in exploratory behavior of rats successfully performing the object-in-place recognition memory test

Male and female rodents display unique search strategies when exploring new and familiar environments. Sex differences are well-documented in the literature and may be observed in tasks that rely on spontaneous exploration (e.g., recognition memory tests). Therefore, we assessed patterns of male and female rat behavior in the object-in-place (OiP) test, a common recognition memory paradigm involving object-location associations. Twelve male and 12 female adult Long Evans rats were tested four times in the 1-h OiP test and exploratory behaviors were compared during habituation, sample, and test phases. Results revealed that females moved faster and farther than males, showed increased immobility frequency and reduced immobility duration, reduced outer zone mobility duration, and increased inner zone entrances, compared to males during habituations. During sample phases, female rats moved faster than males, displayed reduced immobility frequency in the inner zone, and demonstrated consistent distance travelled across repeated sessions; conversely, male rats moved less in later sessions and exhibited increased mobility frequency in the outer zone. Analyses comparing test phase behavior revealed females continued to move faster than males; however, no other sex differences were observed. These findings are consistent with previous literature highlighting unique sex differences in explorative behaviors during recognition testing. Sex differences in locomotion and mobility state behaviors may be more indicative of individual motivation and search strategy between the sexes and less indicative of recognition memory.

Functional reorganization of brain regions supporting artificial grammar learning across the first half year of life.

Pre-babbling infants can track nonadjacent dependencies (NADs) in the auditory domain. While this forms a crucial prerequisite for language acquisition, the neurodevelopmental origins of this ability remain unknown. We applied functional near-infrared spectroscopy in neonates and 6- to 7-month-old infants to investigate the neural substrate supporting NAD learning and detection using tone sequences in an artificial grammar learning paradigm. Detection of NADs was indicated by left prefrontal activation in neonates while by left supramarginal gyrus (SMG), superior temporal gyrus (STG), and inferior frontal gyrus activation in 6- to 7-month-olds. Functional connectivity analyses further indicated that the neonate activation pattern during the test phase benefited from a brain network consisting of prefrontal regions, left SMG and STG during the rest and learning phases. These findings suggest a left-hemispheric learning-related functional brain network may emerge at birth and serve as the foundation for the later engagement of these regions for NAD detection, thus, providing a neural basis for language acquisition.

Predictive pretrained transformer (PPT) for real-time battery health diagnostics

Modeling and forecasting the evolution of battery systems involve complex interactions across physical, chemical, and electrochemical processes, influenced by diverse usage demands and dynamic operational patterns. In this study, we developed a predictive pre-trained Transformer (PPT) model equipped with 1,871,114 parameters that enhance identification of both short-term and long-term patterns in time-series data. This is achieved through the integration of convolutional layers and probabilistic sparse self-attention mechanisms, which collectively enhance prediction accuracy and efficiency in diagnosing battery health. Moreover, the customized hybrid-model fusion supports parallel computing and employs transfer learning, reducing computational costs while enhancing scalability and adaptability. Consequently, this allows for precise real-time health estimations across various battery cycles. We validated this method using a public dataset of 203 commercial lithium iron phosphate (LFP)/graphite batteries charged at rates ranging from 1C to 8C. By using only partial charge data—from an 80 % state of charge to the maximum charging voltage (3.6 V for LFP batteries, 4.2 V for ternary batteries)—and avoiding complex feature engineering, error metrics were achieved below 0.3 % for root mean square error (RMSE), weighted mean absolute percentage error (WMAPE), and mean absolute error (MAE), with an R2 of 98.9 %. The generalization capabilities were further demonstrated across 36 different testing protocols, encompassing 23,480 cycles throughout the entire life cycle, with a total inference time of 9.88 s during the testing phases. Further experiments on 30 nickel cobalt aluminum (NCA) batteries and 36 nickel cobalt manganese (NCM) batteries, across different battery types and operational scenarios, resulted in RMSE, WMAPE, and MAE all below 0.9 %, with R2 values of 94.1 % and 94.4 %, respectively. These findings highlight the potential of our customized deep transfer neural networks to enhance diagnostic accuracy, accelerate training, and improve generalization in real-time applications.

PV Module Soiling Detection Using Visible Spectrum Imaging and Machine Learning

During the last decades photovoltaic solar energy has continuously increased its share in the electricity mix and has already surpassed 5% globally. Even though photovoltaic (PV) installations are considered to require very little maintenance, their efficient exploitation relies on accounting for certain environmental factors that affect energy generation. One of these factors is the soiling of the PV surface, which could be observed in different forms, such as dust and bird droppings. In this study, visible spectrum data and machine learning algorithms were used for the identification of soiling. A methodology for preprocessing the images is proposed, which puts focus on any soiling of the PV surface. The performance of six classification machine learning algorithms is evaluated and compared—convolutional neural network (CNN), support vector machine (SVM), random forest (RF), k-nearest neighbor (kNN), naïve-Bayes, and decision tree. During the training and validation phase, RF proved to be the best-performing model with an F1 score of 0.935, closely followed by SVM, CNN, and kNN. However, during the testing phase, the trained CNN achieved the highest performance, reaching F1 = 0.913. SVM closely followed it with a score of 0.895, while the other two models returned worse results. Some results from the application of the optimal model after specific weather events are also presented in this study. They confirmed once again that the trained convolutional neural network can be successfully used to evaluate the soiling state of photovoltaic surfaces.

Development and cognitive testing of a food frequency questionnaire to assess intake of plant-based protein foods among older adults.

To develop a web-based food frequency questionnaire (FFQ) measuring intake of plant-based protein foods (PBPs) among older adults from the province of Quebec, Canada. The questionnaire was adapted from an existing self-administered FFQ and first underwent expert panel evaluation for face and content validity. Then, three phases of cognitive testing were conducted in French, using the probe and think aloud approaches. Between each phase, the questionnaire was modified based on participants' feedback to improve the clarity and comprehension of the questions. Quebec City, Quebec, Canada. Twenty adults aged 65 years and older. Purposive sampling was used to maximize variation in sociodemographic characteristics, including gender, age, education level, and PBP consumption. The expert panel found the 28-item questionnaire to be a comprehensive measure of PBP intake and suggested minor changes to improve its clarity. The cognitive interviews showed that our PBP-FFQ was generally well understood and identified issues requiring modifications to improve comprehension and accuracy. Our FFQ provides a comprehensive measure of PBP intake, is well understood by older adults in Québec, and will support rigorous assessment of PBP intake in this population, but requires further validation to confirm validity and reproducibility.

Optimizing neural network models for predicting nuclear reactor channel temperature: A study on hyperparameter tuning and performance analysis

This study emphasizes how important accurate prediction of channel temperatures in nuclear reactors is for safety and operational efficiency. While traditional methods require long and complex processes such as kernel modeling and mathematical simulations, artificial neural networks (ANN) provide more efficient predictions by accelerating this process. The superior ability of ANNs to process large data sets is intended to demonstrate that this study will provide a valuable alternative compared to conventional methods and increase the accuracy of reactor temperature predictions. In this study, the training performances of Artificial Neural Network (ANN) developed to determine the nuclear reactor channel temperature with different hyperparameter combinations were analysed. It was conducted several experimental studies to assess the influence of hyperparameters on our model for nuclear reactor parameter data prediction. The training and validation results indicates that learning rate, hidden layer sizes and number have critical effects for the more precisive prediction. It was observed that models with a learning rate of 0.05 and 0.5 achieved successful learning with less fluctuation in training and validation errors. When looking at hidden layer sizes, networks with 32 and 64 neurons performed better than networks with 16 neurons. For the test phase our model can successfully predict data with slight error margin. As a result, we demonstrated that neural networks are a powerful tool in nuclear reactor channel temperature prediction through our proposed model.

Artificial Intelligence and Colposcopy: Automatic Identification of Vaginal Squamous Cell Carcinoma Precursors

Background/Objectives: While human papillomavirus (HPV) is well known for its role in cervical cancer, it also affects vaginal cancers. Although colposcopy offers a comprehensive examination of the female genital tract, its diagnostic accuracy remains suboptimal. Integrating artificial intelligence (AI) could enhance the cost-effectiveness of colposcopy, but no AI models specifically differentiate low-grade (LSILs) and high-grade (HSILs) squamous intraepithelial lesions in the vagina. This study aims to develop and validate an AI model for the differentiation of HPV-associated dysplastic lesions in this region. Methods: A convolutional neural network (CNN) model was developed to differentiate HSILs from LSILs in vaginoscopy (during colposcopy) still images. The AI model was developed on a dataset of 57,250 frames (90% training/validation [including a 5-fold cross-validation] and 10% testing) obtained from 71 procedures. The model was evaluated based on its sensitivity, specificity, accuracy and area under the receiver operating curve (AUROC). Results: For HSIL/LSIL differentiation in the vagina, during the training/validation phase, the CNN demonstrated a mean sensitivity, specificity and accuracy of 98.7% (IC95% 96.7–100.0%), 99.1% (IC95% 98.1–100.0%), and 98.9% (IC95% 97.9–99.8%), respectively. The mean AUROC was 0.990 ± 0.004. During testing phase, the sensitivity was 99.6% and 99.7% for both specificity and accuracy. Conclusions: This is the first globally developed AI model capable of HSIL/LSIL differentiation in the vaginal region, demonstrating high and robust performance metrics. Its effective application paves the way for AI-powered colposcopic assessment across the entire female genital tract, offering a significant advancement in women’s healthcare worldwide.

An experimental investigation to predict the compressive strength of lightweight Ceramsite aggregate UHPC using boosting and bagging techniques

Ceramsite aggregate concrete is renowned for its lightweight properties but often exhibits low compressive strength (CS), limiting its structural applications. This study introduces ultra-high-performance powder into Ceramsite aggregate concrete to enhance its CS while maintaining low density. Achieving optimal CS for this composite material requires a carefully designed mix. In this research, advanced machine learning (ML) techniques, including random forest, extreme gradient boosting, categorical boosting, light gradient boosting machine (LightGBM), natural gradient boosting (NGBoost), and adaptive boosting, were employed to predict the CS of lightweight Ceramsite aggregate UHPC (LWCA-UHPC). The study utilized 189 datasets from the authors' own experimental results for training and testing the models. The results demonstrated that NGBoost and LightGBM outperformed other models, with R² values of 0.9614 and 0.9645, respectively, in the testing phase. The models also showed low errors, with NGBoost having a mean absolute error (MAE) of 2.0624 and a root mean squared error (RMSE) of 2.9140, while LightGBM recorded an MAE of 2.0161 and an RMSE of 2.7778. Moreover, feature importance and Shapley additive explanations (SHAP) analyses revealed that the most influential factors affecting the CS of LWCA-UHPC were cement content, density, sand, and Ceramsite aggregate content, consistent with experimental findings. Additionally, a graphical user interface was developed to integrate the hybrid ML model into a user-friendly desktop tool, achieving an R² value of 0.9671 with a low mean squared error of 3.5204 when tested against experimental datasets. This tool not only enables practical applications in engineering but also helps reduce the need for extensive experimental work by providing accurate predictions of CS.