Search Results (586)

Walnuts are commercially important tree nuts whose moisture content (MC) influences quality, shelf life, and post-harvest processing. This study evaluated the potential of low-cost acoustic sensing combined with machine learning for non-destructive MC prediction. Sixty in-shell walnuts were subjected to controlled drying at 40 °C for 26 h, with acoustic recordings and physical measurements collected every two hours. Acoustic signals were processed using Wavelet Soft Threshold Denoising (WSTD), Short-Time Fourier Transform (STFT), and Variational Mode Decomposition (VMD), and features were extracted from the resulting signals. Predictive models included generalized linear models (GLM), random forests (RF), gradient boosting machines (GBM), and Partial Least Squares (PLS) approaches. Following grouped walnut-level validation, the highest MC prediction performance was achieved by the model combining dimensional and acoustic descriptors (RF: R² = 0.836; GBM: R² = 0.826), while the model combining drying time and acoustic descriptors achieved moderate predictive performance (RF: R² = 0.762; GBM: R² = 0.760). Overall, the results provide proof-of-concept evidence that acoustic descriptors may complement physical measurements for non-destructive walnut moisture-content prediction. However, substantially larger independent datasets collected across multiple cultivars, production batches, acquisition conditions, and external validation studies will be required before practical industrial implementation can be considered. Full article

Dissimilar resistance spot welding of Al-Si-coated B1500HS hot-stamped steel to HC340/590DP dual-phase steel suffers from a narrow process window and HAZ temper softening. A sequential orthogonal-central composite design strategy screened factors and constructed local second-order models for nugget diameter and tensile-shear force. Because the complete tensile-shear CCD dataset is unavailable for independent verification, the tensile-shear model is used strictly as an auxiliary local calibration and is not assigned the same validation level as the nugget-diameter model. Within-batch ANOVA showed that electrode force dominated diameter variation and first-pulse current dominated force variation. A model-assisted variance-aware compromise (7.8/8.5 kA, 2.9 kN, 13/17 cycles) was point-wise validated at 6.5065 ± 0.1366 mm and 15.053 ± 0.1899 kN (n = 20, CV 2.10%/1.26%). The measured performance-optimal orthogonal condition remained Run 11; thus, the compromise is interpreted as a stability-oriented choice rather than a global optimum. A joint-specific HAZ screening envelope (width < 0.7 mm; hardness loss < 50%) is proposed as a descriptive screening criterion only; because HAZ width and microhardness were not measured for the n = 20 validation condition, the envelope was not validated on that condition and remains conditional on the single-factor HAZ data. The framework integrates process optimization with transparent statistical qualification and reports its model calibration limits. Full article

Background: Process analytical technology (PAT) enhances product quality by monitoring and controlling critical quality attributes (CQAs), offering better insights into process performance. Raman spectroscopy is a promising PAT tool to support the development of control systems for continuous and automated processes. In this study, Raman spectroscopy was used to monitor in real-time ultra- and diafiltration (UF/DF) of adeno-associated virus (AAV) and lentiviral vector (LV). Method: The incorporation of a Raman probe in a tangential flow filtration (TFF) system, when operated in open loop, enabled spectral acquisition during a five-fold concentration of clarified bulk followed by a buffer exchange to PBS with five diafiltration volumes. Results: Principal Component Analysis reduced the dimensionality of the training set, and a Partial Least Squares model was developed for each parameter in each phase. After validation, Raman monitoring platforms showed good performance in predicting CQA of both viral vectors. Batch-to-batch variability was identified as the principal source of spectral variation, and its impact on UF/DF operation was observed. Discussion: This study emphasises the potential of Raman spectroscopy as a PAT for monitoring and control strategies in the downstream processing of viral vectors. Conclusions: These monitoring platforms enhance understanding of TFF’s operation and enable timely decision making by providing real-time information on multiple parameters. Full article

Background/Objectives: Oral solid dosage forms of traditional Chinese and ethnic medicines are currently undergoing modernisation. The objective of this study is to explore the scope for formulation variation arising from batch-to-batch fluctuations in intermediates, and to identify the factors influencing key quality attributes of honey-containing tablets. In this regard, a machine-learning-based predictive model is being formulated that will integrate and analyse formulation factors and the results characterised by the SeDeM expert system. Utilising the SeDeM index as a mediating variable, the study endeavours to establish a comprehensible and predictable stepwise research pathway to provide a foundation for industrial-scale upscaling. Methods: Twelve SeDeM expert systems were utilised to characterise honey-containing granules for formulation screening, to evaluate their suitability for use in traditional Chinese medicine honey-containing tablet systems, and to identify key limiting factors and the feasibility space affecting the quality of the final product; Based on the QBD philosophy, a TriAD (Tri-criterion Adaptive Design) design scheme was proposed, integrating the horizontal balance of orthogonal designs, the spatial coverage of uniform designs, and the parameter estimation efficiency of D-optimal designs into the experimental layout of the formulation feasibility space; Through further data aggregation, a multi-layer feature set comprising four formulation factors, six SeDeM indicators, and three critical quality attributes (CQAs) was constructed. The mediating effects of the SeDeM indicators were revealed through different pathways involving 37 combinations of simple, linear, and Bootstrap models. Furthermore, 180 linear and non-linear machine learning models (comprising 12 categories of algorithms) were trained to predict formulation and CQA outcomes, ultimately completing the design space mapping and validation. Results: The results show that the SeDeM parameters effectively bridge the CQA results of different honey formulations, with these indicators acting as selective mediators between formulation factors and CQAs. Compared with a pure data model relying solely on raw formulation variables, the introduction of SeDeM knowledge, combined with high-information-content samples obtained via TriAD, improved the predictive performance and robustness of the SeDeM–ML hybrid model in terms of disintegration time and hardness; its R²_LOO increased by 0.267 and 0.510, respectively, and the overall predictive space was significantly expanded. Experimental validation was conducted using formulations within the design space predicted by the optimal model; the results showed that both the prediction bias and the relative standard deviation were less than 5 percent. Conclusions: The present study demonstrates that SeDeM can not only be used to evaluate formulations of honey-containing TCM tablets but also serves as an intermediary bridge linking formulation factors, granule-mechanism variables, and tablet quality outcomes. TriAD, in turn, further translates the QbD philosophy into an actionable formulation space design, thereby providing a development pathway for honey-containing tablets that combines interpretability, predictability, and QbD consistency, and offers new insights for the industrial application of oral TCM preparations. Full article

Alternative fine aggregates are often assessed using compressive strength at a single reference age, which may conceal an early maximum followed by later strength loss. This preliminary screening study compared mortars containing river sand (RS), standard sand (StS), desert sand (DS), soil sand (SS), coal gangue sand (CGS), and metamorphic rock sand (MRS) under one nominal mixture design. For each aggregate, one mortar batch was prepared and nine 70.7 mm cubes were cast, with three specimens tested at 3, 7, and 28 d. RS, StS, DS, and SS continued to gain strength. Within the single CGS batch, strength decreased from 15.56 ± 0.31 MPa at 7 d to 11.68 ± 0.15 MPa at 28 d; within the single MRS batch, it decreased from 21.63 ± 0.48 MPa to 15.75 ± 0.20 MPa. The corresponding losses were 24.95% and 27.18%, and exploratory within-batch Tukey tests yielded p < 0.001. These statistics describe specimen-level variation within the tested batches and do not establish batch-to-batch reproducibility. Representative 28 d SEM fields, raw-aggregate EDS, and qualitative XRD provide contextual observations but lack the temporal and spatial resolution needed to reconstruct a defect-formation process between 7 and 28 d. Because aggregate moisture state, absorption, flow, air content, and compaction were not independently controlled, the results identify a screening signal that requires independent-batch validation rather than a general material mechanism. Full article

Stopping fermentation is a critical, yet insufficiently standardised, step in the production of aromatic grape distillates, as it influences residual sugar, yeast metabolism and the availability of volatile precursors. This study evaluated the must density at the time of fermentation arrest as a process control parameter for Torontel grape distillates. Five treatments were established by halting fermentation at densities ranging from 1.050 to 1.010 g/mL, followed by batch distillation. The distillates were evaluated using physicochemical analyses, including alcohol strength and the concentration of key volatile congeners such as methanol, acetaldehyde and higher alcohols. Furthermore, these analyses were complemented by gas chromatography with flame ionisation detection (GC-FID) and sensory evaluation by experts. Increased must density was associated with higher concentrations of methanol, acetaldehyde and higher alcohols, whereas ethanol content decreased. Although clear differences in chemical composition were observed among the evaluated process conditions, overall sensory acceptability exhibited only limited variation across the fermentation arrest densities. However, distillates obtained from intermediate density levels showed better aromatic balance and sensory harmony. Multivariate analysis suggested that fermentation arrest density was associated with the main patterns of chemical variability observed among the evaluated distillates, whereas its association with overall sensory perception appeared to be more limited. These findings provide a useful framework for optimising fermentation arrest strategies in the production of aromatic partially fermented grape distillates. Full article

Temperature–time profiles obtained through thermistor-based monitoring provide a rich but noise-sensitive source of information for estimating kinetic and thermal parameters in exothermic batch reactions. Conventional workflows typically combine deterministic smoothing with numerical differentiation, an approach that can amplify measurement noise and fail to propagate preprocessing uncertainty into the resulting reaction-rate and parameter estimates. To address these limitations, this study presents an uncertainty-aware Bayesian machine-learning framework that integrates scalable random-Fourier-feature Gaussian-process (RFF–GP) smoothing, analytical differentiation, temperature-derived apparent conversion, Bayesian parameter inference, posterior validation, predictive calibration, model comparison, ablation, sensitivity analysis, probabilistic benchmarking, simulation of thermal nonideality, and endpoint diagnostics. The framework was applied to $379,631$ cleaned thermistor observations. The production RFF–GP achieved a validation root-mean-square error of $0.04805\mathrm{K}$, yielding a stable latent temperature trajectory and an uncertainty-aware estimate of $dT/dt$. On a smaller matched subset, exact Gaussian-process regression achieved the highest predictive accuracy and the best probabilistic scores, whereas the RFF–GP reduced central-processing-unit runtime by approximately $4.1$-fold and remained applicable to the larger production fit. A Monte Carlo dropout neural comparator produced larger prediction errors and substantially wider predictive intervals. Six apparent thermokinetic structures were evaluated using mean-field variational inference, after which the nth-order and autocatalytic structures were validated using the No-U-Turn Sampler (NUTS). Under mean-field variational inference, the apparent autocatalytic structure achieved the lowest point estimate of the widely applicable information criterion (WAIC), the lowest derivative-domain error, and the lowest full-profile temperature-reconstruction root-mean-square error of $0.2920\mathrm{K}$. Its posterior obtained using NUTS yielded $E_a=40.98\mathrm{kJ}{\mathrm{mol}}^{-1}$, $k_{\mathrm{ref}}=0.005815{\min }^{-1}$, $\Delta T_{\mathrm{ad}}=56.11\mathrm{K}$, $m=0.1694$, and $n=1.0784$. The sampling diagnostics indicated satisfactory convergence, large effective sample sizes, and no divergent transitions. Although the MFVI posterior means and NUTS posterior medians were similar, variational inference produced narrower uncertainty intervals for several correlated parameters. Moving-block bootstrap intervals did not establish a decisive separation in WAIC among the leading structures. Expanded sensitivity, ablation, imperfect-insulation simulation, and endpoint-holdout analyses further showed that the apparent parameter estimates were sensitive to optimization, thermal nonideality, sensor response, and Gaussian-process boundary behavior. The autocatalytic formulation should therefore be interpreted as the best-performing apparent structure among the candidates tested rather than as evidence of a unique chemical mechanism. Overall, the framework extracted physically plausible apparent thermokinetic information from noisy temperature-only measurements while explicitly quantifying uncertainty arising from prediction, parameter estimation, model form, computation, thermal nonideality, and boundary behavior. Full article

Thermoplastic polyolefin (TPO) waterproofing membranes are widely used in building roofs and underground waterproofing systems, and their durability under complex service environments is critical to long-term waterproofing reliability. Self-adhesive TPO membranes from one production batch were evaluated in the unaged condition (C0) and after 5, 7, 14, and 28 complete coupled aging cycles (C5, C7, C14, and C28, respectively). Each complete cycle lasted 48 h; thus, C5, C7, C14, and C28 corresponded to total elapsed times of 10, 14, 28, and 56 d. Mass variation, Fourier-transform infrared spectroscopy (FTIR), peel, tensile, and scanning electron microscopy (SEM) measurements were integrated to evaluate physical, near-surface chemical, interfacial, and mechanical changes. The reported mass-loss ratios varied only slightly from 0.25% to 0.35%, indicating limited sensitivity of mass variation under the adopted conditions. FTIR band-depth indices near 1021, 876, 1462, and 719 cm⁻¹ changed non-monotonically with cycle number, whereas SEM images showed a visually more heterogeneous exposed TPO surface after aging. At C28, the mean peak peel resistance decreased from 94.2 ± 3.1 to 28.5 ± 0.6 N/50 mm (69.8%), while the mean nominal tensile stress at 250 mm crosshead extension decreased from 2.59 ± 0.14 to 2.06 ± 0.05 MPa (20.7%). No tensile specimen ruptured within the investigated extension range. These descriptive results indicate that the self-adhesive joint response was more sensitive to the adopted coupled aging protocol than the fixed-extension tensile response of the membrane. Full article

Global population growth has emphasised the need to have sustainable and alternative sources of nutrients. In this context, microalgae have emerged as a potential solution due to their rich biochemical composition, including high-quality proteins, carbohydrates, lipids, and pigments. This study investigates the variation in microalgal growth and biochemical composition over a light:dark cycle, with a focus on the night period. Batch experiments were performed with eight Chlorella vulgaris cultures over a 7-day period. On the seventh day, biomass samples were collected at four time points in four-hour intervals and stored for subsequent biochemical analyses. During the eight-hour dark period, biomass, carbohydrate, and total chlorophyll concentrations decreased by 9%, 12.5%, and 14.4%, respectively. After four hours of light exposure, these parameters increased significantly by 6%, 15.4%, and 12.7%, respectively. Total protein, carotenoids, and fatty acid contents remained relatively stable throughout the evaluated cycle, although variations were observed in the carotenoid profile. During the dark phase, zeaxanthin decreased by 38.0%, whereas violaxanthin increased by 27.2%, suggesting complementary pigment interconversion consistent with xanthophyll cycle activity. Overall, these results highlight the importance of optimising harvesting time to enhance the production of target compounds in a sustainable production of microalgal biomass. Full article

Traditional Bayesian SAR detection methods suffer poor adaptability to speckle noise, fail to handle severe class imbalance within large-scale multi-target datasets, and incur prohibitive training overheads. To address these drawbacks, this paper develops a lightweight Bayesian detection and recognition framework built upon heavy-tailed Laplacian priors and variational inference. We adopt ResNet-50 as the feature extraction backbone and design a four-stage pipeline: First, a noise-aware Laplacian heavy-tailed prior is proposed to strengthen resistance against speckle outliers. Second, a multi-class variational inference module is constructed to eliminate detection bias induced by uneven sample distribution across target categories. Third, a lightweight uncertainty feedback strategy is introduced to cut computational costs for large-batch training. Evaluated on the MSAR-1.0 dataset, our approach achieves an mAP@0.5 of 94.98% and a macro balanced accuracy ($BA$) of 93.34%. Compared with existing Bayesian detectors, the mAP metric rises by 5.44–6.53%. The model only consumes 4.33 ms per inference frame and completes full training within 1.53 h on a single GPU. Ablation tests validate the independent and combined efficacy of all three core modules. This integrated architecture balances detection precision, classification reliability, and training efficiency, offering a promising prototype for multi-class SAR target interpretation under the evaluated benchmark constraints. Full article

Goose bone paste is an underutilised poultry-processing by-product and a potential source of taste-active peptides. A nominal 1–3 kDa peptide fraction (F2), operationally designated an umami peptide fraction by analogy with comparable bone-hydrolysate fractions reported in the literature, was isolated from a neutral-protease hydrolysate by sequential ultrafiltration and encapsulated in sodium alginate (SA) microcapsules using extrusion–dripping ionic gelation. Single-factor screening identified the following formulation conditions: 2.0% (w/v) SA, 2.5% (w/v) CaCl₂, 0.3% (w/v) SE-15, a core-to-wall mass ratio of 0.3, and a preparation temperature of 50 °C. A verification batch prepared under these conditions gave an encapsulation efficiency of 75.44%, with the ±1.07% denoting the SD of three technical determinations from that batch. The dried microcapsules had a moisture content of 2.98 ± 0.21% and passable flowability. The mean particle diameter was 856 ± 52 μm, with a within-batch coefficient of variation of 5.56 ± 0.28%; a complete particle-size distribution was not recorded. In pepsin-free simplified simulated gastric fluid, the apparent release from the microcapsules rose from about 6% at 1 h to about 13% at 5 h, whereas the apparent detection ratio of free F2 rose from about 56% to about 99%. The calculated concentrations fell at or below the validated limit of quantification, the microcapsule-group absorbances lay near the photometric floor of the instrument, and only three sampling times were used. These percentages and the kinetic fits are therefore qualitative to semi-quantitative. The data support only the relative statement that alginate encapsulation lowered the apparent release of F2 under the tested acidic conditions. They do not establish an exact release rate, an error estimate for the microcapsule group, or a specific release mechanism. Full article

Phenolic compounds are important plant secondary metabolites with broad biological activity and potential applications in pharmaceutical, food, cosmetic, nutraceutical, and veterinary sectors. Conventional production from field-grown plants is limited by environmental variability, seasonality, and difficulties in standardizing metabolite composition. Plant in vitro cultures provide controlled systems for modulating secondary metabolism and producing phenolic compounds under defined conditions. This review summarizes current advances in plant in vitro platforms for phenolic production and discusses regulation through elicitation, metabolic modulation, molecular approaches, and bioreactor cultivation. The distinctive focus of this review is the critical evaluation of how culture type, production stability, metabolite composition, and structural variation affect biological performance and functional equivalence. Current evidence indicates that increased metabolite accumulation alone does not ensure preserved biological properties or translational applicability. Functional equivalence is therefore considered as a framework integrating chemical profiling, batch-to-batch reproducibility, biological validation, bioavailability, and application-oriented evaluation of in vitro-derived phenolics. Future progress will depend not only on increasing yield but also on achieving stable production, predictable composition, reproducible biological performance, and translational reliability. Full article

The non-Gaussian height distribution and power spectral density (PSD) characteristics of rough surfaces have significant effects on the real contact area, local pressure distribution, oil-film formation, and friction and wear behavior of lubricated contact interfaces in mechanical components. Conventional methods for generating non-Gaussian rough surfaces commonly rely on iterative correction under explicit statistical constraints, which limits their computational efficiency in large-scale sample generation. To address this issue, this study proposes a PSD-amplitude-constrained phase conditional variational autoencoder (phase C-VAE) for generating non-Gaussian rough surfaces. Unlike conventional constructive methods that repeatedly correct surface samples under explicit statistical constraints, the proposed method learns the conditional distribution of the Fourier phase, while the spectral amplitude used for reconstruction is directly determined from the prescribed PSD. By taking the target skewness, kurtosis, and PSD as conditional inputs, the proposed method achieves joint control of higher-order statistical characteristics and spectral characteristics within a unified generative framework. Under target conditions derived from measured surfaces, the generated non-Gaussian rough surface samples achieved mean absolute relative errors of $0.056\%$ and $0.044\%$ for skewness and kurtosis, respectively, with a generation time of $24.62\mathrm{s}$. These results indicate that the proposed method can effectively match the target skewness and kurtosis while maintaining good consistency between the generated surfaces and the target PSD. The proposed method alleviates the efficiency limitation of conventional constructive methods in the large-scale generation of non-Gaussian rough surface samples and provides an effective machine-learning-based generative approach for rapid batch modeling of rough surfaces in lubrication, friction, and contact analyses. Full article

Slope stability assessment remains a fundamental challenge in geotechnical engineering because of the complex nonlinear interactions among soil properties, slope geometry, and hydraulic conditions, particularly variations in pore-water pressure. This study investigates the reliability of Artificial Neural Network–Multi-Layer Perceptron (ANN–MLP) models for predicting the Factor of Safety (FoS) of homogeneous soil slopes through a systematic comparison of three gradient-based optimization algorithms: Adam, Mini-Batch Gradient Descent (MBGD), and Nesterov Accelerated Gradient (NAG). A database comprising 2014 slope cases, compiled from published studies and numerically generated using Limit Equilibrium Method (LEM) and Finite Element Method (FEM) analyses, was used for model development and k-fold cross-validation. Beyond statistical evaluation, the developed models were validated using two classical dry-slope benchmark frameworks based on the Taylor stability charts and Bishop–Morgenstern stability coefficients, followed by two documented engineering case studies from Hulu Kelang and Pahang, Malaysia, to assess predictive performance under both dry and variable hydraulic conditions. Adam achieved the highest cross-validated predictive accuracy (R² = 0.988; RMSE = 0.212), whereas MBGD demonstrated the closest overall agreement with the reference LEM solutions across the validation cases and under increasing pore-water pressure ratios. NAG generally produced more conservative predictions while exhibiting greater sensitivity to hyperparameter selection. All models successfully reproduced the expected nonlinear reduction in FoS with increasing pore-water pressure, consistent with established geotechnical behaviour. The results demonstrate that optimizer selection significantly influences ANN–MLP prediction behaviour and that properly validated gradient-based ANN models can serve as efficient decision-support tools for rapid slope stability assessment under hydraulic variability. Full article

Reliable counting of horticultural seeds is important for seed grading, batch consistency assessment, packaging, sowing-rate control, and post-harvest quality evaluation. Manual counting is labor-intensive, whereas conventional image-processing methods can be sensitive to seed adhesion, visual similarity, background variation, and illumination changes. This study developed a lightweight ViT-UNet density regression framework for point-supervised multiclass counting of cucumber, tomato, and pepper seeds. Vision Transformer (ViT) features were integrated with a U-Net-style decoder to predict category-specific density maps from RGB images. A dataset containing 231 images and 7789 annotated seed instances was constructed, and a held-out test set of 36 images was used for final evaluation. For total-count estimation, the proposed model achieved a mean absolute error (MAE) of 4.0678, root mean squared error (RMSE) of 5.1896, mean absolute percentage error (MAPE) of 14.7104%, and R² of 0.8272. Bootstrap resampling yielded a 95% confidence interval of 3.0517–5.1832 for total-count MAE, and paired Wilcoxon signed-rank tests with Holm–Bonferroni correction provided statistical support for lower image-level total-count absolute errors of the proposed model relative to the evaluated deep-learning baselines and the ablation model. The network forward pass reached 94.10 FPS, whereas the full image-processing pipeline reached 2.55 FPS, indicating real-time inference potential at the model level but not yet full-pipeline industrial deployment. These results suggest that ViT-UNet density regression is a useful basis for non-destructive multiclass seed counting, while category-level discrimination, external validation, and pipeline optimization remain priorities for future work. Full article