A methodology is implemented to deform the surface of a magnetorheological elastomer (MRE) exposed to an external magnetic field by means of data matrix manipulation of the surface. The elastomer surface is created randomly using the Garcia and Stoll method to realize a nonuniform morphology similar to that found in real MREs. Deformations are induced by means of the translations of the magnetic particles inside the elastomer, under the influence of a uniform magnetic field, generating changes in the surface roughness. Our model computes these deformations using a three-dimensional Gaussian function bounded at 2 standard deviations from its mean value, taking as the standard deviation value the radius of the particle that causes the deformation. To find the regions deformed by the particles, we created a methodology based on the consultation, creation and modification of a system of matrices that control each point of the random surface created. This methodology allows us to work with external files of initial and subsequent positions of each particle inside the elastomer, and allows us to manipulate and analyze the results in a smoother and faster way. Results were found to be satisfactory and consistent when calculating the percentage of surface deformation of real systems.
The presence of microplastics (MPs) in human tissues raises questions regarding their maternal-fetal impact. Although several countries have confirmed the presence of MPs in the placenta, no previous studies have been conducted in Latin America. This study is the first to identify MPs in the human placenta in Colombia and evaluate the influence of clinical, anatomical, and methodological variables on their detection. Twelve patients were analyzed, and 216 placenta samples were collected, according to the sampling considerations. After digestion with KOH or H2O2, FTIR spectra and morphological characterizations were obtained using SEM–EDS. Associations were evaluated using χ2 and linear models for the carbonyl index (CI). PELD ([[EQUATION]]), PA ([[EQUATION]]), and aminoplastics ([[EQUATION]]) were mainly identified. Anatomical location was significantly associated with polymer type ([[EQUATION]], [[EQUATION]]) and digestion reagent ([[EQUATION]], [[EQUATION]]). CI was higher on the fetal side than on the cord side ([[EQUATION]]; [[EQUATION]]), and KOH significantly reduced CI ([[EQUATION]]; [[EQUATION]]). SEM revealed fragments (112.8 ± 57.6 μm) and fibers (228.8 ± 29.5 μm) in the treated dentin. The results confirmed the presence of MPs in all placental compartments, with higher detection on the fetal side. Reagents, protocols, and locations influence the characterization, highlighting the need for methodological standardization and broader regional studies. The study received institutional ethical approval, and informed consent was obtained from all participants.
Short engineering time series combine limited target-domain data, temporal instability, and deploymentconstraints that hinder reliable training and calibration of high-capacity models. We propose the Small-DataSurrogate Shock-Aware (S3) family, where ”Shock-Aware” denotes residual adaptation informed by localinstability and volatility rather than explicit event-level shock detection. This gradient-free residual-adaptationframework freezes a strictly causal forecasting prior and learns only a low-capacity residual correction. S3-Forecaster combines fixed Echo State dynamics, autoregressive residual lags, and volatility-sensitive correctioncontrol, whereas S3-FastSketch replaces the recurrent state with an interpretable deterministic multiscaleresidual representation regulated by projected shrinkage. Both use sequential Adaptive Conformal Inference(ACI) for interval calibration. Evaluation uses strict chronological splitting on CIF, M3, M4, and Tourismbenchmarks and a real industrial construction-material demand application comprising 15 recurrent-demandproducts with 62 monthly observations each. S3-FastSketch achieved the lowest median sMAPE on CIF(5.06%) and Tourism (16.38%), while S3-Forecaster remained competitive under reduced-history conditionsand obtained a median MSIS of 2.729 in the industrial cohort. Classical statistical and kernel methodsremained stronger in several M3 and M4 conditions, showing that residual adaptation is not universallybeneficial. Ablation identified projected shrinkage as FastSketch’s only consistently supported cross-datasetcomponent, while multi-prior experiments showed that adaptation depends on the frozen prior. These resultssupport S3 as a lightweight complement to established forecasters when short histories require controlledresidual adaptation, interpretable low-capacity representations, and calibrated uncertainty.