As microprocessor architectures continue to increase computing performance under low-energy consumption, the combination of signal integrity, electromagnetic interference, and power delivery (PD) is becoming crucial in the computer industry. In that context, PD engineers make use of complex and computationally expensive models that impose time-consuming industrial practices to reach an adequate PD design. In this article, we propose a general surrogate-based methodology for fast and reliable analysis and design optimization of PD networks (PDN). We first formulate a generic surrogate model methodology exploiting passive lumped models optimized by parameter extraction to fit PDN impedance profiles. This PDN modeling formulation is illustrated with industrial laboratory measurements of a fourth generation server CPU motherboard. We next propose a black box PDN surrogate modeling methodology for efficient and reliable PD design optimization. To build our black box PDN surrogate, we compare four metamodeling techniques: support vector machines, polynomial surrogate modeling, generalized regression neural networks, and Kriging. The resultant best metamodel is then used to enable fast and accurate optimization of the PDN performance. Two examples validate our surrogate-based optimization approach: a voltage regulator (VR) with dual power rail remote sensing intended for communications and storage applications, by finding optimal sensing resistors and loading conditions; and a multiphase VR from a fifth-generation Intel server motherboard, by finding optimal compensation settings to reduce the number of bulk capacitors without losing CPU performance.

As microprocessor design scales to the 10-nm technology and beyond, traditional pre- and post-silicon validation techniques are unsuitable to get a full system functional coverage. Physical complexity and extreme technology process variations severely limits the effectiveness and reliability of presilicon validation techniques. This scenario imposes the need of sophisticated post-silicon validation approaches to consider complex electromagnetic phenomena and large manufacturing fluctuations observed in actual physical platforms. One of the major challenges in electrical validation of high-speed input/output (HSIO) links in modern computer platforms lies in the physical layer (PHY) tuning process, where equalization techniques are used to cancel undesired effects induced by the channels. Current industrial practices for PHY tuning in HSIO links are very time consuming since they require massive laboratory measurements. An alternative is to use machine learning techniques to model the PHY, and then perform equalization using the resultant surrogate model. In this paper, a metamodeling approach based on neural networks is proposed to efficiently simulate the effects of a receiver equalizer PHY tuning settings. We use several design of experiments techniques to find a neural model capable of approximating the real system behavior without requiring a large amount of actual measurements. We evaluate the models performance by comparing with measured responses on a real server HSIO link.

This paper proposes a hybrid predictive maintenance framework that combines the discriminative power of XGBoost with the interpretability of a Bayesian Network automatically learned from sensor data. Targeted at textile manufacturing equipment operating under Industry 4.0 conditions, the system addresses the trade-off between early fault detection and decision transparency. Sensor data, including vibration, temperature, and electric current, were collected from a multi-needle quilting machine using a custom IoT-based platform. A degradation-aware labeling scheme was implemented using historical maintenance logs to assign semantic labels to sensor readings. A Bayesian Network structure was learned from this data via a Hill Climbing algorithm optimized with the Bayesian Information Criterion, capturing interpretable causal dependencies. In parallel, an XGBoost model was trained to improve classification accuracy for incipient faults. Experimental results demonstrate that XGBoost achieved an F1-score of 0.967 on the high-degradation class, outperforming the Bayesian model in raw accuracy. However, the Bayesian Network provided transparent probabilistic reasoning and root cause explanation capabilities—essential for operator trust and human-in-the-loop diagnostics. The integration of both models yields a robust and interpretable solution for predictive maintenance, enabling early alerts, visual diagnostics, and scalable deployment. The proposed architecture is validated in a real production line and demonstrates the practical value of hybrid AI systems in bridging performance and interpretability for predictive maintenance in Industry 4.0 environments.

A practical formulation for EM‐based design optimization of high‐frequency circuits using simple polynomial surrogate functionals is proposed in this paper. Our approach starts from a careful selection of design variables and is based on a closed‐form formulation that yields global optimal values for the surrogate model weighting factors, avoiding a large set of expensive EM model data, and resulting in accurate low‐order low‐dimension polynomials interpolants that are used as vehicles for efficient design optimization. Our formulation is especially suitable for EM‐based design problems with no equivalent circuital models available. The proposed technique is illustrated by the EM‐based design optimization of a Ka‐band substrate integrated waveguide (SIW) interconnect with conductor‐backed coplanar waveguide (CBCPW) transitions, a low crosstalk PCB microstrip interconnect structure with guard traces, and a 10–40‐GHz SIW interconnect with microstrip transitions on a standard FR4‐based substrate. Three commercially available full‐wave EM solvers are used in our examples: CST, Sonnet, and COMSOL. Copyright © 2015 John Wiley & Sons, Ltd.

In this work we use a Neural Network model to forecast Mexican inflation. Related works forecast inflation in countries where this economic variable has a stable behavior while Mexican inflation has been characterized to be very volatile during certain periods. There were implemented different Neural Network models by varying the number of hidden layers (1 and 2) and the number of neurons in the hidden layer (from 1 to 100). The forecasting model results were divided into 3 categories: a volatile inflation phase, where the mean difference was 0.64% between real inflation and forecasted inflation; a transition phase, where the mean difference was 5.44%; and a stability phase, where the mean difference was 0.28%. By doing a comparison between model forecasting results and Bank of Mexico's predictions, Neural Networks model results are clearly more accurate to the real inflation behavior, a critical point during inflationary crisis periods.