Paolo Carbone’s research while affiliated with University of Perugia and other places

What is this page?

This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (91)

Uncertainty Quantification in AI-Based Measurement Systems
  • Article

May 2025

·

3 Reads

IEEE Instrumentation & Measurement Magazine

·

Paolo Carbone

·

Alessio De Angelis

·

[...]

·

The applications of artificial intelligence (AI) are transforming our world and impacting all aspects of society, economy, life, work, and technology, as well as the field of Instrumentation and Measurement (IM). When accurate mathematical modelling of the measurement system is impossible or when the accuracy of the final measurement is more important than understanding exactly how the system works, AI can offer an attractive and practical solution [1].

Optimization workflow for frequency-based regularization: The procedure accepts complex impedance measurements Zm(ω) at discrete frequencies ωk and strategically partitions the frequency spectrum into three subsets. Each subset defines a domain K for evaluating the cost function given by Eq. 13. The algorithm then determines the optimal regularization parameter λsel by minimizing the cost function J(λ).
Synthesis of the measured values Zm(ω) by simulation, based on the theoretical knowledge of Z(ω) and assuming that noise corrupts both the time-domain current and voltage sequences.
Single record of a Monte Carlo simulation assuming an impedance based on the series of three ZARC elements. Shown are the impedance magnitude and phase, and the associated DRT. Shown are the real part, imaginary part, and phase of the error spectrum in estimating the measured impedance by means of the discrete DRT transform. Also shown are the number of peaks and the cost function as a function of the regularization parameter and the comparison of measured and estimated impedance in the Cole-Cole plot. Only the frequency content of the component in low-frequency intervals is used to find the minimum of the cost function.
Boxplot of the results obtained by running 150 times the various optimal λ algorithms using the 3 ZARC impedance. From left to right: λREF obtained by minimizing the rmse between the estimated DRT and the one theoretically known. λmGCV based on the mGCV algorithm. λLC based on the calculation of λ that sets the corner in the L-curve associated with regularized least squares. The rightmost three boxplots refer to the usage of the frequency-selective method that provides the three values associated with the cost function defined in the indicated frequency intervals.
Graphs of the functions used to determine the optimal value of λ when using the (a) mGCV approach, (b) the L-curve technique, (c) the frequency-selective method with the cost function evaluated at low-frequencies.

+7

Optimal Regularization for the Distribution of Relaxation Times via Frequency-Band Selection
  • Article
  • Publisher preview available

February 2025

·

51 Reads

Paolo Carbone

·

Alessio De Angelis

·

Antonio Bertei

·

[...]

·

Electrochemical impedance spectroscopy (EIS) is a powerful analytical technique for characterizing electrochemical energy storage and conversion systems. Among the methods for analyzing EIS data, the distribution of relaxation times (DRT) has emerged as a valuable tool for quantitative interpretation. DRT estimation typically relies on regularized least squares methods, where the selection of an appropriate regularization parameter remains a key challenge. This selection critically affects the balance between the smoothness of the estimated curves and the resolution of physically meaningful features in the DRT. We introduce a novel frequency-domain approach, specifically focused on ridge regression regularization, that provides objective criteria for regularization parameter selection. Our method employs specialized cost functions to analyze user-defined frequency components in the measurement data, complementing existing cross-validation approaches. Extensive Monte Carlo simulations demonstrate the method’s effectiveness in comparison to established techniques. The approach not only yields robust results but also provides intuitive visualizations to support parameter selection. Successful validation across multiple experimental datasets confirms its practical utility.

View access options
Sinewave Fitting Based on Quantization Codes and Threshold Levels

January 2025

·

31 Reads

IEEE Transactions on Instrumentation and Measurement

·

Alessio De Angelis

·

·

Paolo Carbone

This paper illustrates a method for measuring the parameters of a sinewave by using information about both the codes and threshold levels in a quantizer. It is shown that by adding knowledge about the threshold levels, the input signal parameters can still be recovered when very coarse quantization is applied in a case when the conventional three-parameter least squares sinefit method fails to produce precise results. After the description of the procedure, its sensitivity to the accuracy of the frequency estimate is analyzed. Furthermore, the robustness of the method against uncertainties in the knowledge of the threshold levels is assessed. The estimation procedure is validated by extensive simulations and by experimental results obtained by measurement on a 12-bit data acquisition board. Results show that the proposed solution can significantly outperform the three-parameter least squares sinefit if the number of quantizer bits is small.

Download
Damped Sine Wave Parameter Extraction: Application to QCM-D Signals for Accurate Measurements

January 2025

·

5 Reads

IEEE Transactions on Instrumentation and Measurement

·

·

·

[...]

·

Quartz crystal microbalance with dissipation monitoring (QCM-D) is a sensing technique that finds diverse applications in chemical sensing, biosensing, and industrial fields. By monitoring the vibrational characteristics of a thin piezoelectric SiO2 resonator, this technique provides valuable information about the viscoelastic behavior of media in contact with its surface. This paper evaluates the performance of algorithms applied to 10 MHz AT-cut QCM-D signals for estimating the frequency and time constant related to the fundamental resonant mode, extending the study to assess the effectiveness of disturbance rejection, particularly against quartz spurious oscillating modes. To this end, both damped single-tone and three-tone signals are analyzed to cover all potential outputs of a QCM-D in its fundamental bulk shear mode. Various estimation algorithms are considered: one operating in the frequency domain and others in the time domain, based on model fitting techniques with novel preprocessing steps for optimized initialization. The analysis utilizes signals from a QCM-D measurement set-up and signals simulated according to the characteristics of the AT-CUT crystals used. Among model fitting-based estimators, two techniques were developed for single-tone and three-tone signals, splitting the model fitting problem into non-linear and linear subproblems, demonstrating the best accuracy and processing time. The developed techniques meet strict measurement objectives, achieving frequency measurement resolution around 1 ppm (10 Hz) and decay time accuracy around 1 μs, even with AWGN standard deviations of about 1% of the fundamental tone amplitude in water. These techniques are applicable in various contexts where recovering parameters from damped sine waves in the presence of white noise is required.

Citations (37)

... Noise generation.-To obtain valid Monte Carlo results, the behavior of a galvanostat is simulated as in De Angelis et al. 75 : a sinusoidal current is passed through the impedances in Eqs. 18 or 21 to obtain a voltage at its terminals. ...

Reference:

Optimal Regularization for the Distribution of Relaxation Times via Frequency-Band Selection
Characterization of Uncertainty in EIS and DRT Analysis of Lithium Batteries
  • Citing Conference Paper

  • June 2024

Alessio De Angelis

·

Antonio Bertei

·

Paolo Carbone

... In [15], several techniques are analyzed when using one-bit quantization for the estimation of periodic signals in the area of instrumentation and testing. In [16], the first steps in the extension of the approach to a few bits have been taken. The present paper contains a significant extension of this proceedings publication, aiming to restore the input signal, given the extra information on the quantization levels. ...

Reference:

Sinewave Fitting Based on Quantization Codes and Threshold Levels
Calibrated Sinefit Based on Quantized Data
Paolo Carbone

·

·

Alessio De Angelis

·

... 7,22,23 Although some embedded applications of EIS measurement via harmonic perturbation of the LIB in EVs have been advanced, [24][25][26][27] the acquisition of low-frequency impedance data requires excessive time to enable a full EIS reconstruction. Thus, advanced approaches, including multi-sine [28][29][30] and alternative excitation signals, [31][32][33][34] have been proposed to shorten the acquisition time of low-frequency impedance data, additionally including compensation strategies to enable data acquisition even when the battery is not strictly at open circuit. 29,35,36 In most of these cases, the impedance spectrum is reconstructed via time-domain measurements whose basic input signal architecture is, broadly speaking, a step current perturbation or, equivalently, a current interruption. ...

Reference:

Requirements for Applying the Current Interruption Technique for Reconstructing the Impedance of Li-Ion Batteries
Fast Battery EIS Measurement Using Flexible Local Rational Method

IEEE Transactions on Instrumentation and Measurement

Alessio De Angelis

·

·

·

Paolo Carbone

... 7,22,23 Although some embedded applications of EIS measurement via harmonic perturbation of the LIB in EVs have been advanced, [24][25][26][27] the acquisition of low-frequency impedance data requires excessive time to enable a full EIS reconstruction. Thus, advanced approaches, including multi-sine [28][29][30] and alternative excitation signals, [31][32][33][34] have been proposed to shorten the acquisition time of low-frequency impedance data, additionally including compensation strategies to enable data acquisition even when the battery is not strictly at open circuit. 29,35,36 In most of these cases, the impedance spectrum is reconstructed via time-domain measurements whose basic input signal architecture is, broadly speaking, a step current perturbation or, equivalently, a current interruption. ...

Reference:

Requirements for Applying the Current Interruption Technique for Reconstructing the Impedance of Li-Ion Batteries
Battery Impedance Spectroscopy Embedded Measurement System
Gabriele Cicioni

·

Alessio De Angelis

·

·

[...]

·

Paolo Carbone

... On average, the waist-mounted motor consumed battery power at a rate of 0.034 V/min, whereas the shank-mounted motor had a battery consumption rate of 0.0275 V/min, indicating that the waist-mounted motor configuration's battery consumption was 19.12% higher than that of the shank-mounted motor configuration. With a 3.3 V lower cutoff cell voltage and 4.2 V upper cutoff cell voltage [43], and using a 6-cell battery (1800 mAh), a unilateral setup with the waist-mounted motor would have a nominal battery life of 159 minutes -23.6% shorter than the shank-mounted motor's 196 minutes. The average motor current magnitude was 1.48 A for the waist-mounted motor and 1.24 ...

Reference:

Effects of Ankle Exoskeleton Motor Location on Gait Biomechanics and User Perceptions: The Bowden Cable Dilemma
LiPo batteries dataset: Capacity, electrochemical impedance spectra, and fit of equivalent circuit model at various states-of-charge and states-of-health
  • Citing Article

  • September 2023

Data in Brief

·

·

Corrado Giammanco

·

[...]

·

Paolo Carbone

... To apply the ultra-wideband frequency positioning technology for 3D positioning in multi-story buildings, infrastructure improvements are required. For example, Brunacci et al [11] developed a robot operating system that can acquire Bluetooth low energy data from ultra wide band (UWB) devices, achieving high precision of 3 cm in static conditions and 8 cm in dynamic conditions. However, this system requires the installation of four UWB sensors around the robot, which increases the localization cost. ...

Reference:

A multi-sensor fusion approach for centimeter-level indoor 3D localization of wheeled robots
Development and Analysis of a UWB Relative Localization System
  • Citing Article

  • January 2023

IEEE Transactions on Instrumentation and Measurement

·

Alessio De Angelis

·

Gabriele Costante

·

Paolo Carbone

... With ML classification, UQ methods produce a distribution of confidence score values associated with each possible output class. In [24] we characterized an ML based battery state of charge classification system using both the GUM MCS method and MCD method. The two methods give comparable results on the test cases considered (Fig. 3): the distribution resulting from application of MCS and MCD are similar in shape and characterized by a comparable standard deviation. ...

Reference:

Uncertainty Quantification in AI-Based Measurement Systems
Uncertainty characterization of a CNN method for Lithium-Ion Batteries state of charge estimation using EIS data
  • Citing Article

  • July 2023

Measurement

·

Alessio De Angelis

·

·

Paolo Carbone

... Predicting State of Charge (SOC) with an exact mathematical model is quite challenging due to the influence of several external factors, such as the quality of the acquisition equipment, temperature, load conditions, voltage fluctuations, and variations in the battery [13,14]. Building a reliable SOC estimator is crucial for a vehicle's stability and dependability [15,16]. ...

Reference:

State of Charge Estimation Based on Cumulative Error Based‐Extended Kalman Filter for Electric Vehicle Applications
Battery SOC estimation from EIS data based on machine learning and equivalent circuit model
  • Citing Article

  • July 2023

Energy

·

Alessio De Angelis

·

·

[...]

·

Fabrizio Smeraldi

... 29,35,36 In most of these cases, the impedance spectrum is reconstructed via time-domain measurements whose basic input signal architecture is, broadly speaking, a step current perturbation or, equivalently, a current interruption. 17,[37][38][39][40][41][42][43][44][45][46][47] In the current interruption technique, the battery is first held at constant current, which is then rapidly removed. By measuring the following voltage and current relaxations towards open-circuit equilibrium, the impedance spectrum of the battery at a given state-of-charge (SOC) can be reconstructed via Fourier transform of the voltage and current signals. ...

Reference:

Requirements for Applying the Current Interruption Technique for Reconstructing the Impedance of Li-Ion Batteries
Time-Domain Battery Impedance Identification Under Piecewise Constant Current Excitation
  • Citing Article

  • January 2023

IEEE Transactions on Instrumentation and Measurement

Paolo Carbone

·

Alessio De Angelis

·

·

... In the state of charge model, the current usage needs to be tracked by integrating the current draw or input as a function of time. An equivalent circuit model that accounts for nonequilibrium voltage behavior is the equivalent series resistance model, as discussed in Angelis et al. [4]. In this model, a resistance is placed in series with a controlled voltage source. ...

Reference:

Electrochemical–Thermal Model of a Lithium-Ion Battery
On the Usage of Battery Equivalent Series Resistance for Shuntless Coulomb Counting and SOC Estimation
Alessio De Angelis

·

Paolo Carbone

·

·

[...]

·

Luca Ruscitti