1,720,969 research outputs found
A Multidimensional Health Indicator Based on Autoregressive Power Spectral Density for Machine Condition Monitoring
Full text linkCondition monitoring (CM) is the basis of prognostics and health management (PHM), which is gaining more and more importance in the industrial world. CM, which refers to the tracking of industrial equipment’s state of health during operations, plays, in fact, a significant role in the reliability, safety, and efficiency of industrial operations. This paper proposes a data-driven CM approach based on the autoregressive (AR) modeling of the acquired sensor data and their analysis within frequency subbands. The number and size of the bands are determined with negligible human intervention, analyzing only the time–frequency representation of the signal of interest under normal system operating conditions. In particular, the approach exploits the synchrosqueezing transform to improve the signal energy distribution in the time–frequency plane, defining a multidimensional health indicator built on the basis of the AR power spectral density and the symmetric Itakura–Saito spectral distance. The described health indicator proved capable of detecting changes in the signal spectrum due to the occurrence of faults. After the initial definition of the bands and the calculation of the characteristics of the nominal AR spectrum, the procedure requires no further intervention and can be used for online condition monitoring and fault diagnosis. Since it is based on the comparison of spectra under different operating conditions, its applicability depends neither on the nature of the acquired signal nor on a specific system to be monitored. As an example, the effectiveness of the proposed method was favorably tested using real data available in the Case Western Reserve University (CWRU) Bearing Data Center, a widely known and used benchmark
Minimal representations of MIMO time-varying systems and realization of cyclostationary models
No full textThis paper describes a family of minimally parameterized state-space models for completely n-step observable multi-input multi-output time-varying systems and their algebraical links with equivalent input-output models. These results are then used to develop an algorithm for the realization of generic input-output sequences generated by a periodic system. A complete numerical example is also developed to illustrate all steps of the algorithm described in the paper. (C) 2003 Elsevier Ltd. All rights reserved
Residual design for dynamic processes using de-coupling technique
No full textThe work presents some results concerning a fault detection scheme for dynamic processes using disturbance decoupling technique. The first step of the considered approach consists of exploiting input-output descriptions of the monitored system. In particular, the disturbance term of that model can be used to take into account unknown inputs affecting the system. The next step of the scheme leads to define a set of parity relations that can be used as residual signals since they are insensitive to the disturbance term. The proposed fault detection procedure has been tested on an industrial process simulator. Sensor and actuator faults have been simulated on a gas turbine model. Simulation results and concluding remarks have been finally reported
Combining Wavelets and AR Identification for Condition Monitoring of Electric-cam Mechanisms Using PLCopen Readings of Motor Torque
Full text linkThis paper addresses the problem of monitoring the state of health of electric motor driven mechanisms. The proposed condition monitoring procedure belongs to the data-driven methods and employs a combination of wavelet analysis and autoregressive model identification. It exploits the fact that the torque motor signal is a readily available measurement in industrial computers complying with the PLCopen standard and how motion controllers execute electric cams. In particular, the torque provided by the PLC is represented as the sum between the ideal torque and an additional contribution that contains information about mechanism health condition. The procedure completely removes the ideal torque and analyzes the residual component to highlight and classify possible fault conditions. The described condition monitoring procedure is tested on real data in a laboratory setup
Errors-in-Variables Identification of Composite Noncausal-FIR/IIR Models with Application to Transmissibility Identification
No full textModel structures used for system identification include infinite impulse response (IIR) models and finite impulse response (FIR) models. Identification using IIR models requires knowledge of the order of the system, where underestimating or overestimating the order of the system can yield poor parameter estimates. Although identification using FIR models does not require knowledge of the order of the system, FIR models cannot approximate systems with poles on or outside the unit circle. Noncausal FIR models can approximate systems with asymptotically stable and unstable poles, but not systems with poles on the unit circle. A composite noncausal-FIR/IIR (CNFI) model has an IIR part and a noncausal-FIR part, where the IIR part approximates poles on the unit circle, and the FIR part approximates the remaining part of the system. In this paper, we propose an errors-in-variables identification algorithm for CNFI models. We apply the proposed algorithm to identify transmissibilities, which are models that characterize the relationship between the outputs of an underlying system
Thermal Model Identification of Computing Nodes in High-Performance Computing Systems
Full text linkThermal-Aware design and online optimization of the cooling effort are becoming increasingly important in current and future high-performance computing (HPC) systems. A fundamental requirement to effectively develop such techniques is the availability of distributed and compact models representing the system thermal behavior. System identification algorithms allow to extract models directly from the thermal response of the target device. This article proposes a novel thermal identification approach for real, in-production HPC systems, which is capable of extracting thermal models from a computing node affected by quantization noise on the temperature measurements as well as operating in the free-cooling mode, with variable ambient temperature. The approach allows also to identify the physical floorplan of the CPU dies in supercomputing nodes. The effectiveness of the proposed methodology has been tested on a node of the CINECA Galileo Tier-1 supercomputer system
RECURSIVE IDENTIFICATION OF NOISY AUTOREGRESSIVE MODELS VIA A NOISE-COMPENSATED OVERDETERMINED INSTRUMENTAL VARIABLE METHOD
No full textThe aim of this paper is to develop a new recursive identification algorithm for autoregressive (AR) models corrupted by additive white noise. The proposed approach relies on a set of both low-order and high-order Yule-Walker equations and on a modified version of the overdetermined recursive instrumental variable method, leading to the estimation of both the AR coefficients and the additive noise variance. The main motivation behind our proposition is introducing model identification procedures suitable for implementation on edge-computing platforms and programmable logic controllers (PLCs), which are known to have limited capabilities and resources when dealing with complex mathematical computations (i.e., matrix inversion). Indeed, our development is focused on condition monitoring systems, with particular attention paid to their integration onboard industrial machinery. The performance of the recursive approach is tested using both numerical simulations and a laboratory case study. The obtained results are very promising
Condition monitoring of ball bearings using estimated ar models as logistic regression features
No full textBearings are one of the most common components in automatic machines. Diagnosis and prognosis of their working condition is crucial for minimization of downtime and maintenance costs. Different approaches may be adopted to either solve or mitigate the problem of identifying incipient faults during machinery operations. In this paper, we propose a simple and efficient yet effective method to solve this problem by exploiting the edge-computing capabilities of PLCs. Accelerometer signals are modeled as AutoRegressive (AR) processes whose coefficient are used as features for machine learning, based on logistic regression algorithm (LR), to perform Fault Detection and Isolation (FDI). Estimation and prediction are both implementable on-board the PLC, while machine learning can be carried out remotely, in a cloud computing perspective. The exploitation of AR modelling gives a simple and inherent methodology for feature selection. We apply the procedure to the Case Western Reserve University database, a widely known and used benchmark, to highlight its performance with respect to similar fault recognition techniques
Residual generation and identification for dynamic processes
No full textThe work presents a preliminary study concerning fault detection for
dynamic processes using disturbance de-coupling technique. The first step of
the considered approach consists of exploiting input-output descriptions of the
monitored system. In particular, the disturbance term of that model can be used to
take into account unknown inputs affecting the system. The next step of the scheme
leads to the definition of basic parity relations that can be used as de-coupled
residual signals, as they are insensitive to the disturbance term. The proposed fault
detection procedure has been tested on an industrial pro. ~ss prototype. Sensor and
actuator faults have been simulated on a gas turbine model. Simulation results and
concluding remarks have been finally reported
Robust identification of thermal models for in-production High-Performance-Computing clusters with machine learning-based data selection
Full text linkPower and thermal management are critical components of High-Performance-Computing (HPC) systems, due to their high power density and large total power consumption. The assessment of thermal dissipation by means of compact models directly from the thermal response of the final device enables more robust and precise thermal control strategies as well as automated diagnosis. However, when dealing with large scale systems “in production", the accuracy of learned thermal models depends on the dynamics of the power excitation, which depends also on the executed workload, and measurement nonidealities, such as quantization. In this paper we show that, using an advanced system identification algorithm, we are able to generate very accurate thermal models (average error lower than our sensors quantization step of 1∘C) for a large scale HPC system on real workloads for very long time periods. However, we also show that: 1) not all real workloads allow for the identification of a good model; 2) starting from the theory of system identification it is very difficult to evaluate if a trace of data leads to a good estimated model. We then propose and validate a set of techniques based on machine learning and deep learning algorithms for the choice of data traces to be used for model identification. We also show that deep learning techniques are absolutely necessary to correctly choose such traces up to 96% of the times
- …
