Data-Driven Sensor Fault Diagnosis Based on Nonlinear Additive Models and Local Fault Sensitivity

Cartocci, Nicholas; Crocetti, Francesco; Valigi, Paolo; Napolitano, Marcello R.; Fravolini, Mario Luca

doi:10.1109/icar53236.2021.9659449

Cited by 1 publication

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Occasionally in the article, the dependence on

is omitted to simplify the notation. The proposed FD technique is based on analytical redundancy (AR) concepts [ 12 , 13 , 16 ]. It is assumed that a sensor measurement

is approximated by a non-linear additive model consisting of the linear combination of non-linear functions (

and

) defined as follows

where

and

are constant coefficients (to be estimated from data) and

and

are non-linear functions of the variables

and

, respectively, typically representing a rectified linear unit (ReLU) or Gaussians or polynomial splines [ 26 ].…”

Section: Non-linear Additive Models For Fault Diagnosismentioning

confidence: 99%

“…Since real sensor flight data with sensor faults are not easily available, additive faults are artificially injected on the fault-free data to simulate the occurrence of a sensor fault (this approach of generating artificial faulty data is widely used in the FDi community, see for instance [ 11 , 12 , 13 , 14 , 15 , 16 ]).…”

Section: Machine Learning-based Fault Isolation and Estimationmentioning

confidence: 99%

“…In a system characterized by a limited number of monitored variables, directional residuals methods have shown to be a valid alternative to SPM methods. Directional residual methods work by relating possible faulty sensors with the characteristic direction of the faults, isolating the faulty sensor as the one having the smallest distance between the residuals signal and the monitored sensor fault directions [ 12 , 13 , 16 ]. An important variant of the directional residual method is the structured residual approach [ 9 ].…”

Section: Introductionmentioning

confidence: 99%

“…Although GAMs have been used extensively in many application fields, their employment in a data-driven FDi system has not been fully explored. In fact, to date, only a few studies have been presented, such as [ 16 , 27 ]. Motivated by the mentioned issues, in this paper, we propose GAMs to identify non-linear parity relationships in the monitored variables using cubic spline basis functions to characterize the non-linear functions

by exploiting the MARS modeling and estimation algorithm proposed in [ 28 ].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Data-Driven Fault Diagnosis Techniques: Non-Linear Directional Residual vs. Machine-Learning-Based Methods

Cartocci

Napolitano

Crocetti

et al. 2022

Sensors

View full text Add to dashboard Cite

Linear dependence of variables is a commonly used assumption in most diagnostic systems for which many robust methodologies have been developed over the years. In case the system nonlinearities are relevant, fault diagnosis methods, relying on the assumption of linearity, might potentially provide unsatisfactory results in terms of false alarms and missed detections. In recent years, many authors have proposed machine learning (ML) techniques to improve fault diagnosis performance to mitigate this problem. Although very powerful, these techniques require faulty data samples that are representative of any fault scenario. Additionally, ML techniques suffer from issues related to overfitting and unpredictable performance in regions which are not fully explored in the training phase. This paper proposes a non-linear additive model to characterize the non-linear redundancy relationships among the system signals. Using the multivariate adaptive regression splines (MARS) algorithm, these relationships are identified directly from the data. Next, the non-linear redundancy relationships are linearized to derive a local time-dependent fault signature matrix. The faulty sensor can then be isolated by measuring the angular distance between the column vectors of the fault signature matrix and the primary residual vector. A quantitative analysis of fault isolation and fault estimation performance is performed by exploiting real data from multiple flights of a semi-autonomous aircraft, thus allowing a detailed quantitative comparison with state-of-the-art machine-learning-based fault diagnosis algorithms.

show abstract

“…Occasionally in the article, the dependence on

is omitted to simplify the notation. The proposed FD technique is based on analytical redundancy (AR) concepts [ 12 , 13 , 16 ]. It is assumed that a sensor measurement

is approximated by a non-linear additive model consisting of the linear combination of non-linear functions (

and

) defined as follows

where

and

are constant coefficients (to be estimated from data) and

and

are non-linear functions of the variables

and

, respectively, typically representing a rectified linear unit (ReLU) or Gaussians or polynomial splines [ 26 ].…”

Section: Non-linear Additive Models For Fault Diagnosismentioning

confidence: 99%