Performance of hybrid decomposition algorithm under heavy noise condition for health monitoring of structure

• Published in: Journal of civil structural health monitoring Vol. 10; no. 4; pp. 679 - 692
• Main Authors: Das, Swagato, Saha, Purnachandra
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.09.2020; Springer Nature B.V
• Subjects: Algorithms; Benchmarks; Civil Engineering; Condition monitoring; Control; Damage assessment; Damage detection; Decomposition; Dynamical Systems; Empirical analysis; Engineering; Measurement Science and Instrumentation; Noise; Noise monitoring; Original Paper; Parameter identification; Resonant frequencies; Robustness; Sensors; Signal processing; Structural damage; System identification; Vibration; Wavelet analysis; Variational mode decomposition (VMD); Heavy noise contamination; Frequency domain decomposition (FDD); Empirical mode decomposition (EMD); ASCE benchmark structure; Singular sensor output
• ISSN: 2190-5452; 2190-5479
• DOI: 10.1007/s13349-020-00412-5

In this paper, a hybrid damage detection technique involving a combination of variational mode decomposition (VMD) and frequency domain decomposition (FDD) has been applied to study the effectiveness of damage detection in presence of heavily noise contaminated environment. Damage of small magnitude has been tested under Gaussian pulse noise ranging from 0 to 100% for damage analysis. FDD, a signal processing algorithm, has system identification capability in the presence of noise but requires the output acceleration data from all the sensors installed in the nodes of a structure to identify damage. To reduce the number of sensor data needed to identify damage, wavelet-based algorithms have been used to obtain intrinsic mode functions (IMFs) from single sensor output. These IMFs are then fed to FDD algorithm to obtain the natural frequencies of the structure. For comparison purpose, the algorithms (empirical mode decomposition (EMD) + FDD, and VMD + FDD) have been applied to ASCE benchmark building, which has been set as a common platform, using sensor data of first storey. It was observed that the VMD + FDD gives satisfactory damage identification results for 100% noise contamination whereas EMD + FDD was unable to identify damage accurately for noise above 20%. The robustness of VMD + FDD has been established for a different type of noise, random-valued impulse noise, applied on the benchmark structure for detecting the structural parameters. The hybrid algorithm was also checked for system identification using the sensor data of fourth storey to establish its robustness against the sensitivity of the sensor location.