Sparse Transducer Network for Monitoring of Fouling Removal in Pipelines
Affiliation Type:
Academia
Keywords:
Non-destructive testing, Industry 4.0, Wireless sensor network, Machine learning
Abstract:
We present a wireless sensor network (WSN) using ultrasonic transducers and machine learning (ML) methods. This WSN is used for monitoring cleaning processes in industrial environments. The WSN detects fouling by conducting multiple pitch-catch measurements with all possible transmitter-receiver combinations. The ML model estimates the location and extent of the fouled area based on wave packet amplitude comparisons of A0 Lamb waves between measurements. The WSN and the ML model were evaluated by localizing a gypsum fouling proxy on a stainless steel pipe (⌀ = 154 mm, wall thickness 2 mm) using four 400 kHz transducers. Parts of the fouling proxy were sequentially removed to emulate a cleaning process. In addition, the fouling removal process was monitored by comparing data between two fouled states to emulate an environment with no access to a clean reference state. The results show that the WSN along with the ML model can monitor both the accumulation of fouling and the progression of the cleaning process. In conclusion, the proposed methodology could supervise the cleaning process and notify when consecutive cleaning cycles are no longer effective.
