TL;DR
This paper introduces a novel Residual Magnetic Field Testing (RMFT) system using an array of Tunneling Magneto-Resistive (TMR) sensors and permanent magnets to detect cracks in ferromagnetic pipes. The system offers a more compact and potentially lower-power alternative to existing methods like Magnetic Flux Leakage (MFL) by requiring only partial magnetization.
What Problem Does This Paper Address?
Ferromagnetic pipes, commonly used in the oil and gas industry, are susceptible to cracks caused by corrosion, pressure, and fatigue. Early detection of these cracks is crucial to prevent catastrophic failures like leaks, fires, and explosions. While several non-destructive testing (NDT) methods exist, such as Magnetic Flux Leakage (MFL), Ultrasonic Testing (UT), Eddy Current Testing (ECT), and Alternating Current Field Measurement (ACFM), they each have limitations. MFL requires full pipe magnetization, leading to bulky equipment and high power consumption. UT needs a coupling agent, and ECT/ACFM struggle with deep defects in ferromagnetic materials. This paper aims to address these limitations by developing a more efficient and sensitive RMFT system.
Methods
The core of the proposed system is a detection probe that integrates an array of permanent magnets for partial magnetization and an array of TMR sensors for collecting residual magnetic field signals. This allows for simultaneous magnetization and signal acquisition, simplifying the inspection process.
The authors first developed a theoretical framework for RMFT, including a finite element model (FEM) using COMSOL 6.0 to simulate the magnetic field behavior around cracks of varying lengths and depths. This simulation helped identify key signal characteristics (peak-to-peak signal amplitude $\Delta B_x$, peak-to-peak differential signal amplitude $\Delta d(B_x, z)$, and the distance between peak and trough in the differential signal $l_x$) that correlate with crack dimensions.
Based on these simulations, an experimental RMFT system was built. This system comprises the probe, a data transmission cable, a data acquisition card (DAQ), and a PC for signal processing. The probe itself houses 8 permanent magnets and 8 TMR sensors. TMR sensors were chosen for their high sensitivity and low noise levels, making them suitable for detecting the weak residual magnetic fields. The experimental setup involved scanning a carbon steel pipe with machined cracks, collecting the magnetic signals, and analyzing them using the identified characteristics.
Datasets
The paper does not explicitly mention using pre-existing datasets for training or evaluation. Instead, it relies on:
- Simulated data: Generated through the finite element model developed in COMSOL 6.0, simulating the RMFT signals for cracks of various lengths and depths.
- Experimental data: Collected from a new carbon steel pipe with 8 machined cracks of known dimensions. This serves as the benchmark for validating the RMFT system's performance.
Key Insights and Ideas
- Partial Magnetization Sufficiency: The RMFT method, unlike MFL, can effectively detect cracks with only partial magnetization of the pipe wall. This is a significant advantage for system miniaturization and power efficiency.
- TMR Sensor Advantage: TMR sensors, with their high magnetoresistance ratio and sensitivity, are well-suited for detecting the weak residual magnetic fields generated by partially magnetized pipes, outperforming traditional sensors like Hall effect sensors.
- Characteristic $l_x$ as a Key Indicator: The study found that the characteristic $l_x$ (distance between peak and trough in the differential magnetic signal) showed a strong, near-linear correlation with crack length ($L$) in both simulations and experiments. This makes it a reliable parameter for estimating crack length.
- Crack Depth Influence: While crack depth ($D$) influences the overall signal amplitude ($\Delta B_x$), its effect on the characteristic $l_x$ is minimal, suggesting that $l_x$ is primarily indicative of length, not depth.
- Experimental Validation: The experimental results largely align with simulation predictions, confirming the feasibility of the proposed RMFT system for crack detection in ferromagnetic pipes.
Key Results
- Crack Length Estimation: The characteristic $l_x$ demonstrated a strong correlation with crack length, with experimental results showing a near-linear increase in $l_x$ as crack length increased. For example, $l_x$ values ranged from 9.12 mm to 35.70 mm for crack lengths from 10 mm to 40 mm.
- Crack Depth Influence on Amplitude: The peak-to-peak signal amplitude ($\Delta B_x$) was found to increase monotonically with crack depth, agreeing with simulation results. For instance, $\Delta B_x$ values increased from 1.88 × 10³ mV to 2.77 × 10³ mV as depth increased from 3 mm to 7 mm.
- Limited Depth Indication: The characteristic $l_x$ was found to be largely insensitive to crack depth in both simulations and experiments, indicating its primary utility for length estimation.
- System Sensitivity: The TMR sensors used in the probe offer high sensitivity (30 mV/V/Gs) and low intrinsic noise (750 pT/√Hz at 1 Hz), enabling the detection of weak magnetic signals.
Limitations and Future Work
- Crack Depth Quantification: While the system can detect cracks and estimate their length using $l_x$, directly quantifying crack depth from the analyzed characteristics is challenging. The paper notes that $\Delta B_x$ increases with depth, but the relationship is not as straightforward as for length.
- Signal Irregularities: Experimental results for the differential signal characteristic $\Delta d(B_x, z)$ showed some irregularity with crack length and depth, attributed to factors like uneven pipe surfaces and probe vibration.
- Future Work: The authors plan to:
- Develop a novel probe for inspecting the inner surface of pipes.
- Integrate the RMFT system into a pipe pig for in-line inspection of oil and gas pipelines.
- Develop a quantification algorithm for crack sizing, likely incorporating machine learning or polynomial fitting techniques, to improve the accuracy of both length and depth estimation.
Summary updated