Magnetic Particle Inspection...just how good is it?
TWI Bulletin, September - October 2009
...finite element analysis reveals the answer
Capucine Carpentier joined TWI in 2006 as Project Leader in the NDT Technology Group with the equivalent of a Master's degree in Material Science and Mechanics from the Institut Supérieur de Technologie, Toulouse, France. She developed knowledge in NDT during an internship in the EADS Common Research Center prior to joining TWI. At TWI, Capucine has gained experience in procedure writing, site inspection and has concentrated on the development of modelling as a tool (CIVA, COMSOL) for the creation and the optimisation of inspection techniques.
The study aimed to evaluate the performance of magnetic particle inspection (MPI) carried out on pressure vessel/nozzle welds using finite element analysis. This study follows an enquiry from an Industrial Member in relation to an MPI previously performed for the detection of surface breaking flaws. As Capucine Carpentier reports, the client contacted TWI to investigate the possibility of determining the effectiveness of MPI on a given geometry, and the limitations for detection of flaws.
The MPI used a permanent magnet (type Magnaflux Y5 yoke) on the welded joints between a nozzle (41mm thick, diameter 20inch - forged steel) and vessel steel. The MPI was designed for the detection of surface breaking flaws. The study was based on the geometry provided by the client. It was assumed that this inspection was performed in accordance with European standards.
The study was performed in three phases:
- Demonstrate the limits of accuracy of the model and determine the magnetic flux density level on the test surface for an input flux in accordance with EU standard in the cases nine and 12 o'clock.
- Determine the minimum size of detectable flaws for the cases at nine and 12 o'clock by predicting field leakage from three defects and relating them to the formation of magnetic particle indications.
- Demonstrate the validity of the study through MPI tests on representative weld samples.
Phase 1: Prediction of magnetic flux density
Magnetic flux density on test surfaces
The same software package was used for the modelling of the MPI on vessel/nozzle welds cases 9 o'clock and 12 o'clock.
The material property used to define the nozzle and vessel component was the hysteresis curve of C-Mn steel. The magnetic property of the weld and the heat affected zone (HAZ) were assumed to be the same as in the parent material. The shape of the weld was simplified with flat cap and sharp corners.
Shapes and dimension of the yoke 5 type permanent magnet were measured to provide input to the models. The distance from the magnet poles to the weld on the nozzle and vessel surfaces was arbitrarily taken. An air gap of 0.5mm was modelled between the magnet poles and the component surface to simulate irregularity in the contact surfaces. Modelling configuration for the 9 and 12 o'clock cases are shown in Figure 1. Magnetic field input into the material at the magnet poles was one Teslar (1T).
Fig.1. Modelling configuration: a) 9 o'clock
To evaluate whether the given inspection delivered adequate flux density into the component in respect of the EN standards, the magnetic flux density into the vessel and nozzle component and the magnetic flux density along the weld and HAZ surface at 0.1mm from the test surface were computed.
British Standard 6072:1981 requires that magnetic particle flaw detection shall be carried out at a level of magnetic flux density (tangential component) on the test surface equal to, or greater than, 0.72T. This standard also specifies that when a permanent magnet is used as the magnetisation technique, the lifting power of this permanent magnet shall be at least 18kg. It was assumed in this study that the permanent magnet used satisfied this condition. The European standard 9934-1:2001 translates this condition by a minimum level of magnetisation required in the component: the minimum flux density in the component surface shall be 1T.
Phase 2: Prediction of flaws detectability
The capacity to detect surface breaking flaws with MPI is dependent on the:
- Size of the magnetic leakage from the surface breaking flaw
- Creation of an indication by the magnetic particles
- Visual detection of the indication created
To simplify the study, the issues regarding the condition of the visual inspection were not taken in account. It was assumed that all indications once initiated were detectable.
The effect of the surface breaking flaw depth on the magnetic leakage was investigated. Three defects of depths 1, 0.5 and 0.1mm with identical position and orientation, and 0.1mm wide were modelled at the weld toes for both cases 9 and 12 o'clock.
A literature review was carried out to find relationships between magnitude of magnetic leakage induced by a surface breaking flaw on the test surface and the movement of the magnetic particles to create visible indications. Two key papers were found on this subject.
Edward and Palmer (1985) state that the movement of the particles will be initiated when the magnetic force Fm applied on the particle is greater than twice the gravitational force Fg experienced by the particle, ie this is the threshold:
Fm≥2Fg
Where an approximation for Fg= 2x10-14N for a particle in water was used.
The magnetic force applied on a single 1 micron (µm) diameter spherical particle due to the leakage of a surface breaking crack that is greater than 10 particle diameters away from the crack is given by Blakeley et al (1985) as:
The force on any one magnetic particle in a static magnetic field on the test surface is proportional to the product of the field strength (Hx), the field gradient (dHx/dx), the particle volume (v), and the relative permeability of the magnetic particle (µr).
Therefore neglecting the effect of particle-particle interactions, the initial movement of a particle depends on the tangential magnetic field leakage and its rate of change (gradient). These two components were calculated by the modelling computations for each of the defect/weld configurations.
Phase 3: Practical trials
To demonstrate the validity of the study, MPI employing a permanent magnet (Y5 yoke) was conducted on representative weld samples available in TWI.
Tests were performed on a weld joining a nozzle 25mm thick with a vessel 46mm thick, diameter 20inch. The nozzle was oriented at 90° to the vessel. Two cracks were implanted in the weld toe of the thick plate (equivalent in the model to Toe B) as shown in Figure 2.
Fig.2. Location of cracks 1 and 2 on TWI nozzle
An Eddy current test was carried out on the weld to verify the presence of the cracks and evaluate in comparison with calibration notches the depth of these cracks. A dedicated eddy current probe for weld inspection was used at 100kHz.
Results
Phase 1: Prediction of magnetic flux density
9 o'clock: Magnetic flux density in weld
Figures 3a and 3b show the distribution and orientation of the magnetic flux density in the parent material and the weld delivered by the permanent magnet for the 9 o'clock configuration.
Fig.3. Representation of the predicted magnetic flux density and magnetic flux lines orientation for nozzle/vessel weld of the 9 o'clock configuration: a) In nozzle and vessel component, weld and permanent magnet
The permanent magnet delivers 1.3T in the thin plate and 0.9T in the thick plate. With the same magnetisation for both magnet poles based on 1T input, the magnetic flux density in the thick plate is 30% smaller than in the thin plate. This implies that the thickness of the component in which the MPI is performed will have an influence on the test.
Looking at the measurements along the test surface, HAZ and weld, the tangential magnetic flux density is largely above the 0.72T threshold required to satisfy the standard. Hence the use of a DC permanent magnet type yoke is acceptable for the configuration 9 o'clock.
From Fig.3, it can be noted that the magnetic flux is concentrated at the weld toes. Saturation of the magnetic field at the weld toes can create spurious magnetic particle indications. This can lead to a decrease in the capability for flaw detection. The phase two and three studies on the capability of flaw detection of this inspection will demonstrate whether this situation will lead to spurious indications.
12 o'clock: Magnetic flux density in weld
Similar conclusions for the 12 o'clock are noted for the 9 o'clock configuration on the reduction of the magnetic flux in thick sections and also the concentration of magnetic fields.
Also looking at the measurement along the test surface, HAZ and weld, the tangential magnetic flux density is largely above the 0.72T threshold required to satisfy the standard. Hence the use of a DC permanent magnet type yoke is acceptable for the configuration 12 o'clock.
Phase 2: Prediction of flaws detectability
9 o'clock: Flaw detectability
Figures 4a and 4b show the leak of the magnetic field induced by the weld toe and a surface breaking flaw. This confirms the possible interaction between the magnetic field leakage and magnetic particles. Particles are disturbed mostly by the tangential magnetic field on the test surface.
Fig.4. Magnetic field leakage at the weld toe in configuration 9 o'clock a) From the toe itself
b) From 1mm deep surface breaking flaw at the toe
Curves in Fig.5 represent the measurements of the magnetic field 0.1mm in air along the test surface at positions A (thin parent material section) and B (thick parent material section). The origin of the horizontal axis is taken at the central point of the flaws. In both positions, flaws on the weld toe produce a stronger magnetic leak, as well as a flaw deeper into the weld producing a stronger leak. Flaws occurring at the weld toe in position A (thinner parent material section) also generate a larger leak as shown in Section 3.1.
Fig.5. Predicted magnetic flux leakage from surface breaking flaw at the weld toe at position A for the 9 o'clock configuration
Figure 6 demonstrates whether the excessive leakage from flaws at positions A and B are enough to create an indication. For each position and flaws, the magnetic force induced from the magnetic leakage on the test surface is compared with the magnetic force threshold at which the movement of particles is initiated leading to an indication.
Fig.6. Magnetic force induced from weld toe and surface breaking flaw in position A for the 9 o'clock configuration compared with the magnetic force threshold (blue line) for: a) Weld toe
The analysis of the variation of the magnetic force along the test surface shows:
- Magnetic particles are not affected by the weld toe in positions A and B.
- A surface breaking flaw 0.1mm deep can generate an indication in position A. However the creation of an indication in position B is on the borderline with regard to the threshold required.
- Particles are more sensitive to a 0.5 and 1mm deep flaws. These flaws will create an indication in positions A and B.
12 o'clock: Flaw detectability
Results were similar to those found for the nine o'clock configuration.
Figures similar to those in Figure 4 show the magnetic leakage produced by the weld toe and the surface breaking flaw.
Tangential magnetic field leakage is stronger when a surface breaking flaw is present at the toe and the magnitude of the magnetic leakage increases with the depth of the flaw. Also, the leak at the weld toe from flaws on the thick parent material section (position B) is smaller than on the thin parent material section (position A).
The analysis of the variation of the magnetic force along the test surface shows:
- Magnetic particles are not affected by the weld toes.
- A surface breaking flaw 1mm deep in configuration 12 o'clock will create an indication both on the thick and the thin sections.
- A surface breaking flaw 0.5mm deep will be detected on the thin section. However the predicted magnetic leakage on the thick section will not be strong enough to create a clear indication.
- A surface breaking flaw 0.1mm deep at either of the weld toes will be impossible to detect in configuration 12 o'clock.
Moreover flaw detection in configuration 12 o'clock is less sensitive than for the configuration 9 o'clock. In configuration 9 o'clock, with the same magnetic input, the magnetic field in the material is stronger, leakage from potential flaws at the weld toes are larger and surface breaking flaws down to 0.5mm deep can be detected.
Phase 3: Practical trials
In comparison with eddy current signatures from calibration slots, the depths of the cracks were estimated in:
- Location 1 to be minimum 0.2 to 0.5mm
- Location 2 to be minimum 0.5 to 1mm
In location 1, an MPI indication was seen at the weld toe. However it was not clear if it was due to the crack or spurious indications from the weld toe (Fig.7).
Fig.7. MPI indication for crack detection in location 1.
In location 2, the crack was clearly detected by the MPI (Fig.8). As predicted in the model, a crack greater or equal to 0.5mm deep can be detected using the MPI procedure employing a permanent magnet yoke.
Fig.8. MPI inidcation for crack detection in location 2.
Conclusions
- The use of the finite element modelling for this particular application was validated.
- The inspection cases (at 9 and 12 o'clock) using a DC yoke magnet are in accordance with the requirements of EN standards.
- The results imply that for 9 o'clock configuration:
- On the weld toe to the thin wall (Toe A), a crack greater than or equal to 0.1mm deep can be detected.
- On the weld toe to the thick wall (Toe B), a crack greater than or equal to 0.5mm deep can be detected.
- The results imply that for 12 o'clock configuration:
- On the weld toe to the thin wall (Toe A), a crack greater than or equal to 0.5mm deep can be detected.
- On the weld toe to the thick wall (Toe B), a crack greater than or equal to 1mm deep can be detected.
- MPI on weld sample confirms predictions.
Recommendation
The initial study shows the possibility of using the finite element model COMSOL to predict the complete MPI inspection for a given case. The methodology proposed in this study should be qualified and refined using a much larger set of flaws to make it available as a routine tool for inspection design.
This study is an example of using modelling to design and validate NDT inspection.
TWI has lengthy experience of using modelling for ultrasonic inspection. This study is the first time it has been performed for MPI.
