|FHWA > Engineering > Pavements > Research > Concrete > Use of Magnetic Tomography Technology to Evaluate Dowel Placement|
Use of Magnetic Tomography Technology to Evaluate Dowel Placement
LABORATORY EVALUATION (continued)
The overall standard deviation and confidence interval for horizontal and vertical alignment results of MIT Scan-2 were determined based on laboratory testing results and consideration of other sources of error. The laboratory testing results confirmed that the repeatability error and device error are within the MITspecified limits as follows:
Other sources of error include the following:
The magnitudes of errors listed above are peak-to-peak values. The peak-to-peak values are conservatively assumed to represent 95-percentile values. The standard deviation of measurement error from each source was obtained by dividing the peak-to-peak error values by the standard normal variate corresponding to 95 percent probability (1.64). Assuming all sources of errors are independent, the overall variance was obtained by summing the component variance. The components of measurement error on rotation (horizontal and vertical alignment) are summarized in Table 5. The overall standard deviation of measurement error is 3.0. Since the uneven surface will have minimal effect on horizontal alignment, the overall standard deviation for horizontal misalignment may be somewhat less than that for vertical misalignment, but the difference is not significant (2.8 mm [0.1 in.], rather than 3.0 mm [0.12 in.]).
Based on overall standard deviation of 3.0 mm (0.12 in.), MIT Scan-2 may be expected to provide measurement accuracy of +5 mm (0.20 in.) with 95 percent reliability. However, it is important to note that any metal objects (tie bars, nails in the joint, coins, pieces of wire, or other metal) near the measurement region can introduce significant errors. The influence of the foreign metal objects cannot reasonably be incorporated in the overall standard deviation of measurement errors. The presence of significant metal objects (e.g., tie bars, nails, pieces of wire, or other significant mass of metal) essentially invalidates the results for the bars within the influence region of the metal objects. In most cases, the presence of foreign metal objects is easily detectable on the signal intensity plot. Whenever MIT Scan-2 results indicate a large dowel misalignment, close inspection of the signal intensity plot and the evaluation results is advisable to ensure that the results are real and not due to external influence.
The most problematic of the external influences is the presence of tie bars in close proximity of the joint being evaluated. Typically, but not always, the presence of tie bars is clearly visible on the signal intensity plot (Figure 23). Figure 24 shows a case in which the presence of tie bars is not easily discernible from the signal intensity plot. In this case, the only indication of external influence is the sudden jump in side shift results that are not consistent with the signal intensity plot, as indicated in Table 6. If tie bars occur at a consistent location (relative to the dowel bars), the effects of tie bars can be filtered out. But the tie bar locations are highly variable. The inability to obtain accurate results for dowel bars influenced by tie bars is a critical limitation of MIT Scan-2.
The accuracy of MIT Scan-2 results, determined based on laboratory testing, were verified using the results of testing conducted at the MIT GmbH test track (Figure 6). The test track offers the means of taking very accurate measurements of the actual dowel bar positions under simulated field conditions. A factorial of tests was conducted covering the full operating range of MIT Scan-2 using seven dowel bars. In each case, six of the dowel bars were placed with similar side shift and minimal misalignment, and the position of one of the middle dowels was varied according to the factorial design of the testing program. This test simulates the conditions in real joints, in which most of the bars are in good alignment but one bar has a varying degree of misalignment. The results of this test will show the effects of any influence of neighboring dowel bars on measurement accuracy.
The results of the validation tests are shown in Figures 25 through 34. Similar to the laboratory testing results, this series of tests showed that, within the operating range, the magnitudes of measurement errors are independent of other selected parameters. For example, Figure 25 shows that the errors in horizontal misalignment results are not dependent on side shift. Similarly, Figure 27 shows that the errors in horizontal misalignment results are not dependent on depth. Figure 33 still shows a slight bias in the depth results, but not to the extent shown in Figure 21. Figure 34 shows that the side shift results are well within the specified range of accuracy, which suggests that the bias shown in Figure 22 for the laboratory testing may very well be due to errors in manual measurements.
The validation test results confirmed the findings of the laboratory evaluation, that MIT Scan-2 provides accuracy of +5 mm (+0.2 in.) with 95 percent reliability. The measurement errors exceeded +5 mm (+0.2 in.) on a few occasions. However, the error in horizontal misalignment measurements exceeded +5 mm (+0.2 in.) only 0.6 percent of the time, and the error in vertical misalignment measurements exceeded +5 mm (+0.2 in.) only 1.2 percent of the time. As noted under laboratory testing, it is important to remember that the conclusions regarding the accuracy and reliability of MIT Scan-2 are valid only in the absence of any external influence (mainly, the presence of foreign metal, such as tie bars in close proximity).
For dowel bars placed in baskets, the presence of the metal basket interferes with MIT Scan-2 results. However, if the transport ties in the basket are cut, good results can be obtained, even without any special considerations for the dowel basket. Without basket calibration, the results for dowel baskets are more sensitive to side shift. If the basket is well centered under the joint saw cut (e.g., side shift less than about +38 mm [1.5 in.]), the horizontal and vertical misalignment results are very good. The reported depth is less (by about 6-7 percent), because the presence of the additional metal (basket) makes the bars appear closer to the surface. With specific calibration and basket software, a similar level of accuracy can be obtained for dowel baskets as for bare bars.
MIT GmbH is developing a special version of MagnoProof and MagnoNorm to handle dowel baskets. For accurate quantitative results, the basket software must be used with specific calibration for the type of basket scanned. Because dowel baskets are calibrated as a whole unit, the baskets for different size dowel bars must be calibrated separately.
Limited testing was conducted using the beta-test version of the basket software. The results for a test conducted in open air are summarized in Table 7. As shown in Table 7, when the basket is calibrated, a similar level of accuracy can be obtained for dowel baskets as for bare bars.
For dowel baskets, the qualitative results from MIT Scan-2 (e.g., approximate numerical results and graphical output) can be as useful as the more precise numerical results obtained by having the correct basket calibration. In general, with proper inspection prior to paving, excellent dowel alignment can be achieved as long as the basket does not move during paving. The type and number of pins used to anchor the basket in place is the critical factor. If the basket is properly anchored, the dowel alignment will not change during paving. However, if the baskets are not adequately anchored, the basket can deform, burst open, or move during paving. The occurrences of such problems are very easy to detect from graphical output of MIT Scan-2. Figure 35 provides examples, clearly visible in the graphical output, of the problems resulting from inadequate anchoring of dowel baskets. These types of problems cause severe dowel misalignment, which can also be reliably detected from the approximate numerical results obtained using standard calibration (i.e., without basket-specific calibration).