The goal is to design sensor data fusion algorithms that can synergistically combine defect-related information from heterogeneous sensors used in gas pipeline inspection to reliably and accurately predict the condition of the pipe wall in an attempt to help maintain the nation’s natural gas transmission and distribution infrastructure through enhanced evaluation of pipeline inspection data.
Rowan University – project management and research product
Glassboro, NJ 08028
Accurate and reliable characterization of the pipe wall condition in gas transmission pipelines requires inspection using more than one method of non-destructive testing. The objective of this research is to develop a suite of sensor data fusion algorithms that synergistically combine signals to obtain information that is not present in data from heterogeneous sensors (i.e. magnetic, ultrasonic, and thermal).
The multi-sensor data fusion algorithms are employed in two stages. In the first stage, data from multiple inspection modalities are fused to identify and separate pipe-wall anomalies from benign indications. A machine learning algorithm that is capable of incremental learning is employed for this purpose. In the second stage, the multi-sensor data is fused to predict the size and shape of those indications, identified as anomalies. Models that are based on human stereoscopic vision are used to design a data fusion processes in order to extract redundant and complementary information among different sets of sensors.
Data fusion techniques (Learn++ and Geometric transformations) were exercised on a test-specimen suite that was fabricated to exhibit common indications (both benign and anomalous) that occur in gas transmission pipelines. This test-specimen suite was subjected to multi-sensor interrogation using ultrasonic testing, magnetic flux leakage, and thermal imaging techniques.
The Learn++ algorithm’s data fusion capabilities were evaluated on the magnetic flux leakage (MFL) and ultrasonic testing (UT) data obtained from the test-specimen suite. Of particular interest was whether the generalization performance of the algorithm could be improved when data from two different non-destructive evaluation (NDE) modalities (i.e. magnetic flux leakage (MFL) and ultrasonic testing (UT) are fused. To test this performance, the test-specimen suite was partitioned into two sections, ten to be used for training the algorithm, and eleven to be used for validation. A total of five categories (namely pitting, crack, mechanical damage, weld, and no defect) were used for classification. The goal was to compare the classification performances when trained with only MFL or UT data to the performance when the data was combined through the Learn++ algorithm. The initial results were very promising.
In order to train the artificial neural network for defect characterization, it is necessary to indicate the desired complementary and redundant information between the two NDE inspection methods. Complementary information in two NDE images are defined as those distinct pixels in each of the NDE signatures that are present in the defect region, but are not shared between them. Redundant information in two NDE images is defined as those common pixels that are present in both NDE signatures and present in the defect region. The results of tests indicated that the proposed technique is able to successfully extract redundant and complementary information related to geometry of the defect. The best results are obtained for the fusion of UT and MFL data, and the poorest performance results combined thermal imaging and any other method. These results are consistent with the quality and amount of information contained in these respective modalities.
Inspection and evaluation of natural gas pipeline condition is a critical element in assuring delivery reliability, safety and operational functionality of the nation’s energy infrastructure. Inspection is carried out using different non-destructive evaluation techniques that concentrate on different aspects of potential pipeline damage which can reduce safe and reliable operating pressure and / or life of the pipeline. This technology’s potential impact lies in its ability to take unique evaluation data from different sources and generate a more robust and reliable picture of the true condition of the infrastructure by using complementary data to confirm areas of concern and signal differences. This limits false alarms which would cause unnecessary excavations and repairs. By offering an opportunity to enhance the clarity of pipeline conditions, the project holds the potential to: 1) reduce lost product through leaks developed from unfound damage, 2) reduce emissions which could result from leaks and 3) increase the overall safety and reliability of the pipeline system by ensuring thorough evaluation of the available inspection data.
The project is complete and the final report is listed below under "Additional Information".
Final Report [PDF-2932KB]