In addition to advances in hybrid (cooled) thermal sensors, technological progress in low-cost microbolometer (uncooled) infrared focal-plane array sensors have significantly contributed to the broadening of the application of this sensing technique in many fields: Leisure, Manufacturing, Process Control, Building insulation diagnostic, Civil Engineering, etc. The performances, size and power consumption of available uncooled systems have increased the number of use cases, but new performance needs are still emerging.

In this paper, we address the use case of hot axle bearing monitoring by infrared thermography for railway safety.

We first propose a brief background overview and discuss the advantages and disadvantages of current practices. We then introduce and analyze the concept of wayside monitoring.

After this introduction, we present preliminary in-situ experiments conducted at various times of the day on an operational rail line. The image sequences were collected using two camera models, a cooled camera, operating at 250Hz (SWIR - MWIR), and an uncooled camera, operating at 100 Hz (LWIR).

The generated infrared thermal images database is presented and discussed. Two processing strategies for hotbox detection and tracking are presented and discussed. For both strategies, Deep learning (DL)-based object detection algorithm is compared to classical images processing ones. In particular, attention is paid to rejection of false positive induced by environmental conditions during measurements. To complete, preliminary results obtained on a working station and an edge computer are presented and discussed.

The lessons learned from this work are then used to define the requirements for more suitable infrared cameras. As an illustration, we discuss the performance of uncooled sensors and present preliminary laboratory results with new prototypes of uncooled IRFPA cameras developed within the CHIPS JU BRIGHTER project (https://project-brighter.eu/).

Finally, conclusions, challenges and perspectives are proposed for infrared sensors performance requirements, fast processing methods, and computing hardware platforms.

The service life of freight railroad bearings can be extended anywhere from 50,000 to 250,000 miles through reconditioning. The visual inspection process for reconditioning bearing components currently does not account for subsurface defects, which are likely responsible for the variation in performance. In this presentation, nondestructive ultrasonic methods are discussed as a way to account for this inspection weakness. Here, we compare ultrasonic backscatter results from 15 MHz 45° shear scans with backscatter measurements based on surface waves. All measurements were made on the outer rings (cups) of eight reconditioned bearings, which are the subcomponents that experience the highest level of rolling contact fatigue. Shear waves and surface waves have different effective focal depths such that subsurface defects on these bearing components can be effectively compared. Subsurface regions were selected because the maximum shear stress from rolling contact fatigue is subsurface and can be the sites of microcrack initiation. 360° scanning profiles were conducted for both raceways with a horizontal overlap of 0.5 mm before being vertically indexed by 1 mm. After the cups were scanned, predictions for the defective regions were made through correlation of the data to the expected Gumbel distribution, selected for its proficiency in detecting extreme outliers. These predictions were validated experimentally through the use of dynamic bearing test rigs that ran until a defect propagated or until 120,000 simulated service miles were reached. These results are expected to have an impact on future inspection strategies for reconditioned railroad bearings.

Wheels are one of the most safety-relevant components in rail transport. To ensure operational safety, regular ultrasonic testing is essential. Challenges arise in both passenger and freight transport due to longer operating times, greater loads and higher speeds. Time for inspection has to be reduced to a minimum, while number and complexity of inspection areas increase.
Thus, maximum in flexibility concerning probe or specimen handling on the one hand but also steering of ultrasonic beam within specimen to optimize inspection area on the other hand is necessary.
In this context, coupling control plays a prominent role in ultrasonic testing. However, in many cases, precise and direct coupling control is not possible with conventional test systems due to the complex geometries and test requirements.

These challenges are solved by forward-looking technologies including robotics and next level sensor design. It is further supported by operator workflow guidance and intelligent decision support.
The lecture will deal with this combination and its implementation in field. Beneficial integration of test specimen for validation of inspection system as well as results of 2D-matrix-arrays are shown.

Keywords: Automated Testing Systems, Ultrasonic Testing, Wheel Testing, Reliability

The work was aiming at the problem that the fatigue crack of train frame is detected by magnetic powder paint removal in most cases, the eddy current detection method is innovatively introduced to realize the fatigue crack detection at the weld position of the frame surface. Develop a mobile intelligent flaw detection robot platform equipped with SLAM laser navigation and binocular camera hand-eye calibration technology to achieve sub-millimeter positioning accuracy of weld landing point. The finite element simulation model is established, and the orthogonal coil structure that is not affected by crack direction is designed to verify the reliability of eddy current detection method for fatigue crack in frame weld. A robot system is developed to automatically detect the flaw of the new frame and the scrapped frame of the actual vehicle. The results show that when the scanning speed of the system is 150 mm/s, the detection time of the whole vehicle frame is nearly about 1 h, and the detection and location of fatigue crack defects with a depth of not less than 0.5 mm can be realized. The false alarm rate of the system algorithm can be stable at 2.5%. The detection efficiency of the robot system exceeds that of the magnetic particle method, which provides relevant technical reference for improving the batch detection efficiency of the vehicle frame assembly line.