We present a novel approach for real-time defect detection and classification in laser welding processes based on the use of uncooled PbSe image sensors working in the MWIR range. The spatial evolution of the melt pool was recorded and analyzed during several welding procedures. A machine learning approach was developed to classify welding defects. Principal components analysis (PCA) is used for dimensionality reduction of the melt pool data. This enhances classification results and enables on-line classification rates close to 1 kHz with non-optimized code prototyped in Python. These results point to the feasibility of real-time defect detection.
The combination of flexibility, productivity, precision and zero-defect manufacturing in future laser-based equipment are a major challenge that faces this enabling technology. New sensors for online monitoring and real-time control of laserbased processes are necessary for improving products quality and increasing manufacture yields. New approaches to fully automate processes towards zero-defect manufacturing demand smarter heads where lasers, optics, actuators, sensors and electronics will be integrated in a unique compact and affordable device.
Many defects arising in laser-based manufacturing processes come from instabilities in the dynamics of the laser process. Temperature and heat dynamics are key parameters to be monitored. Low cost infrared imagers with high-speed of response will constitute the next generation of sensors to be implemented in future monitoring and control systems for laser-based processes, capable to provide simultaneous information about heat dynamics and spatial distribution.
This work describes the result of using an innovative low-cost high-speed infrared imager based on the first quantum infrared imager monolithically integrated with Si-CMOS ROIC of the market. The sensor is able to provide low resolution images at frame rates up to 10 KHz in uncooled operation at the same cost as traditional infrared spot detectors. In order to demonstrate the capabilities of the new sensor technology, a low-cost camera was assembled on a standard production laser welding head, allowing to register melting pool images at frame rates of 10 kHz. In addition, a specific software was developed for defect detection and classification. Multiple laser welding processes were recorded with the aim to study the performance of the system and its application to the real-time monitoring of laser welding processes. During the experiments, different types of defects were produced and monitored. The classifier was fed with the experimental images obtained. Self-learning strategies were implemented with very promising results, demonstrating the feasibility of using low-cost high-speed infrared imagers in advancing towards a real-time / in-line zero-defect production systems.
Gas detectors are nowadays widely spread for safety purposes in industrial facilities. They are categorized by the type of gas they detect: combustible and/or toxic. Whereas electrochemical sensors have limited lifetime and maintenance issues, infrared sensors are reliable and free of maintenance devices used for detecting a wide variety of VOCs and inflammable gases such as hydrocarbon vapors. They usually work via a system of transmitters (light sources) which power is interfered when a gas is present in the optical path. A spectral analysis of this optical interference allows the gas detection and identification. Optical flame detectors are sensors intended to sight and respond to the presence of a flame, faster than a smoke detector or a heat detector would do. Many of these systems operate in the infrared band in order to detect the heat radiation, most of the times by comparison of three specific wavelength bands.
Most of the present infrared gas and optical flame detectors traditionally make use of MWIR single point sensors rather than imaging sensors; this is mainly due to the lack of affordable imaging sensing technologies in this band of the infrared spectrum. However, the appearance of uncooled imaging MWIR sensors made of VPD PbSe, with spectral detection range from 1 to 5 microns, opens the possibility to incorporate these sensors into gas and flame detection systems to allow area monitoring.
The paper presents some aspects of muzzle flash detection using low resolution polycrystalline PbSe 32×32 and 80×80 detectors FPA operating at room temperature (uncooled performance). These sensors, which detect in MWIR (3 - 5 microns region) and are manufactured using proprietary technology from New Infrared Technologies (VPD PbSe – Vapor Phase Deposition of polycrystalline PbSe), can be applied to muzzle flash detection. The system based in the uncooled 80×80 FPA monolithically integrated with the CMOS readout circuitry has allowed image recording with frame rates over 2000 Hz (true snapshot acquisition), whereas the lower density, uncooled 32×32 FPA is suitable for being used in low cost infrared imagers sensitive in the MWIR band with frame rates above 1000 Hz. The FPA detector, read-out electronics and processing electronics (allows the implementation of some algorithms for muzzle flash detection) of both systems are presented. The systems have been tested at field test ground. Results of detection range measurement with two types of optical systems (wide and narrow field of view) have been shown. The theoretical analysis of possibility detection of muzzle flash and initial results of testing of some algorithms for muzzle flash detection have been presented too.
The paper is focused on the application of uncooled MWIR imaging sensors for the monitoring of industrial welding
processes: resistance spot welding, resistance seam welding and laser welding. During the last 40 years, there has been
little advancement in sensor systems for inline quality control monitoring of the welding process. Most of the existing
systems are oriented for current, voltage and welding force monitoring. However, the temperatures reached during the
majority of the welding processes lead to infrared sensing as a powerful tool, and to the MWIR band in particular as the
most useful spectral band for monitoring this type of industrial processes. Infrared image information is a powerful tool
to study the energy distribution in the HAZ (Heat Affected Zone).
The work presents some experimental results obtained with uncooled MWIR imaging sensors, by monitoring several
welding processes. These results may be applied for real-time quality assurance of the process leading to better
throughputs in industrial manufacturing. The high-speed capability of the sensors used helped also to characterize the
dynamics of the welding process.
In this work a breakthrough in the field of low cost uncooled infrared detectors is presented: an 80x80 MWIR VPD PbSe detector monolithically integrated with the corresponding Si-CMOS circuitry. Fast speed of response and high frame rates are, until date, non existing performances in the domain of low cost uncooled IR imagers. The new detector presented fills the gap. The device is capable to provide MWIR images to rates as high as 2 KHz, full frame, in real uncooled operation which converts it in an excellent solution for being used in applications where short events and fast transients dominate the system dynamics to be studied or detected. VPD PbSe technology is unique because combines all the main requirements demanded for a volume ready technology: 1. Simple processing 2. Good reproducibility and homogeneity 3. Processing compatible with big areas substrates 4. Si-CMOS compatible (no hybridation needed) 5. Low cost optics and packagin The new FPA represents a milestone in the road towards affordable uncooled MWIR imagers and it is the demonstration of VPD PbSe technology has reached industrial maturity. The device presented in the work was processed on 8-inch Si wafers with excellent results in terms of manufacturing yield and repeatability. The technology opens the MWIR band to SWaP concept.
Dfgfdg Due to international environmental regulations, aircraft turbojet manufacturers are required to analyze the gases exhausted during engine operation (CO, CO2, NOx, particles, unburned hydrocarbons (aka UHC), among others).Standard procedures, which involve sampling the gases from the exhaust plume and the analysis of the emissions, are usually complex and expensive, making a real need for techniques that allow a more frequent and reliable emissions measurements, and a desire to move from the traditional gas sampling-based methods to real time and non-intrusive gas exhaust analysis, usually spectroscopic. It is expected that the development of more precise and faster optical methods will provide better solutions in terms of performance/cost ratio. In this work the analysis of high-speed infrared emission spectroscopy measurements of plume exhaust are presented. The data was collected during the test trials of commercial engines carried out at Turbojet Testing Center-INTA. The results demonstrate the reliability of the technique for studying and monitoring the dynamics of the exhausted CO2 by the observation of the infrared emission of hot gases. A compact (no moving parts), high-speed, uncooled MWIR spectrometer was used for the data collection. This device is capable to register more than 5000 spectra per second in the infrared band ranging between 3.0 and 4.6 microns. Each spectrum is comprised by 128 spectral subbands with aband width of 60 nm. The spectrometer operated in a passive stand-off mode and the results from the measurements provided information of both the dynamics and the concentration of the CO2 during engine operation.
MWIR spectroscopy shows a large potential in the current IR devices market, due to its multiple applications (gas
detection, chemical analysis, industrial monitoring, combustion and flame characterization, food packaging etc) and its
outstanding performance (good sensitivity, NDT method, velocity of response, among others), opening this technique to
very diverse fields of application, such as industrial monitoring and control, agriculture, medicine and environmental
However, even though a big interest on MWIR spectroscopy technique has been present in the last years, two major
barriers have held it back from its widespread use outside the laboratory: the complexity and delicateness of some
popular techniques such as Fourier-transform IR (FT-IR) spectrometers, and the lack of affordable specific key elements
such a MWIR light sources and low cost (real uncooled) detectors. Recent developments in electrooptical components
are helping to overcome these drawbacks. The need for simpler solutions for analytical measurements has prompted the
development of better and more affordable uncooled MWIR detectors, electronics and optics.
In this paper a new MWIR spectrometry device is presented. Based on linear arrays of different geometries (64, 128 and
256 elements), NIT has developed a MWIR Variable Filter Array Spectrometer (VFAS). This compact device, with no
moving parts, based on a rugged and affordable detector, is suitable to be used in applications which demand high
sensitivity, good spectral discrimination, reliability and compactness, and where an alternative to the traditional scanning
instrument is desired. Some measurements carried out for several industries will be also presented.
In 2007 the compatibility of VPD PbSe and Silicon CMOS technologies was demonstrated. At that time, the first
monolithic device, a laboratory demonstrator with 16x16 elements, was processed successfully. Since then the
technology has evolved towards its industrial maturity and a number of new devices based on this material have been
developed. Their performances have converted the VPD PbSe in one of the most promising technologies in the market
for fast and low cost IR imagers.
In this paper a brief historical review and the state of the art of the PbSe technology are presented, as well as the ultimate
performances commercially available,the latest experimental results obtained for three relevant applications, and the
future technology evolution to fulfill the requirements associated with more complexes and demanding applications.
Several sensors are used during the performance tests of a turbojet engine to record parameters such as temperature,
pressure, vibration, etc. However, most of these sensors have long time constants and are unable to measure fast
transients and fluctuations.
In this paper we show an alternative sensor to characterize some phenomena observing the flow resulting from the
combustion at the outlet of a turbojet engine, using a very high-speed uncooled MWIR COTS imaging sensor from New
Infrared Technologies, which provides over 1,600 fps.
The experiments include monitoring of flow stability during a long observation time (accelerations and stationary
regimes), and higher frequency events such as surges. Compressor surges produce extremely loud bangs from the engine,
and may be accompanied by a fast increment of the exhaust gas temperature and an increase in rotor speed due to the
reduction in work done by the stalled compressor, causing severe stresses within the engine from the intense
aerodynamic buffeting within the compressor.
The main conclusion of this study is the demonstration of the potential and worthiness of high-speed uncooled IR
imagers for detecting and analyzing fast transient events and stationary events during combustion test studies from
thrusters such as turbojets and rockets.