CENTER FOR NON DESTRUCTIVE EVALUATION

News

CNDE Laravel

RESEARCH THEMES

UBIQUITOUS SENSING

Continuous Distributed Sensing using Fiber Optic Sensors:

The current state of the art for fiber optic sensing is the use of Fiber Bragg Grating and similar discrete sensing elements for localising the measurements. These techniques are now well known and applied in practical applications such as for structural health monitoring and process measurements. However, when faced with large volume of monitoring, the discrete approach becomes very limited in scope and increasing the number of discrete sensors becomes the limiting factor. Hence, continuous sensing will be the preferred solution where the sensing element is non-discrete and information anywhere along the length of the fiber can be obtained. Currently, continuous sensing is feasible using Rayleigh Scattering. This covers large volume, but with low resolution. Hence, it is necessary to explore new paradigms in fiber optic sensing techniques where large volume and higher resolution of measurement becomes feasible. In addition, the current sensors have an upper operational temperature range and increasing this operational temperature ranges will also be desirable.

Structured Materials for Imaging

Metamaterials-based Ultrasonic/Acoustic imaging & sensors

The CNDE group has been a leader in the field of employing meta-crystals and meta-materials for realizing extraordinary properties in imaging, vibration damping and mode filtering. This sub-theme will take a leap forward from these efforts, to achieve next generation sensing and device capabilities. In keeping with the goals of NDE 5.0, this sub-theme will explore technologies that can be integrated and embedded into structural members, such that self-sensing and self-warning capabilities can be in-built into them. Topological devices such as material-contrast and step-change lenses will be explored for integration into structural and machine elements, such that using passive excitation consisting of random external vibrations, local defect generation events can be flagged for immediate attention soon after a threshold indication. This process requires studying mechanical filtering mechanisms such that coherent information can be extracted from such random excitations. Once extracted, the signals will trigger an alarm through excitation of electromagnetic indicators for remote logging. The other broad topic that will be studied under this sub-theme, is that of metamaterial-based material sensors for online materials characterization. For example, integration of metamaterial layers into the walls of sensitive locations such as fuel or water tanks will be explored such that the level of toxic/ unwanted contaminants can be self-monitored, triggering an alarm beyond a threshold. Another example is the incorporation of metamaterial ridges into piping, wiring and cabling to achieve self-sensing and self-warning using passive random excitation, while embedded bandgap layers for blocking vibration and seismic disturbances is the other topic to be studied. These topics also contribute to the overall NDE 5.0 theme of ‘ubiquitous and distributed sensing and mitigation’

Edge Intelligence & Soft-Sensing

AI enabled super-fast computations

Currently, numerical computational methods such as Finite Element, Finite Difference, Finite Volume, Boundary Element, etc. are extensively used for calculations. However, the speed of computation and the resource requirements will necessitate the off-line mode of computation. In this proposed work, a disruptive approach to exploring the translation of these numerical calculations to the edge computing platforms such as GPUs and TPUs will be demonstrated for very sophisticated calculations. This is a relatively new paradigm with very limited literature available. The CNDE group has been developing this approach over the past 2 years and have achieved more than 7 orders of magnitude reduction in computational time and with very small computational resources for classic problems in stress, thermal, and flow analysis. In this work, the numerical models are used to train a sophisticated AI engine using a range of computational models. The trained AI engine is expected to then provide, near instant, calculations in edge computing platforms. Up until now, about 1000% extrapolation has been demonstrated on 2D and 3D Problem sets. Using this approach, edge problem solving will be explored to enhance the last mile intelligence at the sensor end and provide real-time problem solving capability to the sensors. This, to the best of our knowledge, has never been attempted.

Pervasive Inspection

AI enabled super-fast computations

Currently, numerical computational methods such as Finite Element, Finite Difference, Finite Volume, Boundary Element, etc. are extensively used for calculations. However, the speed of computation and the resource requirements will necessitate the off-line mode of computation. In this proposed work, a disruptive approach to exploring the translation of these numerical calculations to the edge computing platforms such as GPUs and TPUs will be demonstrated for very sophisticated calculations. This is a relatively new paradigm with very limited literature available. The CNDE group has been developing this approach over the past 2 years and have achieved more than 7 orders of magnitude reduction in computational time and with very small computational resources for classic problems in stress, thermal, and flow analysis. In this work, the numerical models are used to train a sophisticated AI engine using a range of computational models. The trained AI engine is expected to then provide, near instant, calculations in edge computing platforms. Up until now, about 1000% extrapolation has been demonstrated on 2D and 3D Problem sets. Using this approach, edge problem solving will be explored to enhance the last mile intelligence at the sensor end and provide real-time problem solving capability to the sensors. This, to the best of our knowledge, has never been attempted.

Remote Large Area Inspection

AI enabled super-fast computations

Currently, numerical computational methods such as Finite Element, Finite Difference, Finite Volume, Boundary Element, etc. are extensively used for calculations. However, the speed of computation and the resource requirements will necessitate the off-line mode of computation. In this proposed work, a disruptive approach to exploring the translation of these numerical calculations to the edge computing platforms such as GPUs and TPUs will be demonstrated for very sophisticated calculations. This is a relatively new paradigm with very limited literature available. The CNDE group has been developing this approach over the past 2 years and have achieved more than 7 orders of magnitude reduction in computational time and with very small computational resources for classic problems in stress, thermal, and flow analysis. In this work, the numerical models are used to train a sophisticated AI engine using a range of computational models. The trained AI engine is expected to then provide, near instant, calculations in edge computing platforms. Up until now, about 1000% extrapolation has been demonstrated on 2D and 3D Problem sets. Using this approach, edge problem solving will be explored to enhance the last mile intelligence at the sensor end and provide real-time problem solving capability to the sensors. This, to the best of our knowledge, has never been attempted.