Printed Electronics for Structural Health Monitoring: Automating the Identification of Cohesive Damage Parameters

Structurally integrated sensors are the basis of many structural health monitoring (SHM) systems [1]. Although integrated sensors allow a more compact design, they change the mechanical properties of the original load-bearing structure. Catastrophic failure due to unfavorable diminution of the structure’s mechanical strength and malfunction of sensors due to mechanical overload have to be avoided. Numerical tools, such as the finite element method (FEM), can help to anticipate damage initiation [2]. These simulations can be used not only to support the design of multifunctional structures but also to facilitate the inverse identification of damage relevant material parameters. [3] To describe the mechanical properties of a particular type of SHM sensors, i.e. metalbased printed electronics on composite materials, a cohesive zone model (CZM) is employed [4]. Based on experimentally obtained reference data, ABAQUS was used to determine the full set of cohesive parameters in a semi-automated way. Semi-automated and empirical parametrization approaches strongly depend on the user’s ability to identify the correct parameters. As a result, automation of these approaches may be limited, inefficient or incapable of dealing with changing experimental and thus numerical setups. To overcome this problem, fully automated algorithms which do not replicate the user behavior are desired. Based on existing experimental data, a procedure for the inverse identification of CZM parameters is implemented. A custom PYTHON script triggers ABAQUS calculations with varying CZM parameter configurations. Quality criteria are introduced to compare the numerical results with the experimental data. Using a response surface methodology in MATLAB, optimal parameters are found that produce the best match between simulation and experiment. Descriptions are given how the employed optimization procedures cope or struggle with common issues such as numerical divergence, efficiency and quality of produced parameters. The procedures are then benchmarked against each other, considering speed and applicability.