Verification and Validation of manufactured structures requires certified inspection techniques for specific structures and flaw types. Like all activities on-orbit inspection is expensive and time consuming when it is even possible. We are proposing a method where physical inspection is augmented with simulated inspection data to reduce the time and cost associated with certifying an on-orbit inspection process. Augmenting physical inspection with simulation has two positive outcomes it reduces cost and time required for a full certification of inspection and it creates a larger dataset that can be used in a state tracking framework such as digital twin.

Currently POD studies are required for an inspection technique looking for a specific kind of flaw on a specific structure such as subsurface impact damage on a IM7-8552 laminated skin of launch vehicle that is being detected using thermography. The study is done by manufacturing many representative skins, impacting them with a range of different sized impactors with a range of energies and inspecting them all with the exact technique that will be used. This included hardware and software and even sometime the inspector is part of the certification. Multiple samples of the same size of damage are required. All of this data is combined to determine the probability of detecting each size of damage. Typically, the POD establishes the smallest size of indication where 95% of those indications will be detected 90% of the time or a 90% probability of detection with a confidence of 95%. These types of studies can take years and have significant costs. Changing the material system, the inspection process, the load cases, or the inspector can require additional data to be collected or a whole new study. This new approach of using simulated data to replace and augment the physical data has not previously been adopted because inspection has been a legacy highly manual process. One inspector teaches another inspector. If for example one person gets hit by a bus or wins the lottery large inspection systems could be crippled. They have agreed upon standards and inspectors have to pass certification tests where they look at standards and identify indications of flaws but the processes are still highly variable. Two different inspectors could look at the same inspection data and have two different results but over a large collection of data they must be statistically similar. The process is not ideal but at the same time there is a lot of hesitancy to change a process that has mostly worked in the past. However, it is becoming impossible to ignore that the field of NDE needs transformation. It must more away from a manual process to an automated one. Data can no longer be discarded after an inspection (lest a failure occur, and someone were to second guess the decisions made) it must be incorporated into a larger dataset for a structure and analyzed for trends. If we are successful and MAPOD becomes the normal course of operation we will be able to reduce the number of parts that need to be physically manufactured, damaged, and inspected by replacing them with simulated data that represents a wider range of damage sizes, locations and configurations. Modeling can create an exponentially larger dataset in a fraction of the time with no addition manufacturing costs. We will assess the success of our demonstration study by comparing the cost of simulation with cost of acquiring the physical data.