We have developed a method to realize piezoelectric polymers into disposable and large-area flexible sensors. A machine learning algorithm that uses both the output from the piezoelectric PVDF and the amount of charge in the vibration energy harvester as features is shown to enable battery-less and highly accurate fault diagnosis. Furthermore, the matrix-array PVDF stress sensor was found to be highly accurate using a crosstalk elimination algorithm based on linear programming.