A weak signal detection method combining Van der Pol-Duffing chaotic oscillator and convolutional neural network(CNN)is proposed to address the problem that current weak signal detection methods for chaotic oscillator heavily rely on the uncertainty of phase diagram detection.Taking the defect detection of composite materials using coplanar capacitive sensors as the research background

the Van der Pol-Duffing chaotic system is used for weak signal detection.Combined with CNN

the phase diagrams generated by the chaotic system are detected and classified to solve the problem of high sensitivity to small defects in noisy backgrounds.Simulation verification results show that the confusion matrix of CNN classification results has good performance

with accuracy

precision

and recall rates reaching 95.84%

95.35%

and 95.34%

respectively.Comparative experiments are carried out with digital lock-in amplifiers and Duffing chaos detection methods

and the results show that the Van der Pol-Duffing chaotic oscillator method has higher precision and signal sensitivity in weak signals detection in materials.