Most of the issues arise while managing infrastructure in urban places. To overcome these problems, we put forward an innovative AI-driven approach for predictive maintenance in the smart urban environments by having focus on the infrastructure management. Using a deep learning neural network called MLP (Multi-layer Perceptron) the data is obtained from various sensors in the infrastructure which includes, health, age, vibration, traffic load, and other environmental factors. While performing the feature importance in the predictive maintenance, the age and vibration are the important ones. The accuracy decreases at the slightest for the old-age infrastructure. With high precision and recall rates our model was able to achieve accuracy of 95% in the predictive maintenance. Efficient use of resources, cost-cutting, enhancements in safety measures, and finally, the infrastructure management in smart urban environments.