Machine Learning and IoT Integrated Piezoelectric Smart Textiles and Wearables for Point-of-Care Biomedical Applications

Abstract

In recent years, the convergence of wearable sensors with digital technologies has provided new avenues for healthcare solutions. However, the development of wearable electronics with better functionalities, flexibility, and comfort remains a significant challenge. This thesis addresses these challenges by integrating machine learning (ML) and Internet of Things (IoT) technologies into piezoelectric smart e-textiles and wearable devices, paving the way for innovative biomedical applications such as remote health monitoring, energy-efficient wearable sensors, and wound healing. In this context, a novel lead-free flexible PVDF composite is introduced, which is prepared by incorporating the Cs 3 Bi 2 I 9 perovskite into PVDF matrix. This composite achieves a remarkable electroactive phase of 92% yield in the PVDF matrix, resulting in a highly stable and flexible piezoelectric sensor for mechanical energy harvesting, bio-physiological signal detection and voice recognition. This study demonstrates the potential of lead-free perovskites in creating optical signal-modulated, piezo-responsive wearable sensors with enhanced environmental stability. In continuation with the above work, a wearable piezoelectric textile sensor is fabricated using Cs 3 Bi 2 I 9 -PVDF composite through continuous electrospinning process. This three-layered textile device features Cs 3 Bi 2 I 9 -PVDF nanofibers as the active layer and PEDOT-coated PVDF nanofibers as electrodes. The composite achieves full β-phase crystallization and generates an output voltage of 12 V, a current of 7 µA, and a power density of 3 µW⋅cm⁻². With outstanding breathability (b ~ 1.13 kg⋅m⁻²⋅d⁻¹) and water resistance (contact angle ~ 138°), the textile sensor efficiently harvests biomechanical energy while also monitoring human physiological motions, demonstrating its applicability in wearable healthcare and self-powered robotics. To enhance the performance of the smart textiles, an all-electrospun piezoelectric smart e- textile is developed and integrated with IoT and ML technologies, for advanced point-of-care diagnostics. The e-textile sensor consists of Cs 2 AgBiBr 6 -PVDF nanofibers as piezoelectric layer and polypyrrole-coated PVDF nanofibers as electrodes. The sensor achieves an open circuit voltage of 10.5 V, a short circuit current of 7.7 µA, and a power density of 4.2 µW⋅cm⁻². Exceptional breathability (b ~ 4.13 kg⋅m⁻²⋅d⁻¹), flexibility, and water resistance (contact angle ~ 137°) make this e-textile ideal for continuous monitoring of vital signs such as arterial pulse and respiration rate. With an impressive 96% accuracy in respiratory signal detection through viML, this e-textile showcases the immense potential for remote patient monitoring and early disease detection in wearable healthcare systems. The limitation of electrospinning in mass-production of smart textiles is overcome by introducing rotary jet spinning technique. This study demonstrates a rotary jet-spun textile piezoelectret, which improves the piezoelectric output by 150% (voltage) and 200% (current) through electrical poling. With an outstanding sensitivity of 400 mV/kPa, waterproofing (contact angle ~ 134°), and exceptional breathability (b ~ 10 kg⋅m⁻²⋅d⁻¹), this textile sensor is capable of deep learning-aided pressure mapping with 98% accuracy. Moreover, it accelerates the proliferation and migration of L929 cells, demonstrating its application in wound healing through piezoelectricity-induced electrical stimulation. This study highlights the scalability and versatility of the rotary jet-spun textile for biomedical technology. Furthermore, a machine learning-based approach is developed for the rapid characterization and quantification of electroactive phases in PVDF. Using a comprehensive spectral dataset, the machine learning model achieves high precision in identifying and quantifying α, β, and γ phases. This approach reduces the time and complexity involved in traditional techniques such as FTIR, allowing real-time monitoring of electroactive phase compositions in PVDF-based materials. Thereafter, to show the importance of this approach in biomedical technology, blood pressure monitoring system is developed utilizing a PVDF wearable sensor as a proof of concept. The integration of machine learning with polymer characterization enhances both material analysis and real-time sensor performance in healthcare applications. For customizing the design of smart textile sensor, additive manufacturing is a vital technique. In this work, a cylindrical textile piezoelectret sensor, is introduced first time using 3D printing technology. It is comprised of PLA fibers as the piezoelectret layer and PEDOT-coated PLA fibers as electrodes. The sensor efficiently converts large axial impact forces into uniform radial pressures, achieving an output voltage of 11.4 V and a current of 7.2 µA, marking 250% and 160% improvements respectively over traditional flat devices. Integrated with IoT and ML, the sensor is capable of intelligent pressure mapping, achieving 100% accuracy in special character identification, demonstrating its potential in sustainable energy harvesting and real-time sensor applications. In conclusion, this thesis presents a comprehensive framework for the integration of ML, IoT, and advanced piezoelectric materials in smart e-textiles and wearable sensors. These innovations not only enhance the functionality and comfort of wearable devices but also viiintroduce novel approaches for energy harvesting, healthcare monitoring, and tissue engineering, offering promising solutions for future biomedical applications.