Low-cost, long-range, and highly reliable wireless connections are required to support industrial IoT (IIoT) applications, such as smart manufacturing facilities, automated logistics centres, and chemical enterprises with high process intensity. LoRaWAN has gained widespread adoption due to its energy efficiency and scalability. Nevertheless, LoRaWAN is subject to significant performance degradation in busy commercial areas due to severe co-channel interference, packet collisions, multipath fading, and erratic electromagnetic noise from industrial machinery. With thousands of enrolled gadgets, network congestion grows exponentially, resulting in package delivery rates falling below 70 per cent, long latency spikes beyond safe limits, and excess energy waste from frequent retransmissions. These performance bottlenecks degrade critical industrial processes, including safety monitoring, robotic coordination, and real-time quality control. To overcome them, we propose a new LoRaWAN protocol redesign, called IR-LoRaWAN (Interference-Resilient LoRaWAN), optimised for ultra-dense IIoT networks. IR-LoRaWAN proposes a three-level approach to interference management: Adaptive Frequency Hopping Spread Spectrum (AFHSS) to switch between congested and clear channels dynamically, Density-Aware Transmission Scheduling (DATS) to group nodes into clusters with optimal transmission slots, and a Predictive Interference Learning Engine (PILE) running at the gateways uses lightweight machine learning to predict the interference patterns in advance and optimally modify the network accordingly. Moreover, a QoS-aware priority layer ensures that mission-critical nodes, e.g., safety alarms or emergency controls, have priority access with low latency. Large-scale simulations in a 10,000-node smart factory would demonstrate that IR-LoRaWAN would deliver a 92% packet delivery ratio, 46% lower latency, and 27% lower energy consumption relative to plain LoRaWAN and more recent improvements such as LoRa-IC and HybMAC. IR-LoRaWAN can scale, reliably and efficiently, to an industrial IoT implementation because it does not have to work around interference; instead, it implements an interference-resistant industrial Internet.