Smart water purification for domestic environments: AI-driven dual filtration with multilingual voice-assisted interaction

Abstract

Abstract Household water safety depends not only on contaminant removal efficiency but also on consistent user adherence and timely maintenance. However, usability barriers and lack of accessible guidance limit the real-world effectiveness of point-of-use purification systems, particularly in multilingual settings. This study presents Automated Quality Understanding and Voice Assistant i.e. AQUA-VA, a smart domestic water purifier integrating dual-stage RO + UV/UF filtration with edge artificial intelligence and multilingual voice-assisted interaction. The system employs low-power sensing of pH, total dissolved solids (TDS), turbidity, flow, and temperature, coupled with a machine-learning-based water quality index model and Bayesian filter health estimation for predictive maintenance. A mixed-integer programming scheduler optimizes energy usage and purification cycles, while an on-device speech interface enables real-time multilingual guidance. The system was evaluated over 12 weeks across 36 households under diverse water conditions. Results show a 71.3% reduction in unsafe water dispenses, a 54.8% decrease in delayed maintenance events, and a 26.4% improvement in first-attempt task completion compared to a baseline system without voice support. The speech module achieved 93.1% intent recognition accuracy with a median latency of 186 ms across multiple languages. These findings demonstrate that integrating edge AI with multilingual voice-assisted interaction significantly enhances water safety, usability, and maintenance adherence in household purification systems.