An artificial intelligence algorithm greatly increases accuracy in diagnosing the health of complex mechanical systems. In blind tests using data coming from highly unpredictable and real-life situations, the algorithm consistently achieves over 90 percent accuracy, say the researchers.