System and method for the assessment of condition in complex operational systems based on multi-level pattern recognition

A method comprising:

• receiving a plurality of time series signals generated by a plurality of sensors of one or more machines;
• for each signal of the plurality of time series signals:
• identifying a base range transformation window for the signal through an iterative process of range reduction, each iteration corresponding to a level and a level range;
• for each level:
• transforming the signal into a plurality of digests using the corresponding level range;
• and
• inputting the digests for each level as pseudo time-series signals to a transformer to generate a transformation block for the level;
• composing the transformation blocks for the levels in a range digestion block;
• generating a resolver block from the range digestion block and known condition episodes;
• composing a condition recognition block from the range digestion block and the resolver block;
• and
• applying a combination of unsupervised and supervised learning to transform the condition recognition block into a feedback signal applied to the or more machines.