Industry 4.0, in addition to simple digitization itself, proposes the development of a complex innovation chain based on the combination of multiple technologies that inevitably forces companies to rethink all of their maintenance, business and process management methods. The use of artificial intelligence, in turn, through techniques of Artificial Neural Networks (ANN) enables the construction of unsupervised mathematical and statistical systems capable of managing, diagnosing and acting on maintenance and fault detection systems. It was proposed in this work a discussion about the identification of the fault criticality condition in each load category in an internal combustion engine through the use of a neuron map. Measurements were made under the following load conditions: no load (0 Kw), 0.5 Kw, 1.0 Kw and 1.5 Kw, through insertion in standard condition (without failure), and in conditions with failures of the following types: wear on the valve stem exhaust, clearance between the valve guide and the stem of the exhaust valve and radial wear of the compression segment ring. A neural fault map was generated for each load condition, using clustering data mining tools using the Wards method, artificial neural networks using the Kohonen method and Confusion Matrix prediction matrix. The results showed that, in the standard condition (without load) and with 0.5 kW of load, the highest level of criticality was related to the clearance between the valve guide of the exhaust valve stem. While in the load conditions of 1kW and 1.5kW, the level of criticality indicated radial wear of the compression segment ring.