Sensor networks are becoming ubiquitous, enabling to improve decision-making and reducing human interaction by means of automatic or semi-automatic responses. However, due to deterioration or induced effects, sensors measures can be affected and produce anomalies that could alter decision-making. Most of the existing methods to identify sensors irregularities focus basically on detecting and discarding anomalous values, without looking for complementary information to understand generated anomalies. This paper presents an approach to obtain such complementary information by categorizing sensor anomalies, based on multidimensional quality assessment. It consists of two processing stages: an evaluation of data and information streams to estimate data quality imperfections and information quality dimensions; followed by the determination of agreement limits, compliant with normal states, to identify and categorize anomalies. The case study of discrete and analog sensors system installed in a simulator training platform of fuel tanks is presented, to illustrate an application of the proposed approach, considering 13 experimentally evaluated anomalies.