The computer engineer Susana Ferreiro has produced a thesis entitled ‘Contributions towards the diagnosis and prognosis of industrial problems by means of Supervised Classification Techniques’. This work has been developed at the IK4-TEKNIKER R+D centre under the supervision of the EHU/UPV lecturer Basilio Sierra-Araujo (Head of the Department of Robotics and Autonomous Systems of the Computing Faculty in Donostia-San Sebastian). The aim of the research in this thesis has been to apply artificial intelligence techniques, data mining and machine learning to problems linked to the aeronautics industry. “These are algorithms and classifying models that extract information from large volumes of data and infer knowledge on the basis of these data,” explains Ferreiro.
Specifically, three problems have been studied through these techniques:aircraft brake wear prognosis for predictive maintenance, the prediction of the appearance of burrs during the drilling process in the manufacture of components, and the prediction of the basicity number (BN) of oil on the basis of spectroscopic data.
Aircraft brake wear prognosis
The ultimate aim was to cut the costs of aircraft line maintenance, in other words, the maintenance carried out after landing between one flight and the next,by deferring it to a more convenient time and place. The study also sought to reduce waiting times between flights and ensure punctuality when eliminating the delays caused by current corrective maintenance.“ A series of components of the aircraft are usually checked between one flight and the next.Sometimes an unanticipated problem arises; so the aim is to have an estimate of the wear of certain components to anticipate all the resources that are going to be necessary,” says Ferreiro. “The aim is also to optimize airline routes because sometimes there is an interest in having the maintenance done in a specific country, and what is needed for this is the forward planning of the state of the aircraft.” This line of research came out of the European TATEM project.
Predicting the appearance of drilling burrs
This problem has to do with the manufacturing process. When the components are manufactured, a check needs to be done to make sure that the burr, the notch, that has come away during the drilling,does not exceed 127 microns as specified by the aeronautics industry. “We have developed a process using the internal signals of the machine which detects in real time when the limit has been exceeded,” explains Ferreiro.Normally, after drilling, a process is always applied to eliminate the remaining burr, but thanks to this study, the process would be applied only when the limit is exceeded. This part of the research was started in the ARKUNE project.
Prediction of the basicity number (BN) of the oil on the basis of spectroscopic data
This problem affects the measuring of the oil degradation level. “The basicity number (BN) is used to estimate what state it is in:whether it is satisfactory, whether it needs to be monitored because it has started to degrade, or whether it needs to be replaced,” says the author. The aim of the research was to obtain a model for detecting the BN state to be able to make an assessment on the degradation state of the oil without having to run a laboratory test. The obtaining of the BNusing laboratory equipment is an assessment involvingperchloric acid, a very expensive task not only in terms of equipment and material, but also in terms of personnel and time. The idea developed in this thesis is to replace this method of analysis by near infrared FTIR spectrometry. With this method, “it is possible to develop a sensor and incorporate it into the machine and in what is being monitored without having to run a lab test,” explains Ferreiro.
For more information visit www.ehu.es.
Filed Under: Aerospace + defense, AI • machine learning
