No abstract

Lactate threshold is considered an essential parameter when assessing performance of elite and recreational runners and prescribing training intensities in endurance sports. However, the measurement of blood lactate concentration requires expensive equipment and the extraction of blood samples, which are inconvenient for frequent monitoring. Furthermore, most recreational runners do not have access to routine assessment of their physical fitness by the aforementioned equipment so they are not able to calculate the lactate threshold without resorting to an expensive and specialized center. Therefore, the main objective of this study is to create an intelligent system capable of estimating the lactate threshold of recreational athletes participating in endurance running sports.The solution here proposed is based on a machine learning system which models the lactate evolution using recurrent neural networks and includes the proposal of standardization of the temporal axis as well as a modification of the stratified sampling method. The results show that the proposed system accurately estimates the lactate threshold of 89.52% of the athletes and its correlation with the experimentally measured lactate threshold is very high (R=0,89). Moreover, its behaviour with the test dataset is as good as with the training set, meaning that the generalization power of the model is high.Therefore, in this study a machine learning based system is proposed as alternative to the traditional invasive lactate threshold measurement tests for recreational runners.

Theoretical models of manufacturing processes provide a valuable insight into physical phenomena but their application to practical industrial situations is sometimes difficult. In the context of Industry 4.0, artificial intelligence techniques can provide efficient solutions to actual manufacturing problems when big data are available. Within the field of artificial intelligence, the use of deep learning is growing exponentially in solving many problems related to information and communication technologies (ICTs) but it still remains scarce or even rare in the field of manufacturing. In this work, deep learning is used to efficiently predict unexpected events in wire electrical discharge machining (WEDM), an advanced machining process largely used for aerospace components. The occurrence of an unexpected event, namely the change of thickness of the machined part, can be effectively predicted by recognizing hidden patterns from process signals. Based on WEDM experiments, different deep learning architectures were tested. By using a combination of a convolutional layer with gated recurrent units, thickness variation in the machined component could be predicted in 97.4% of cases, at least 2 mm in advance, which is extremely fast, acting before the process has degraded. New possibilities of deep learning for high-performance machine tools must be examined in the near future.

Grinding is an advanced machining process for the manufacturing of valuable complex and accurate parts for high added value sectors such as aerospace, wind generation, etc. Due to the extremely severe conditions inside grinding machines, critical process variables such as part surface finish or grinding wheel wear cannot be easily and cheaply measured on-line. In this paper a virtual sensor for on-line monitoring of those variables is presented. The sensor is based on the modelling ability of Artificial Neural Networks (ANNs) for stochastic and non-linear processes such as grinding; the selected architecture is the Layer-Recurrent neural network. The sensor makes use of the relation between the variables to be measured and power consumption in the wheel spindle, which can be easily measured. A sensor calibration methodology is presented, and the levels of error that can be expected are discussed. Validation of the new sensor is carried out by comparing the sensor's results with actual measurements carried out in an industrial grinding machine. Results show excellent estimation performance for both wheel wear and surface roughness. In the case of wheel wear, the absolute error is within the range of microns (average value 32 μm). In the case of surface finish, the absolute error is well below Ra 1 μm (average value 0.32 μm). The present approach can be easily generalized to other grinding operations.

Grinding is a critical machining process because it produces parts of high precision and high surface quality. Due to the semi-artisan production of the wheel, it is not possible to know in advance the performance of the wheel. One of the most useful parameters to characterize the grinding process is the specific grinding energy, which varies with the wear of the grinding wheel during its lifecycle. Thus, it would be useful to model the specific grinding energy in order to get information about the performance of the wheel before buying it. Unlike the typical applications of time series forecasting, in this work, a totally different issue is presented: the prediction of new and complete time series bounded in time without initial or historic values. In this context, an analysis of the effect of the time characteristics and the number of points of the time series on the prediction capabilities of the ANN is presented. The results of the analysis show that 200 points are enough to predict a complete time series up to 2000 mm 3 /mm of specific volume of material removed. Actually, it is shown that modelling the evolution of the grinding specific energy up to 2000 mm 3 /mm is possible. The net shows good capability to generalize to new grinding conditions, with errors below 23.65 %, and to new wheel characteristics, with errors below 20.01 %, which are satisfactory from the grinding process perspective.