Machine learning generative models have opened up a new perspective for automated machine diagnostics. These methods improve decision-making by extracting features, classifying, and creating new observations using deep neural networks. Generative modeling aims to determine the joint distribution of input data. This contrasts traditional methods used in diagnostics based on discriminative models and the conditional probability distribution of the target variable at known feature values. In the variational autoencoder (VAE) algorithms trained by the authors, the parameters of diagnostic features are random variables, the distributions of which can be approximated based on data, and the identification of probability distributions is based on variational inference. Variational inference is a tool that deals with difficult statistical problems and is usually faster than classical methods. VAEs can detect anomalies, predict failures, and optimize processes. This paper proposes an unsupervised approach to fault diagnosis using only healthy data with automatic feature extraction from the continuous probabilistic latent subspace of the VAE encoder and reduction in PCA or t-SNE. The solution, verified in the example of simulation data, is a response to a common problem related to the lack or difficulty of obtaining marked data in defected states of devices and mechanical structures.

Diagnosing the condition of rotating machines by non-invasive methods is based on the analysis of dynamic signals from sensors mounted on the machine—such as vibration, velocity, or acceleration sensors; torque meters; force sensors; pressure sensors; etc. The article presents a new method combining the empirical mode decomposition algorithm with wavelet leader multifractal formalism applied to diagnosing damages of rotating machines in non-stationary conditions. The development of damage causes an increase in the level of multifractality of the signal. The multifractal spectrum obtained as a result of the algorithm changes its shape. Diagnosis is based on the classification of the features of this spectrum. The method is effective in relation to faults causing impulse responses in the dynamic signal registered by the sensors. The method has been illustrated with examples of vibration signals of rotating machines recorded on a laboratory stand, as well as on real objects.

The article presents a method for diagnosing cycloidal gear damage on a laboratory stand. The damage was simulated by removing the sliding sleeves from two adjacent external pins of the cycloidal gearbox. Damage to the sliding sleeves may occur under operating conditions and can lead to the destruction of the gear unit. Hence, early detection is essential. Signals from torque sensors, rotational speed sensors and vibration acceleration sensors of input and output shafts for various rotational speeds and transmission loads were recorded. The frequency analysis of these signals was carried out. Due to the fluctuation of the rotational speed, the frequency spectrum gives an approximate picture and is not useful in detecting this type of damage. The statistical characteristics of the signal were determined. However, only statistical moments of higher orders, such as kurtosis, are sensitive to the tested damage. Therefore, the use of multifractal analysis of the vibration signal using the wavelet leader method (WLMF) was considered. Then log-cumulants of the multifractal spectrum were selected as the new signal features.

A driving cycle is a time series of a vehicle’s speed, reflecting its movement in real road conditions. In addition to certification and comparative research, driving cycles are used in the virtual design of drive systems and embedded control algorithms, traffic management and intelligent road transport (traffic engineering). This study aimed to develop an adaptive driving cycle for a known route to optimize the energy consumption of an electric vehicle and improve the driving range. A novel distance-based adaptive driving cycle method was developed. The proposed algorithm uses the segmentation and iterative synthesis procedures of Markov chains. Energy consumption during driving is monitored on an ongoing basis using Gaussian process regression, and speed and acceleration are corrected adaptively to maintain the planned energy consumption. This paper presents the results of studies of simulated driving cycles and the performance of the algorithm when applied to the real recorded driving cycles of an electric vehicle.

The exponential development of technologies for the acquisition, collection, and processing of data from real-world objects is creating new perspectives in the field of machine maintenance. The Industrial Internet of Things is the source of a huge collection of measurement data. The performance of classification or regression algorithms needs to take into account the random nature of the process being modelled and any incomplete observability, especially in terms of failure states. The article highlights the practical possibilities of using generative artificial intelligence and deep machine learning systems to create synthetic measurement observations in monitoring the vibrations of rotating machinery to improve unbalanced databases. Variational Autoencoder VAE generative models with latent variables in the form of high-level input features of time-frequency spectra were studied. The mapping and generation algorithm was optimised and its effectiveness was tested in the practical solution of the task of diagnosing the three operating states of a demonstration gearbox.