The main concept of the paper is to place Reinforcement Learning (RL) into various fields of manufacturing. As a first attempt, RL for Statistical Process Control (SPC) in production is introduced in the paper; it is a promising approach owing to the adaptability and continuous application capability of reinforcement learning. The well-known Q-Table method was applied for get more stable, predictable and easy to overview results, therefore, quantization of the values of the time series to stripes was required. The formulated goal was to predict the time series value in a certain number of production steps ahead as manufacturing trend forecast. The recent values of the analysed time series were selected as states for the RL and the future probabilities of its values being in the formulated stripes were defined as RL actions. For action update, the Bellman equation was applied and the RL reward depends on how accurate the predicting is. Furthermore, two concepts were introduced, the Reusing Window (RW) and the Measurement Window (MW). The RW is a sliding window that determines how many times one measured value of the time series will be reused during the RL repeatedly, while the MW is defined for enabling the comparison of learnings with different RWs by sampling them with the same evaluation frequency.

The paper presents the measurement system for the diagnostics of the ratchet mechanisms based on the acoustic signal analysis. Such mechanisms are common in such devices, as drive elements in bicycles, safe locks or socket wrenches. Their exploitation may lead to wearing out teeth in gears, which is significant in the sport bicycles. To ensure safety of their user, the mechanism’s state has to be evaluated. Based on the symptoms extracted from the audio recordings the Artificial Intelligence-based classifier can be used to detect and locate faults related with the pawl degradation. Experiments were performed to verify the system’s efficiency during the analysis of the selected gears in the sport (BMX-type) bicycle. They included determining the ability to operate in the noisy conditions and identifying faults by the decision tree, both in the standalone and boosted version. The system’s accuracy above 90% proves its applicability to analyze the real-world objects.

In the production process of natural gas one of the major problems is the formation of hydrate crystals creating hydrate plugs in the pipeline. The hydrate plugs increase production losses, because the removal of the plugs is a high cost, time consuming procedure. One of the solutions used to prevent hydrate formation is the injection of modern compositions to the gas flow, helping to dehydrate the gas. Dehydratation obviously means that the size of hydrate crystals does not increase. The substances used in low concentrations, have to be locally injected at the gas well sites. Inhibitor dosing depends on the amount of gas hydrate present. In the article two Artificial Neural Network (ANN)-based predictive detection solutions are presented. In both cases the goal is to predict hydrate formation. Data used come from two solutions. In the first one measurements were performed by a self-developed and -produced equipment (in this case, differential pressure was used as input). In the second solution data are used from the measurement system of a motorised chemical-injector device (pressure, temperature, quantity and type of inhibitor were used as inputs). Both systems are presented in the article.

The following paper presents the methodology of RIAA equalizer condition analysis based on measurements of its amplitude and phase characteristics. The RIAA equalizer is used during the signal recording and is an integral part of modern turntables. It’s parameters determine the quality of the music being played. The task is to determine the critical values of electronic components (capacitors) based on the characteristics of signals observed at the circuit’s output. It is considered difficult due to the presence of noise, elements’ tolerances, and simultaneous drift of several system’s parameters. The presented methodology uses the Artificial Intelligence (AI) module that implements the task of parameter identification. The knowledge exploited by the AI-based module is extracted during machine learning, based on the dataset obtained during the simulations of the equalizer’s computer model. For the decisionmaking module, the standard tool for the regression tasks, i.e. RBF-type Artificial Neural Network (ANN) was used. The obtained results allow for considering the potentially high usefulness of the presented approach for the parameters identification in electronic circuits used in audio technology.

Anomaly of rotating machines are usually inferred from vibration measurements. However, it is not easy to determine the normal range for conventional crest factor or primary component analysis. In this paper, we try to use the Artificial Neural Network technique to make judgments based on the degree of deviation from the learned normal range. Specifically, we evaluated VAE which compresses the measured sensor data into the latent space of smaller number of dimensions with standard normal distributions. We propose an anomaly score which indicates the deviation from the center of the normal distribution using linear VAE calculation and dimensionality compensation. The proposed anomaly score shows good performance with several test data sets and measured real data sets.