The paper presents an application of signal processing and computational intelligence methods to detect presence of the wood boring insects larvae in the wooden constructions (such as the furniture of buildings). Such insects are one of the main sources of the degradation in such objects, therefore they should be detected as quickly as possible, before inflicting serious damage. The presented work involved the acoustic monitoring for detecting the presence of the larvae inside pieces of wood. An accelerometer was used to record the sound, further analyzed by a computer algorithm extracting features important for artificial-intelligence (AI) based classification employed to detect the old house borer’s (Hylotrupes bajulus L.) activity. The presented task is difficult, as the sounds made by the larvae are of relatively low amplitude and the background noise caused by people, electrical appliances or other sources may significantly degrade the accuracy of detection. The classification of sounds is needed to separate sources of noise which deteriorate the proper larva detection and should be suppressed if possible. The employed classification was based on features defined in the time domain followed by the support vector machine used as the binary classifier. The results allowed us to assess the effectiveness of the old house borer’s detection by the acoustic analysis enhanced with the AI algorithm.

This paper presents an automated method for analog system diagnostics, which aims to detect and localize multiple faults in noisy conditions. The generic architecture of the diagnostic scheme and its stages of denoising, stamp extraction, and fault detection are explained. The method is tested on three systems of various physical nature. Then, approaches to automated diagnostics of the different classes of the systems are proposed. Machine learning methods (decision-tree-based fuzzy logic) are used to effectively detect faults. Their advantages are explained and confirmed by examples.

Series arc faults cause the majority of household fires involving electrical failures or malfunctions. Low-fault current amplitude is the reason for the difficulties faced in implementing effective arc detection systems. The paper presents a novel arc detection and faulty line identification method. It can be easily used in the low-voltage Alternate Current (AC) household network for arc detection in the Non-Intrusive Load Monitoring (NILM). Unlike existing methods, the proposed approach exploits both current and voltage signal time domain analysis. Experiments have been conducted with up to six devices operating simultaneously in the same circuit with an arc fault generator based on the IEC 62606:2013 standard. Sixteen time-domain features were used to maximize the arc-fault detection accuracy for particular appliances. Performance of the random forest classifier for arc fault detection was evaluated for 28 sets of features with five different sampling rates. For the single period analysis arc, detection accuracy was 98.38%, with F-score of 0.9870, while in terms of the IEC 62606:2013 standard, it was 99.07%, with F-score of 0.9925. Location of a series arc fault (line selection) was realized by identifying devices powered by the faulty line. The line selection was based on the Mean Values of Changes feature vector (MVC₅₀), calculated for absolute values of differences between adjacent current signal periods during the arc fault. The fault location accuracy was 93.20% for all cases and 98.20% for cases where the arc fault affected a single device.

The paper presents a novel method for non-intrusive appliances identification. It can be used for energy load disaggregation in a smart grid. The approach identifies changes in the state of the particular appliance by measuring and processing the common supply current signal. Analysis of the instantaneous changes in the aggregated current on the output of the analyzed circuit in the power network is exploited here. The signal is processed using the time alignment of the current and voltage signals samples represented in the array form. The scheme includes filtering, event detection and identification, which is performed by comparing parameters of the detected event against previously determined signatures of monitored appliances. The analysis is performed in the time domain; therefore (unlike other existing methods), the information contained in the original signal is not lost. The approach was tested in the laboratory designed specifically for this purpose. All tests have been conducted with up to 12 appliances operating at the same time in the single power supply circuit. The measurement setup was developed and used to record appliances’ switching on/off events. During tests, 2300 events for devices were recorded. Collected data were processed to identify particular devices with the accuracy of 98.8% and macro-averaged F-score measure of 0.9874. High identification accuracy was achieved despite the high number of devices operating in the background.

This paper presents an application of the rough sets algorithms to the analog systems diagnostics. Multiple methods of discretization and reduction of the data sets obtained from measurements are tested to find effective combinations for diagnostic purposes. Useful features of rough sets are identified. Versatility and scalability of the proposed method are verified, by application to multiple analog systems belonging to various technical domains with different complexities. Practical remarks about fault detection and location in the presented systems are included. This paper is supplemented by conclusions about the effectiveness of the method and its future prospects.