Real-time diagnostics of complex technical systems such as power plants are critical to keep the system in its working state. An ideal diagnostic system must detect any fault in advance and predict the future state of the technical system, so predictive algorithms are used in the diagnostics. This paper proposes a novel, computationally simple algorithm based on the Auto-Regressive Integrated Moving Average model to solve anomaly detection and forecasting problems. The good performance of the proposed algorithm was confirmed in numerous numerical experiments for both anomaly detection and forecasting problems. Moreover, a description of the Autoregressive Integrated Moving Average Fault Detection (ARIMAFD) library, which includes the proposed algorithms, is provided in this paper. The developed algorithm proves to be an efficient algorithm and can be applied to problems related to anomaly detection and technological parameter forecasting in real diagnostic systems.

И з в е с т и я в у з о в • Я д е р н а я э н е р г е т и к а • № 4 • 2 0 1 9 5

Offline changepoint detection (CPD) algorithms are used for signal segmentation in an optimal way. Generally, these algorithms are based on the assumption that signal’s changed statistical properties are known, and the appropriate models (metrics, cost functions) for changepoint detection are used. Otherwise, the process of proper model selection can become laborious and time-consuming with uncertain results. Although an ensemble approach is well known for increasing the robustness of the individual algorithms and dealing with mentioned challenges, it is weakly formalized and much less highlighted for CPD problems than for outlier detection or classification problems. This paper proposes an unsupervised CPD ensemble (CPDE) procedure with the pseudocode of the particular proposed ensemble algorithms and the link to their Python realization. The approach’s novelty is in aggregating several cost functions before the changepoint search procedure running during the offline analysis. The numerical experiment showed that the proposed CPDE outperforms non-ensemble CPD procedures. Additionally, we focused on analyzing common CPD algorithms, scaling, and aggregation functions, comparing them during the numerical experiment. The results were obtained on the two anomaly benchmarks that contain industrial faults and failures—Tennessee Eastman Process (TEP) and Skoltech Anomaly Benchmark (SKAB). One of the possible applications of our research is the estimation of the failure time for fault identification and isolation problems of the technical diagnostics.

No abstract

Introduction. Electronic devices capable of collecting individual telemetry data have opened up prospects for preclinical detection of COVID-19 signs. Known solutions involve the analysis of information that is difficult to obtain at the moment. We are talking, specifically, about the blood condition or a PCR test. This significantly limits the possibility of integrating algorithms with wrist gadgets. At the same time, the cardiovascular system as an object of observation is quite informative, the data collection is well developed. The article describes the problem of detecting covid anomalies in rhythm strips. The work aims at creating a mathematical model based on machine learning algorithms to automate the process of detecting covid abnormalities in the heart rhythm. The possibility of integrating the results obtained with fitness bracelets and smart watches is shown.Materials and Methods. The work involved an open technology stack: Python, Scikit-learn, Lightgbm. When assessing the quality of models for binary classification, metric F1 was used. 229 cardiac rhythm strips (сardiointervalographies) of patients with COVID-19 were studied. The presence or absence of signs of an anomaly was determined taking into account the time of the rhythm strip and the intervals between heartbeats. Deviations that could indicate infection were shown graphically. Based on the exploratory analysis results, a list of signs indicating an anomaly was made.Results. As a result of the work done, a mathematical model was obtained that detected heart rate abnormalities specific to COVID-19 with an accuracy of 83 %. The basic features determining the predictive ability of the model were identified and ranked. They included the current value of the interval between heartbeats, the derivatives at the subsequent and previous points of measuring the duration of the heartbeat, the first derivative at the current point, and the deviation of the current value of the duration of the RR-interval from the median. The first indicator in this list was recognized as the most significant, the last — the least. For machine learning purposes, the potential of five algorithms was evaluated: IsolationForest, LGBMClassifier, RandomForestClassifier, ExtraTreesClassifier, SGDOneClassSVM. The normal and abnormal results of observations in isolation trees were visualized. A parameter was set that corresponded to the probability of regular observation outside the norm, and its value was selected — 0.11. Taking into account this indicator, a graph was constructed for the SGDOneClassSVM model. Based on the data set, using the cross-validation technique, the quality metric was calculated. The case in hand was a rhythm strip with a time series of observations taken in one continuous time interval from one person. A step-by-step process of obtaining averaged metric values for each model was described. In comparison, the highest indicator was recorded for the LGBMClassifier model, the lowest — for SGDOneClassSVM and IsolationForest.Discussion and Conclusions. The resulting mathematical model takes up little space in the memory of a mobile device, i.e., it does not impose significant requirements on computing resources. The solution has an acceptable detection quality for preclinical screening of COVID-19-related cardiovascular disorders. The algorithm detects anomalies in 83 % of cases. Four minutes is enough to record a rhythm strip. The proposed scenario for using an integrated solution is concise and easy to implement. Widespread use of the development can contribute to the detection of COVID-19 at an early stage.