Search

This paper presents an alternative approach to the task of control performance assessment. Various statistical measures based on Gaussian and non-Gaussian distribution functions are evaluated. The analysis starts with the review of control error histograms followed by their statistical analysis using probability distribution functions. Simulation results obtained for a control system with the generalized predictive controller algorithm are considered. The proposed approach using Cauchy and Lévy α-stable distributions shows robustness against disturbances and enables effective control loop quality evaluation. Tests of the predictive algorithm prove its ability to detect the impact of the main controller parameters, such as the model gain, the dynamics or the prediction horizon.

Freight forwarding and transportation are the backbone of the modern economy. There are thousands of transportation companies on the market whose sole purpose is to deliver ordered goods from pickup to delivery. Transportation can be carried out by two types of fleets. A company can have its own trucks, or it can use third-party companies. This transportation can be carried out in a variety of formulas, with full truckload being the most common for long routes. The shipper must be aware of the potential cost of such a service during the process of selecting a particular transport. The presented solution addresses this exact issue. There are many approaches, ranging from detailed cost calculators to machine learning solutions. The present study uses a dedicated hybrid algorithm that combines different techniques, spanning clustering algorithms, regression and kNN (k Nearest Neighbors) estimators. The resulting solution was tested on real shipping data covering multi-year contract data from several shipping companies operating in the European market. The obtained results proved so successful that they were implemented in a commercial solution used by freight forwarding companies on a daily basis. [ABSTRACT FROM AUTHOR]

Statistical approach to Control Performance Assessment (CPA) is of great practical importance. This is particularly visible in process industry, where there are many PID loops. They are often assessed with measures derived from the Gaussian probabilistic density function. Standard deviation, variance, skewness or kurtosis form the majority of applied indexes. The review of data originating from process industry shows, however, to the contrary, that these signals have rather non-Gaussian properties and are mostly characterized by fat-tailed distribution disable the ability. Investigations show that strong disturbances may significantly disable the capacity of proper assessment. Standard measures often fail in such cases. It is shown that non-Gaussian measures can help with this problem. Various disturbances are tested and compared. Results show that fat-tailed distributions are an interesting alternative. They are less sensitive to disturbance shadowing and still make possible loop dynamic assessment. [ABSTRACT FROM AUTHOR]