In general we can say that scheduling track control is based on analysis of statistical data about their actual condition, research on experimental plots and development on the basis of their algorithm for determining the timing of diagnostics of rails in different parts of the railways (with different specifications).
The development of periodicity is currently performed manually by experts of the railway. It does not take into account a number of important factors that directly affect the rate of development of defects in the rails and, therefore, be considered when optimizing the intervals between checks the state of the rails and minimizing the cost of diagnosis. These data will add to the calculation of frequency, providing more accurate test period diagnosed plot ultrasonic flaw detectors.
Object of study is diagnosed part of the path, namely, its two components: the base metal rails and welded joints.
Rails [8] - a special section of steel beams, are laid on sleepers or other supports for education, as Typically, path that moves the rolling stock of railway transport, urban railways, a specialized structure in mines, quarries, cranes (Figure 4a).
Welded joint (Fig. 4b.) [9] - Compound of metal elements, carried out by welding. All that does not relate to the rail to the weld joint, is considered to be the main metal rail. That it produced the largest number of severe defects, so the base metal should be checked more frequently than the welded joint.
а)
b)
Figure 4 - а)Rails; b) Welded joint
There is also a notion of welded rail (velvet path) (Fig. 5.) [10]- A track with a completely welded joints. The advantages of this path are that it reduces wear and tear of rolling stock, exposed in the joints of the shock (dynamic) effects, prevents the breakdown of the permanent way, usually occurring in the joints, reducing the ground resistance of the train, reducing the costs of the road.
Figure 5 - Velvet path
We need to install the base and recommended frequency of testing diagnosed stretch of track on the basis of factors specified in the program.
Factors influencing the frequency of monitoring are given below:
information about the exit (withdrawal of the way) broken (to be replaced immediately on the day of detection) of rails for a certain statistical period (usually for 12 months in the past);
traffic density controlled the railway section on the results over the past year (measured in millions of tonnes transported by rail cargo area per kilometer per year);
Missed the total tonnage on the tracks (measured in millions of tonnes gross weight);
train speed at the site (the higher the speed, the higher the yield broken rails);
presence of surface defects and damage in operated the way;
The status of the path based on measurements of travel (actual assessment of a number of dynamic parameters of the path, such as track width (pattern), the level of yarn path (distortions and subsidence), the nonlinear acceleration due to changes in the stability of the path);
Type of rail (light or heavy).
Using the parameters of speed of trains, evaluation railmeter and availability of surface defects is impractical because their influence is minimal. Preferably use the reference of the facts. But the inclusion of these parameters in the development of diagnostic frequency of the base metal rails will give a more precise value, unlike the base frequency.
Can be used in ways and distances in the laboratory inspection, and further developed in the drafting of the optimal schedule of inspection sites and the way the required number of ultrasonic flaw and crack detector operators of trucks.
One type of structuring rules and facts is to provide information in the form of a decision tree.
Method of decision tree (decision trees) [3] is one of the most popular methods for solving the job classification and prediction. Sometimes this method of Data Mining is also called tree of decision rules, classification and regression trees. As can be seen from the last name, using this method solves the problem of classification and prediction. If the dependent, ie target variable takes a discrete value, then using the decision tree classification problems are solved. If the dependent variable takes continuous values, the decision tree determines the dependence of this variable on the independent variables, ie solves the problem of numerical prediction.
Thus, decision trees - this is a way of representing the rules in a hierarchical, coherent structure, where each object corresponds to a single site that provides solutions. Under the rule refers to a logical construct that is represented in the form of "if ... then ...". The first time, decision trees were proposed Hovilendlom and Hunt (Hoveland, Hunt) in the late 50's of last century. The first and most famous work (Hunt, EB), Merina (Marin J.) and Stone (Stone, PJ), which deals with the essence of the decision tree - "Experiments in Induction» («Experiments in Induction») - was published in 1966.
A decision tree consists of the following:
the root of the tree;
branches of a tree;
internal nodes (nodes check);
tree nodes (decision nodes).
As a result of the passage from the root to its top (leaves) solves the problem of classification, ie select one of the classes. In our case, we choose not a class, and the solution for a given rule (runs through every level of decision tree from root to tip).
