Systems and methods for automatic tuning of classification yard parameters

Abstract

Methods and systems for automatically tuning control parameters for operations of a classification yard. In embodiments, production predictions for car events at a segment is made using current tuning coefficients. Analysis on real-world measurements associated with the car events is used to obtain a set of candidate tuning coefficients. Backoffice predictions for the car events are made using the candidate tuning coefficients. The production predictions and the backoffice predictions are compared against the real-world measurements. If the backoffice predictions are found to better approximate the real-world measurements at the segment or device, the candidate tuning coefficients are accepted and the current tuning coefficients for the segment or device are replaced by the candidate tuning coefficients. In this manner, the present disclosure provides a system with functionality that allows the system to automatically adjust the tuning coefficients to real-world conditions.

What is claimed is: 
     
         1 . A method of automatically tuning control parameters for operations of a classification yard, comprising:
 generating a set of production predictions associated with one or more car events at a first point of a route within the classification yard using a production set of control parameters associated with the first point of the route, wherein the production set of control parameters includes one or more parameters associated with the rollability of one or more railcar cuts through the first point of the route;   obtaining actual measurements associated with the one or more car events at the first point of the route;   estimating a candidate set of control parameters associated with the first point of the route based on the actual measurements associated with the one or more car events at the first point of the route;   generating a set of backoffice predictions associated with the one or more car events at the first point of the route using the candidate set of control parameters associated with the first point of the route;   comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route to determine which of the production set of control parameters or the candidate set of control parameters for the first point of the route yields more accurate predictions for car events at the first point of the route; and   determining to replace the production set of control parameters for the first point of the route with the candidate set of control parameters in response to a determination that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters.   
     
     
         2 . The method of  claim 1 , wherein the first point of the route includes one or more of a route segment and a device of the classification yard. 
     
     
         3 . The method of  claim 2 , wherein the device of the classification yard includes one or more of:
 a switch;   a retarder; and   a wheel detector.   
     
     
         4 . The method of  claim 1 , wherein the one or more car events include one or more of:
 a railroad cut traveling through the first point of the route at a first speed;   the railroad cut arriving at the first point of the route at a first time;   the railroad cut entering at an entry point of the first point of the route at an entry speed; and   the railroad cut exiting at an exit point from the first point of the route at an exit speed.   
     
     
         5 . The method of  claim 1 , wherein the production set of control parameters for the first point of the route includes one or more of:
 rolling resistance coefficients;   temperature coefficients;   regression coefficients;   switch coefficients;   retarder coefficients;   detector coefficients; and   angle coefficients.   
     
     
         6 . The method of  claim 1 , wherein estimating the candidate set of control parameters associated with the first point of the route based on the actual measurements associated with the one or more car events at the first point of the route includes applying a regression algorithm to the actual measurements associated with the one or more car events at the first point of the route to obtain the candidate set of control parameters associated with the first point of the route. 
     
     
         7 . The method of  claim 1 , wherein one or more of the set of production predictions and the set of backoffice predictions include predictions of one or more of:
 energy of a railroad cut at the first point of the route;   speed of the railroad cut at the first point of the route; and   arrival time of the railroad cut at the first point of the route.   
     
     
         8 . The method of  claim 1 , wherein comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route includes applying a statistical comparison between the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route. 
     
     
         9 . The method of  claim 1 , wherein comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route includes:
 calculating a production absolute value average difference between the set of production predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   calculating a backoffice absolute value average difference between the set of backoffice predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   comparing the production absolute value average difference and the backoffice absolute value average difference to determine which one of the production absolute value average difference and the backoffice absolute value average difference is smaller;   determining that the production set of control parameters yields more accurate predictions for car events at the first point of the route than the candidate set of control parameters in response to a determination that the production absolute value average difference is smaller than the backoffice absolute value average difference for the first point of the route; and   determining that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters in response to a determination that the production absolute value average difference is not smaller than the backoffice absolute value average difference for the first point of the route.   
     
     
         10 . The method of  claim 1 , wherein the one or more car events at the first point of the route are classified into a bucket classification, the bucket classification including one or more of:
 a wet classification to classify car events occurring during wet weather conditions;   a dry classification to classify car events occurring during dry weather conditions;   a cold classification to classify car events occurring during cold weather conditions;   a warm classification to classify car events occurring during warm weather conditions;   a hot classification to classify car events occurring during hot weather conditions; and   a resilience bearing type classification to classify car events associated with a railroad cut including one or more train cars having a resilience type bearing.   
     
     
         11 . A system for automatically tuning control parameters for operations of a classification yard, comprising:
 at least one processor; and   a memory operably coupled to the at least one processor and storing processor-readable code that, when executed by the at least one processor, is configured to perform operations including:
 generating a set of production predictions associated with one or more car events at a first point of a route within the classification yard using a production set of control parameters associated with the first point of the route, wherein the production set of control parameters includes one or more parameters associated with the rollability of one or more railcar cuts through the first point of the route; 
 obtaining actual measurements associated with the one or more car events at the first point of the route; 
 estimating a candidate set of control parameters associated with the first point of the route based on the actual measurements associated with the one or more car events at the first point of the route; 
 generating a set of backoffice predictions associated with the one or more car events at the first point of the route using the candidate set of control parameters associated with the first point of the route; 
 comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route to determine which of the production set of control parameters or the candidate set of control parameters for the first point of the route yields more accurate predictions for car events at the first point of the route; and 
 determining to replace the production set of control parameters for the first point of the route with the candidate set of control parameters in response to a determination that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters. 
   
     
     
         12 . The system of  claim 11 , wherein the first point of the route includes one or more of a route segment and a device of the classification yard. 
     
     
         13 . The system of  claim 11 , wherein the one or more car events include one or more of:
 a railroad cut traveling through the first point of the route at a first speed;   the railroad cut arriving at the first point of the route at a first time;   the railroad cut entering at an entry point of the first point of the route at an entry speed; and   the railroad cut exiting at an exit point from the first point of the route at an exit speed.   
     
     
         14 . The system of  claim 11 , wherein the production set of control parameters for the first point of the route includes one or more of:
 rolling resistance coefficients;   temperature coefficients;   regression coefficients;   switch coefficients;   retarder coefficients;   detector coefficients; and   angle coefficients.   
     
     
         15 . The system of  claim 11 , wherein estimating the candidate set of control parameters associated with the first point of the route based on the actual measurements associated with the one or more car events at the first point of the route includes applying a regression algorithm to the actual measurements associated with the one or more car events at the first point of the route to obtain the candidate set of control parameters associated with the first point of the route. 
     
     
         16 . The system of  claim 11 , wherein one or more of the set of production predictions and the set of backoffice predictions include predictions of one or more of:
 energy of a railroad cut at the first point of the route;   speed of the railroad cut at the first point of the route; and   arrival time of the railroad cut at the first point of the route.   
     
     
         17 . The system of  claim 11 , wherein comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route includes:
 calculating a production absolute value average difference between the set of production predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   calculating a backoffice absolute value average difference between the set of backoffice predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   comparing the production absolute value average difference and the backoffice absolute value average difference to determine which one of the production absolute value average difference and the backoffice absolute value average difference is smaller;   determining that the production set of control parameters yields more accurate predictions for car events at the first point of the route than the candidate set of control parameters in response to a determination that the production absolute value average difference is smaller than the backoffice absolute value average difference for the first point of the route; and   determining that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters in response to a determination that the production absolute value average difference is not smaller than the backoffice absolute value average difference for the first point of the route.   
     
     
         18 . The system of  claim 11 , wherein the one or more car events at the first point of the route are classified into a bucket classification, the bucket classification including one or more of:
 a wet classification to classify car events occurring during wet weather conditions;   a dry classification to classify car events occurring during dry weather conditions;   a cold classification to classify car events occurring during cold weather conditions;   a warm classification to classify car events occurring during warm weather conditions;   a hot classification to classify car events occurring during hot weather conditions; and   a resilience bearing type classification to classify car events associated with a railroad cut including one or more train cars having a resilience type bearing.   
     
     
         19 . A computer-based tool for automatically tuning control parameters for operations of a classification yard, the computer-based tool including non-transitory computer readable media having stored thereon computer code which, when executed by a processor, causes a computing device to perform operations comprising:
 generating a set of production predictions associated with one or more car events at a first point of a route within the classification yard using a production set of control parameters associated with the first point of the route, wherein the production set of control parameters includes one or more parameters associated with the rollability of one or more railcar cuts through the first point of the route;   obtaining actual measurements associated with the one or more car events at the first point of the route;   estimating a candidate set of control parameters associated with the first point of the route based on the actual measurements associated with the one or more car events at the first point of the route;   generating a set of backoffice predictions associated with the one or more car events at the first point of the route using the candidate set of control parameters associated with the first point of the route;   comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route to determine which of the production set of control parameters or the candidate set of control parameters for the first point of the route yields more accurate predictions for car events at the first point of the route; and   determining to replace the production set of control parameters for the first point of the route with the candidate set of control parameters in response to a determination that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters.   
     
     
         20 . The computer-based tool of  claim 19 , wherein comparing the set of production predictions associated with the one or more car events at the first point of the route and the set of backoffice predictions associated with the one or more car events at the first point of the route includes:
 calculating a production absolute value average difference between the set of production predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   calculating a backoffice absolute value average difference between the set of backoffice predictions associated with the one or more car events at the first point of the route and the actual measurements associated with the one or more car events at the first point of the route;   comparing the production absolute value average difference and the backoffice absolute value average difference to determine which one of the production absolute value average difference and the backoffice absolute value average difference is smaller;   determining that the production set of control parameters yields more accurate predictions for car events at the first point of the route than the candidate set of control parameters in response to a determination that the production absolute value average difference is smaller than the backoffice absolute value average difference for the first point of the route; and   determining that the candidate set of control parameters yields more accurate predictions for car events at the first point of the route than the production set of control parameters in response to a determination that the production absolute value average difference is not smaller than the backoffice absolute value average difference for the first point of the route.