System and method for coordination of acceleration values of locomotives in a train consist

Abstract

A train control system includes independent virtual in-train forces modelling engines onboard each of a plurality of locomotives in a train. Each of the plurality of locomotives may also include an analytics engine and a calibration engine configured to assimilate, analyze, and calibrate real time information from other locomotives and from draft gears and couplers interconnecting the locomotives with determinations made by the independent virtual in-train forces modelling engine onboard the respective locomotive, with the plurality of locomotives of the train being configured to operate collectively and coordinate their own acceleration values based on a common goal of minimizing in-train forces without being dependent on a command from a lead locomotive or central command.

What is claimed is: 
     
         1 . A train control system including a plurality of independent virtual in-train forces modelling engines onboard each of a plurality of locomotives in a train, the train control system comprising:
 a data acquisition hub communicatively connected to one or more of sensors and databases associated with one or more of the locomotives or other components of the train and configured to acquire real-time and historical operational and structural data from one or more of systems and components of the train;   an analytics engine configured to assimilate and analyze real time information from other locomotives and from components such as draft gears and couplers interconnecting the locomotives with determinations of acceleration values for each respective locomotive made by the independent virtual in-train forces modelling engine onboard the respective locomotive, with the plurality of locomotives of the train being configured to operate collectively and coordinate respective, acceleration values independently determined on each of the locomotives based on a common goal for all of the locomotives of minimizing in-train forces without being dependent on a command from a lead locomotive or central command;   each of the plurality of independent virtual in-train forces modelling engines being associated with a machine learning engine and being disposed onboard a respective one of the plurality of locomotives and configured to simulate in-train forces and train operational characteristics using physics-based equations, kinematic or dynamic modeling of behavior of the train or components of the train during operation when the train is accelerating or slowing down, and inputs derived from stored historical contextual data comprising one or more of a number of locomotives in the train, age or amount of usage of one or more locomotives of the train or other components of the train, weight distribution of the train, length of the train, speed of the train, control configurations for one or more locomotives or consists of the train, power notch settings of one or more locomotives of the train, braking implemented in the train, grade, temperature, or other characteristics of train tracks on which the train is operating, and engine operational parameters that affect performance of one or more locomotive engines for the train;   a virtual system model database configured to store one or more virtual system models simulated by the independent virtual in-train forces modelling engines, wherein each of the one or more virtual system models includes a mapping between different combinations of the stored historical contextual data and corresponding simulated in-train forces and train operational characteristics that occur when the train is accelerating or slowing down; and   an energy management system associated with each of the locomotives of the train and configured to adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more locomotives of the train based at least in part on the virtual system model.   
     
     
         2 . The train control system of  claim 1 , wherein
 the one or more sensors acquiring real-time data include one or more of LIDAR sensors, RADAR sensors, accelerometers, gyroscopic sensors, optical recognition sensors, and physical strain gauges configured to measure displacements or forces at the draft gears and couplers interconnecting the locomotives, and   the analytics engine is configured to integrate the data acquired by each of the sensors to produce more consistent and accurate force measurements for in-train forces than would be possible with any one of the sensors by itself.   
     
     
         3 . The train control system of  claim 2 , wherein the machine learning engine includes at least one of a neural network, a support vector machine, a Markov decision process engine, a decision tree based algorithm, or a Bayesian based estimator. 
     
     
         4 . The train control system of  claim 3 , wherein the machine learning engine includes a neural network, and the machine learning engine is configured to train the neural network by providing the inputs derived from stored historical contextual data as the input to a first layer of the neural network, wherein the output generated by a learning function of the machine learning engine includes a plurality of first outputs from the neural network generated based on the inputs, and the at least one learning parameter includes a characteristic of the neural network which is modified to reduce a difference between the plurality of first outputs and the training data.. 
     
     
         5 . The train control system of  claim 1 , wherein the virtual system modeling engine is configured to simulate the in-train forces and train operational characteristics during a period of time when the train includes at least one locomotive with an associated energy management system that is transitioning from a braking control to an acceleration control. 
     
     
         6 . The train control system of  claim 1 , wherein the real-time and historical operational and structural data acquired by the data acquisition hub for use as training data includes one or more of structural stresses on one or more couplers or draft gears interconnecting one or more of locomotives or locomotives and non-powered rail cars of the train. 
     
     
         7 . The train control system of  claim 6 , wherein the energy management system is configured to adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more associated locomotives of the train using a microprocessor-based locomotive control system, a cab electronics system, and an electronic pneumatic brake system mounted within a cab of each of the one or more locomotives. 
     
     
         8 . The train control system of  claim 7 , wherein the energy management system is configured to adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more associated locomotives of the train while transitioning and ramping up from heavy dynamic braking at the bottom of a hill to full throttle on the way back up an adjacent hill in a direction of travel of the train along the train tracks, and while increasing a ramp rate and maintaining the structural stresses on one or more of the couplers or draft gears. 
     
     
         9 . The train control system of  claim 1 , wherein the machine learning engine is configurable by a user in order to adjust the relative weights that are assigned to each of different types of the stored historical contextual data of each of the one or more virtual system models, and wherein one or more of the predetermined acceptable ranges of values for simulated in-train forces and train operational characteristics are configurable by the user. 
     
     
         10 . A method of using independent virtual in-train forces modelling engines onboard each of a plurality of locomotives in a train in order to coordinate the accelerations of each of the locomotives determined independently onboard each of the locomotives to minimize in-train forces without being dependent on a command from a lead locomotive or central command, the method comprising:
 assimilating, analyzing, and calibrating real time information from other locomotives and draft gears and couplers interconnecting the locomotives with determinations made by the independent virtual in-train forces modelling engine onboard the respective locomotive; and   operating the plurality of locomotives of the train collectively based on a common goal of minimizing in-train forces without being dependent on a command from a lead locomotive or central command;   using machine learning performed by a machine learning engine associated with each of the independent virtual in-train forces modelling engines for controlling the acceleration values for the locomotive on which it is mounted independently from any command received from offboard the locomotive;   acquiring real-time and historical operational and structural data from one or more of systems and components of the train at a data acquisition hub communicatively connected to one or more of sensors and databases associated with one or more locomotives or other components of the train, wherein the one or more sensors acquiring real-time data may include one or more of LIDAR sensors, RADAR sensors, accelerometers, gyroscopic sensors, optical recognition sensors, and physical strain gages configured to measure displacements or forces at the draft gears and couplers interconnecting the locomotives;   simulating in-train forces and train operational characteristics using each of the independent virtual in-train forces modelling engines mounted on a respective locomotive based on physics-based equations, kinematic or dynamic modeling of behavior of the train or components of the train during operation when the train is accelerating or slowing down, and inputs derived from stored historical contextual data comprising one or more of a number of locomotives in the train, position of the locomotive in the train, age or amount of usage of one or more locomotives of the train or other components of the train, weight distribution of the train, length of the train, speed of the train, control configurations for one or more locomotives or consists of the train, power notch settings of one or more locomotives of the train, braking implemented in the train, positive train control characteristics implemented in the train, grade, topology, temperature, or other characteristics of train tracks on which the train is operating, and engine operational parameters that affect performance of one or more locomotive engines for the train;   storing one or more virtual system models simulated by the independent virtual in-train forces modelling engine in a virtual system model database, with each of the one or more virtual system models including a mapping between different combinations of the stored historical contextual data and corresponding simulated in-train forces and train operational characteristics that occur when the train is accelerating or slowing down; and   adjusting one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for each of the locomotives using an energy management system onboard each of the locomotives of the train based at least in part on real time information from other locomotives and draft gears and couplers interconnecting the locomotives assimilated with determinations made by the independent virtual in-train forces modelling engine onboard the respective locomotive, with the plurality of locomotives of the train operating collectively based on a common goal of minimizing in-train forces.   
     
     
         11 . The method of  claim 10 , wherein the machine learning engine includes at least one of a neural network, a support vector machine, a Markov decision process engine, a decision tree based algorithm, or a Bayesian based estimator. 
     
     
         12 . The method of  claim 11 , wherein the machine learning engine includes a neural network, and the machine learning engine is configured to train the neural network by providing the inputs derived from stored historical contextual data as the input to a first layer of the neural network, wherein the output generated by a learning function of the machine learning engine includes a plurality of first outputs from the neural network generated based on the inputs, and the at least one learning parameter includes a characteristic of the neural network which is modified to reduce a difference between the plurality of first outputs and the training data. 
     
     
         13 . The method of  claim 10 , wherein the virtual system modeling engine is configured to simulate the in-train forces and train operational characteristics during a period of time when the train includes at least one locomotive with an associated energy management system that is transitioning from a braking control to an acceleration control. 
     
     
         14 . The method of  claim 13 , wherein the virtual system modeling engine is configured to simulate the in-train forces and train operational characteristics during a period of time when the train includes at least one locomotive with an associated energy management system that is transitioning and ramping up from heavy dynamic braking at the bottom of a hill to full throttle on the way back up an adjacent hill in a direction of travel of the train along the train tracks. 
     
     
         15 . The method of  claim 10 , wherein the real-time and historical operational and structural data acquired by the data acquisition hub for use as training data includes one or more of structural stresses on one or more draft gears or couplers interconnecting one or more of locomotives and non-powered rail cars of the train, and measured vibrations of engine components caused by harmonic nodes encountered while ramping up power output of one or more of the locomotive engines of the train. 
     
     
         16 . The method of  claim 15 , wherein the energy management system is configured to adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more associated locomotives of the train using a microprocessor-based locomotive control system, a cab electronics system, and an electronic pneumatic brake system mounted within a cab of each of the one or more locomotives. 
     
     
         17 . The method of  claim 16 , wherein the energy management system is configured to adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more associated locomotives of the train while transitioning and ramping up from heavy dynamic braking at the bottom of a hill to full throttle on the way back up an adjacent hill in a direction of travel of the train along the train tracks, and while increasing a ramp rate and maintaining the structural stresses on one or more knuckles and the vibrations of engine components within predetermined acceptable ranges of values. 
     
     
         18 . The method of  claim 10 , wherein the machine learning engine is configurable by a user in order to adjust the relative weights that are assigned to each of different types of the stored historical contextual data of each of the one or more virtual system models, and wherein the predetermined acceptable ranges of values for simulated in-train forces and train operational characteristics are configurable by the user. 
     
     
         19 . A locomotive control system, comprising:
 a learning system configured to:
 receive real-time and historical operational and structural data for use as training data from one or more systems or components of the train at a data acquisition hub communicatively connected to one or more of sensors and databases associated with one or more locomotives or other components of a train; 
 simulate, using a virtual system modeling engine, in-train forces and train operational characteristics using physics-based equations, kinematic or dynamic modeling of behavior of the train or components of the train during operation when the train is accelerating or slowing down, and inputs derived from stored historical contextual data comprising one or more of a number of locomotives in the train, age or amount of usage of one or more locomotives of the train or other components of the train, weight distribution of the train, length of the train, speed of the train, control configurations for one or more locomotives or consists of the train, power notch settings of one or more locomotives of the train, braking implemented in the train, positive train control characteristics implemented in the train, grade, temperature, or other characteristics of train tracks on which the train is operating, and engine operational parameters that affect performance of one or more locomotive engines for the train; 
 store one or more virtual system models simulated by the virtual system modeling engine in a virtual system model database, wherein each of the one or more virtual system models includes a mapping between different combinations of the stored historical contextual data and corresponding simulated in-train forces and train operational characteristics that occur when the train is accelerating or slowing down; 
 calculate, using a machine learning engine, relative weights to assign to each of different types of the stored historical contextual data of each of the one or more virtual system models and assigning the relative weights to the stored historical contextual data; 
 train the learning system with the machine learning engine using the weighted stored historical contextual data and the training data to determine a probability of each of the one or more virtual system models providing an accurate representation of actual in-train forces and train operational characteristics that occur during acceleration of the train after braking, and using a learning function including at least one learning parameter, wherein training the learning system includes:
 providing the weighted stored historical contextual data as an input to the learning function, the learning function being configured to use the at least one learning parameter to generate an output based on the input; 
 causing the learning function to generate the output based on the input; 
 comparing the output to the training data, wherein the training data includes data produced by sensors having captured actual information on in-train forces and train operational characteristics during acceleration of the train after braking; 
 comparing the determined probabilities of each of the virtual system models to a predetermined threshold probability level; and 
 initiating adjustments to one or more of the calculated relative weights assigned to each of the different types of the stored historical contextual data of each of the one or more virtual system models to improve the determined probabilities of each of the virtual system models based on actual information on in-train forces and train operational characteristics acquired from a plurality of different trains operating under different conditions; and 
 
 adjust one or more of throttle requests, dynamic braking requests, and pneumatic braking requests for the one or more locomotives of the train using an energy management system associated with the one or more locomotives of the train based at least in part on one of the virtual system models with the highest probability of providing an accurate representation of actual in-train forces and train operational characteristics and with one or more of the simulated in-train forces and train operational characteristics falling within a predetermined acceptable range of values. 
   
     
     
         20 . The locomotive control system of  claim 19 , wherein the training data includes configuration and operational data associated with the inputs derived from stored historical contextual data, the training data being generated by one or more systems or components of the train while the train is being operated by an experienced train operator, wherein the output generated by the learning function represents a goal or objective that the machine learning engine is configured to cause the learning system to match by modifying the at least one learning parameter until the difference between the output and the training data is less than a predetermined threshold difference.