METHOD FOR DETERMINING THE LOCAL VEHICLE SPEED OF A VEHICLE
Patent Information
- Application Number
- DE502020009636
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-08-08
- Filing Date
- 2020-07-28
- Publication Date
- 2024-11-21
- Estimated Expiration
- 2040-07-28
AI Technical Summary
Current methods for determining the speed of rail vehicles face significant measurement errors due to noise in satellite navigation, wheel diameter inaccuracies, wheel slip during braking or acceleration, and drift in acceleration sensors, which affect the accuracy of position, speed, and acceleration measurements.
A method using an observer model, preferably a Luenberger observer or Kalman filter, that acquires and processes sensor data such as longitudinal acceleration, wheel speed, and braking/drive forces to continuously determine the carriage speed, adapting to specific vehicle conditions and compensating for measurement errors through sensor fusion and drift correction.
This approach enables precise and continuous determination of carriage speed, improving accuracy for anti-skid protection, adhesion detection, and tilt detection, regardless of driving conditions, by using a consistent method and accounting for individual measurement inaccuracies and drift.
Description
[0001] The invention relates to a method for determining a car speed of a car and of a train with at least one car, a device designed to carry out the method, a car with this device and a computer program product.
[0002] Precise knowledge of the position, speed, deceleration, or acceleration of rail vehicles is the basis for a wide range of opportunities to improve operational processes in rail vehicle traffic.
[0003] However, measured quantities for determining position, speed, deceleration or acceleration, which can be measured directly with reasonable effort, are in many cases subject to large measurement errors.
[0004] When determining a position using satellites, namely a global navigation satellite system (GNSS), such as GPS, Galileo, or GLONASS, the true position is superimposed with a zero-mean measurement noise, the standard deviation of which can be estimated from the number of satellites received. However, in poor reception conditions, the accuracy can be drastically reduced, or reception may fail completely, for example, in a tunnel.
[0005] Position determination using train control systems is generally only possible at specific points or in sections and is therefore not suitable for continuous evaluation.
[0006] A speed signal from the so-called wheel slide protection, which is either provided by a reference wheelset or determined over several or all wheelsets, for example by averaging, has the problem that a systematic measurement inaccuracy can be caused by an unknown wheel diameter, which changes slowly over time.
[0007] Further measurement inaccuracies can arise from wheelset slip caused by a power transmission during braking or acceleration. While a non-driven and unbraked reference wheelset can minimize measurement errors caused by slip, this requires additional components and assembly work, and an unbraked wheelset is undesirable.
[0008] While it is possible to measure longitudinal acceleration cost-effectively from measurement signals from pulse generators, for example, through temporal differentiation, this is also inaccurate due to the sampling rate. Furthermore, the measured longitudinal acceleration must be compensated for a component of gravity. If acceleration sensors are used to measure longitudinal acceleration, this error does not need to be compensated; however, they can exhibit drift, which experience has shown to significantly distort or even render unusable speeds and distances determined by integration.
[0009] A braking or acceleration force can be approximately determined from the state variables of a drive or brake. However, inaccuracies in the determination of these forces can arise, for example, due to a lack of knowledge of the exact coefficient of friction between the brake disc and brake pad and the precise point of force application, or due to non-ideal force transmission.
[0010] Document US 9 517 780 B2 discloses a method for determining a vehicle speed. Sensor data available in the vehicle are recorded as a respective input variable for an observer model, the input variables are input into the observer model, and a continuous determination of the vehicle speed is performed by the observer model.
[0011] The invention is based on the object of providing a possibility to determine a speed of a wagon and of a rail vehicle consisting of at least one wagon as reliably as possible from the available information.
[0012] This object is achieved by the subject matter of the independent claims. Advantageous developments of the invention are contained in the dependent claims.
[0013] According to the invention, a method for determining a vehicle speed of a vehicle is provided, which comprises the steps: Acquiring sensor data available in the vehicle as a respective input variable for an observer model; inputting the input variable into the observer model; performing a continuous determination of the vehicle speed by the observer model.
[0014] "Detecting" here can be understood as both the measurement of corresponding quantities and their determination or calculation from other quantities. For example, a wheel speed can be detected by measuring it. On the other hand, the application force of a friction brake can be detected by calculating it from a brake pressure.
[0015] The observer model is preferably designed as a control engineering observer model, wherein in a preferred embodiment it is designed as a Luenberger observer or Kalman filter.
[0016] Preferably, the sensor data includes a longitudinal acceleration of the vehicle and / or a rotational speed of a vehicle element whose rotational speed is proportional to a wheel speed.
[0017] The carriage element is preferably a wheel, an axle of the carriage or a shaft of a motor of the carriage, wherein the proportionality factor is preferably a wheel radius, a gear ratio and / or another factor which is designed to determine a peripheral speed of the wheel from the detected rotational speed, which is preferably used as the wheel speed.
[0018] The speed is preferably measured using a speed sensor or pulse generator.
[0019] The acceleration is preferably measured using acceleration sensors, which are preferably designed in the form of a MEMS (microelectromechanical system), and preferably used in an inertial measuring system.
[0020] The acceleration can particularly preferably be measured from a high-precision aligned single-axis sensor, or the sensor system has additional sensors in orthogonal spatial directions and can be corrected using a calibration procedure.
[0021] Alternatively or additionally, the input variables include measurable forces or, preferably by means of an observer model, indirectly determined forces introduced by the braking or drive systems of the vehicle.
[0022] These forces can be determined, for example, by measuring the application forces of friction brakes or the drive torque of a motor. Preferably, these are forces that are transmitted to the rail via the corresponding wheel of the wagon whose wheel speed is being used. This provides physical quantities that influence the correlation between the wheel speed and the actual wagon speed.
[0023] Preferably, the observer model uses a speed and an acceleration of the car as state variables, wherein preferably the speed is the determined car speed and / or the acceleration is a longitudinal acceleration of the car.
[0024] According to a specific embodiment of the invention, a method for determining a car speed of a car of a rail vehicle includes the steps of: acquiring speed-relevant sensor data available locally in the car as a respective input variable for a control-engineering observer model; inputting the input variables into the control-engineering observer model; performing continuous processing of the input variables and determining state variables by the control-engineering observer model, wherein one of the state variables is preferably the car speed.
[0025] This method allows the vehicle speed to be accurately and continuously determined from the locally available sensor data. Due to the continuous determination of the local vehicle speed, this can be reliably used for anti-skid protection, adhesion detection, or inclination detection, regardless of the driving condition.
[0026] In this context, continuous determination means, in particular, that there is no need to switch between different calculation methods due to different driving conditions. For example, in switching methods, a mode is switched in which wheel speed or wheel speed is no longer used if the wheel is skidding or locking. Instead, in continuous determination, the same measured variables are always used to determine the vehicle speed. The method according to the invention is preferably designed to adapt the influence of the corresponding variables, such as acceleration or longitudinal acceleration or wheel speed, on the determined vehicle speed itself.
[0027] With the help of the observer model, which preferably functions as a simulation model of a system and into which measurement results from sensors can be input as output variables of the system as input variables of the observer model, state variables can be determined. The state variables can, for example, be more precise or corrected measurement results from the sensors. Furthermore, hidden state variables can also be determined that cannot be measured directly or can only be measured with great effort. The state variables are determined using equations, in particular differential equations, which define the behavior of the state variables in relation to the respective input variables, but also, if necessary, the behavior of the state variables in relation to one another. This way, by taking into account several relevant measurement results, precise state variables can be determined through sensor fusion, even when individual measurement results are inaccurate or erroneous.In addition, the observer model can estimate its own errors in determining the state variables in a specific way.
[0028] At least some of the sensor data is preferably acquired by sensors arranged on a chassis unit, in particular a bogie. If the sensor data, particularly speed-relevant data, is advantageously acquired by sensors arranged on a chassis unit, in particular a bogie, influences on the measurement results of the sensors can be locally taken into account in order to accurately determine the local speed as the state variable.
[0029] The sensor data preferably contains measurement values an acceleration sensor, an inclination sensor, an inertial measuring unit, a speed sensor, a brake pressure sensor, a brake current sensor, and / or a motor.
[0030] These sensor data are preferably entered into the observer model as input variables.
[0031] If the sensor data advantageously includes measurements from an acceleration sensor, the local vehicle speed can be determined by integrating the acceleration sensor measurements over time, whereby non-mean errors can be eliminated without filtering and thus without phase shift. Any inherent drift of the acceleration sensors can be compensated for by merging the sensors with other speed-relevant sensors, such as a wheel speed sensor. The drift can also be used as a state variable (see below) in the observer model.
[0032] Conditions in which the corresponding wheel circumferential speed is skidding or locking, i.e., in which the wheel speed no longer correlates with the actual vehicle speed, can be problematic when using the wheel circumferential speed. If such a condition exists, using a wheel speed determined from the wheel speed can introduce an error into the determination of the vehicle speed. Therefore, such conditions must be detected, and the observer model must be able to react accordingly.
[0033] According to the invention, a drift quantity is determined, whereby this determination is carried out by the observer model. The drift quantity is determined as a state variable of the observer model, whereby the drift quantity is the amount of a Difference between an actual longitudinal acceleration of the vehicle and the measured longitudinal acceleration, and / or difference between a longitudinal acceleration determined from the rotational speed, preferably from the wheel speed, and the measured longitudinal acceleration, and / or difference between the longitudinal acceleration determined from the rotational speed, preferably from the wheel speed, and the actual longitudinal acceleration is defined.
[0034] Preferably, a drift dynamic is determined by determining a temporal rate of change of the drift variable, wherein the amount of the drift dynamic is preferably compared with predetermined maximum values and, further preferably, if the predetermined maximum values are exceeded by the amount of the drift dynamic, it is concluded that there is an incorrect provision process of the rotational speed, in particular the wheel speed, and the predetermined maximum values preferably include changes in gradient and inclination over a route, installation and calibration errors of the acceleration sensors used and usual acceleration and deceleration dynamics of the train.
[0035] Based on physical considerations for the car or a train with multiple cars, the maximum magnitude of this drift can be estimated. These physical considerations primarily include the gradient change over the track, as well as installation errors and calibration errors of the acceleration sensors used, as well as typical train-dependent acceleration and deceleration dynamics.
[0036] This information can be used to appropriately parameterize an observer model so that, without knowledge of discrete operating conditions, the measured variables, in particular the wheel speed, can be optimally taken into account with regard to the determination of the actual car speed.
[0037] This results in a continuous tracking of the drift size and a subsequent continuous plausibility check of the correlation of the rotational speed or the wheel speed to the wheel speed determined from the wheel speed.
[0038] Preferably, at least one of the input variables is continuously weighted based on an evaluation criterion in order to adjust the influence of this input variable in the observer model. In a special embodiment, the weighting can also be completely reduced to zero, so that no influence of the corresponding input variable is effective in the observer model. Preferably, however, the influence is merely reduced by means of the weighting. Preferably, the measured rotational speed, wheel speed, or the wheel speed determined therefrom is weighted.
[0039] Preferably, a Kalman filter is used as the observer model. In this case, a variance of a measurement noise of at least one of the input variables of the observer model is further preferably adjusted. In this way, the variance of the measurement noise, i.e. the confidence level that indicates how well a measured value corresponds to reality, can be influenced, so that the influence of the corresponding input variable can be reduced here as well. In a special embodiment, the change in the variance can be modified such that no influence of the corresponding input variable is effective in the observer model. Preferably, however, the influence is merely reduced. The measured rotational speed, wheel rotational speed, or the wheel speed determined therefrom is preferably affected.
[0040] Preferably, the value of the variance of the measurement noise of at least one of the input variables is adjusted or the weighting of at least one of the input variables is determined depending on a characteristic resulting from the vehicle's drive and / or braking processes. For example, if a strong driving force is applied to the wheel, resulting in relatively high wheel slip, the weighting of the wheel speed, i.e., the speed calculated from the wheel rotational speed, is reduced, or the variance of the measurement noise of the wheel speed is changed to a value corresponding to a low confidence level. In this way, the influence of the wheel speed as an input variable can be reduced in this case.
[0041] Preferably, one of the input variables of the observer model is the wheel speed and / or an input variable is determined from the wheel speed. Preferably, a wheel speed is determined from the wheel speed, and the wheel speed is particularly preferably also one of the input variables of the observer model.
[0042] Furthermore, the observer model is preferably designed to model acceleration and braking processes, thereby enabling precise observation of the actual driving speed. For this purpose, the observer model is particularly designed to determine the driving and braking forces acting on the vehicle. This can preferably be achieved by processing measured values, such as currents from drive motors or electrodynamic brakes, application forces or braking pressures from friction brakes, or similar values.
[0043] In a preferred embodiment, a speed, an acceleration, and a deviation of a mean value of a measured acceleration from an acceleration related to a movement of the vehicle mass are determined as state variables. The accelerations are, in particular, longitudinal accelerations.
[0044] The Kalman filter is a control-engineering observer model with a special mathematical structure that allows it to be used in real-time systems. In addition to the state variables, the Kalman filter also estimates its own error.
[0045] If the speed-relevant sensor data advantageously contain measurement results of a wheel speed sensor, and a variance of a measurement noise of the state variable speed is variably adapted over time, the variance can be considered as a confidence measure for a speed obtained from a wheel speed.
[0046] With an advantageous adjustment of a variance value proportional to a difference between a requested and actually realized braking or drive value, or proportional to a factor determined from at least one of the requested and actually realized braking or drive values, the variance can be considered a confidence measure for a speed value. The braking or drive values can be a wide variety of measurable or observable values that characterize the state of the vehicle when driving, rolling, braking, or at a standstill. Characteristic quantities are measured, internally calculated, or observable quantities from the traction, drive, and brake control, such as brake pressures, motor or converter currents, acceleration values, slip values, or request values from a drive or brake lever position.
[0047] If the input variables advantageously include measurable forces introduced into the car from the outside, a prediction of resulting changes in state is possible.
[0048] In addition to the inaccuracies described above, which can arise from deviations between wheel speed and the actual vehicle speed, inaccuracies in the measured acceleration can also pose a problem when determining speed. This is particularly the case when the vehicle is traveling uphill or downhill, i.e., a stretch of road with a gradient other than zero. Therefore, such situations must be detected, and the observer model must be able to react accordingly.
[0049] Preferably, a gradient value is determined, which preferably represents one of the input variables of the observer model, wherein the gradient value is preferably measured and / or estimated and / or determined from a database, wherein the gradient value is particularly preferably used to correct the measured longitudinal acceleration. The gradient value describes the gradient of the route on which the vehicle is moving. The gradient value can be specified, for example, as an angle or as a percentage. In this way, an acceleration component can be identified in the measured longitudinal acceleration that is to be expected based on the determined gradient. This acceleration component can then be used to correct the measured longitudinal acceleration.
[0050] In particular, the car has a longitudinal direction in which it travels.
[0051] Preferably, a vertical acceleration is measured in a vertical direction of the vehicle perpendicular to the longitudinal direction, and the measured longitudinal acceleration is corrected by the vertical acceleration. Alternatively or additionally, the vertical acceleration is an input variable of the observer model, wherein a correction of the longitudinal acceleration is preferably performed by the vertical acceleration in the observer model.
[0052] Alternatively or additionally, a vehicle inclination about a horizontal axis perpendicular to the longitudinal axis and perpendicular to the vertical axis is used as one of the state variables of the observer model. Measurement variables required to determine the vehicle inclination are preferably input to the observer model as input variables, with the vertical acceleration being input in particular. Thus, a vehicle inclination can be determined from the determined vertical acceleration. The vehicle inclination can be used, similar to the previously determined gradient, to correct the measured longitudinal acceleration.
[0053] By advantageously detecting an acceleration in a direction perpendicular to a longitudinal direction of the car in a vertical plane and correcting the data of the acceleration sensor, in particular the detected longitudinal acceleration, by the acceleration in the direction perpendicular to the longitudinal direction of the car, the state observation is not influenced by the drift in the acceleration measurement.
[0054] If the acceleration in the direction perpendicular to the longitudinal direction of the car in the vertical plane is detected, and the acceleration perpendicular to the longitudinal direction of the car in the vertical plane is advantageously one of the input variables of the observer model, the state observation is not affected by the drift in the acceleration measurement.
[0055] If, advantageously, a vehicle inclination about a horizontal axis perpendicular to the longitudinal axis is one of the state variables, and measured variables required for determining the vehicle inclination are entered as input variables, the state observation is not influenced by the drift in the acceleration measurement.
[0056] If, advantageously, one of the measured variables required for determining the vehicle inclination is an acceleration in the direction perpendicular to the longitudinal direction in the vertical plane as an input variable of the observer model, the influence of the vehicle inclination can be easily captured.
[0057] For the entire document, it should be noted here that further processing of the measured longitudinal acceleration can take place both in the corrected state, as described above, and in the uncorrected state.
[0058] To evaluate the wheel speed, in particular to assess the extent to which the wheel speed can be further used, the method preferably comprises a further step: Determining a quality of the determined wheel speed, where the quality is preferably formed as a characteristic value taking into account at least one of the following variables: a brake pressure and / or an application force of a brake designed to brake the wheel; a current and / or a voltage of a device designed to impart a torque to the wheel; a drive or braking torque applied to the wheel; a weight of the vehicle, which is determined by load detection; a wheel slip of the wheel; a reference speed and / or a vehicle speed; a coefficient of friction between the wheel and a surface on which the wheel rolls; the detected acceleration; a drive or braking torque applied to the wheel; a global vehicle speed, which is determined in particular from a satellite navigation system; information as to whether an anti-skid intervention is present; information as to whether a braking intervention is present; information as to whether the corresponding wheel is driven.
[0059] A poor quality of the wheel speed is preferably given when the detected rotational speed or the wheel speed determined from it does not correlate with the actual vehicle speed and / or deviates from a reference value, such as a reference speed. This is particularly the case when the wheel skids or locks, i.e. when the wheel cannot fully transfer an applied drive or braking torque to a surface on which the wheel is rolling. The wheel speed then determined can no longer be reliably used for the further determination of the vehicle speed, so that as a result the influence of the wheel speed on the determination of the vehicle speed is preferably reduced and particularly preferably the vehicle speed is essentially determined from the detected acceleration.
[0060] Preferably, the quality of the wheel speed is determined as a characteristic value. Its determination is preferably designed such that the characteristic value changes to a value corresponding to a poorer quality when a braking intervention occurs. Alternatively or additionally, the characteristic value changes to a value corresponding to a poorer quality when a wheel-skid protection intervention occurs. Alternatively or additionally, the characteristic value changes to a value corresponding to a poorer quality when a deviation of the detected wheel speed from the reference speed exceeds or falls below a predetermined value.
[0061] The quality characteristic value is preferably determined in such a way that, starting from a certain car speed, in a state in which the car rolls freely, i.e. is neither driven nor braked, it assumes a value which corresponds to a higher quality, compared to states in which the car is driven or braked.
[0062] In a particular embodiment, a reference speed is formed according to the method disclosed in EP 0 499 947 A1 (reference speed of the anti-skid device).
[0063] The reference speed is preferably calculated in such a way that a theoretical speed of the wagon, i.e. the reference speed, is calculated from all drive or braking variables or drive or braking forces acting on the wagon or the wheel whose speed is being recorded, whereby it is assumed that the drive or braking forces are optimally transferred to the rail. This means that the reference speed corresponds to a theoretically possible speed of the wagon if the corresponding wheel does not slip or spin. The influence of the wagon's inertia is taken into account when determining the reference speed. Preferably, the inertia from the wagon's mass is taken into account. Particularly preferably, a rotational inertia of rotating elements of the wagon, such as the wheels, is also taken into account.
[0064] This allows the quality of the determined wheel speed to be advantageously determined purely by comparing the reference speed with the actual wheel speed. If the car is driven, the wheels are at risk of skidding. They would therefore rotate faster than the car's inertia would allow. In this case, the wheel speed would be higher than the corresponding reference speed. If the car is braked, the wheels are at risk of locking. They would therefore rotate slower than the car's inertia would allow, since the car's inertia continues to "push" the wheels. In this case, the wheel speed would be lower than the corresponding reference speed. In both cases, the quality of the wheel speed can be determined by calculating the difference between the absolute values of the wheel speed and the reference speed.If the difference is small or even zero, the wheel speed is close to the actual vehicle speed or completely matches it. Alternatively or additionally, a percentage deviation or another way to compare the wheel speed with the reference speed can be provided.
[0065] The reference speed can also preferably be formed from at least two of the determined wheel speeds. The reference speed is preferably suitable for identifying implausible wheel speeds or a poor quality of a wheel speed by comparing it with individual wheel speeds.
[0066] Furthermore, it is also conceivable to use several reference speeds that are determined in different ways.
[0067] Preferably, the variance of the measurement noise of at least one of the input variables and / or the weighting of at least one of the input variables is influenced by the quality, in particular by the characteristic value, by exceeding or falling below a limit value. Thus, by appropriately calculating the quality of the wheel speed, the influence of the wheel speed on the determination of the vehicle speed or on the observer model can be changed.
[0068] According to the invention, a method for determining a train speed of a train with at least one car is further provided, comprising the steps: Determining at least one car speed of a car belonging to a train with at least one car according to the method described above, and determining a train speed from the at least one car speed, preferably by averaging the at least one car speed; preferably performing a plausibility check of the at least one car speed, in particular by means of a diagnostic function which checks the plausibility of individual ones of the at least one car speed; and preferably ignoring implausible, in particular local, car speeds for determining the train speed, and / or preferably ignoring car speeds which were determined on the basis of wheel speeds with a quality which is below a predetermined quality level.
[0069] Preferably, a Kalman filter is used as the observer model for determining the at least one car speed, wherein a variance of a measurement noise in the Kalman filter is determined as a measure of confidence for checking the plausibility of the local car speed, and if a minimum level of confidence of one of the local car speeds is undershot, the associated wheel speed or the associated wheel speed is ignored as an input variable.
[0070] By ignoring car speeds that have wheel speeds of lower quality than a certain value in their calculation, the overall train speed calculation can be improved. This advantageously ensures that car speeds determined based on high-quality wheel speeds are included in the calculation.
[0071] Preferably, the method for determining the train speed comprises the following step: Adjusting or correcting the determined train speed with a train speed formed from other sources, wherein another source is preferably a radar measurement, a GPS measurement or an acceleration and / or speed-independent measurement, wherein the adjustment or correction is preferably carried out by means of fusion using a Kalman filter.
[0072] According to a further aspect of the invention, a method for determining a train speed advantageously includes the step of determining a train speed from a plurality of local car or wheel speeds, e.g., by calculating the arithmetic mean. This allows the train speed to be easily determined from the determined local car speeds.
[0073] By advantageously performing a diagnostic function by checking the plausibility of the individual local car speeds and ignoring implausible local car speeds for determining the train speed, the accuracy of the determined train speed can be increased.
[0074] If a Kalman filter is advantageously used as the observer model for a local car or wheel speed, a variance of a measurement noise in the Kalman filter is determined as a measure of confidence for the plausibility check, and if a minimum level of confidence is exceeded, the speed value obtained from the corresponding wheel speed is ignored, the accuracy of the determined train speed can also be increased.
[0075] In the case where the determined train speed is advantageously compared with a train speed formed from other sources, the accuracy of the determined train speed can be increased.
[0076] According to the invention, a device for determining a car speed or a train speed is further provided, which device is designed to carry out the method for determining the car speed or the train speed as described above.
[0077] According to the invention, a vehicle, in particular a rail vehicle, in particular comprising at least one carriage, is further provided, wherein the vehicle is designed to carry out the method for determining the carriage or train speed as described above, and / or wherein the vehicle has a device as described above.
[0078] According to the invention, a computer program product is further provided with code means stored on a machine-readable carrier and which
[0079] Cause a data processing device, preferably a device as described above, to carry out the method for determining the car or train speed as described above when the code means are executed on the data processing device.
[0080] The invention will now be explained by means of embodiments with reference to the accompanying drawings.
[0081] In particular, Fig. 1 shows a schematic representation of an observer model for determining a local car speed using a method according to a first embodiment of the invention; Fig. 2 shows a schematic representation of an observer model for determining a local car speed using a method according to a second embodiment of the invention; Fig. 3 shows a schematic representation of an observer model for determining a local car speed using a method according to a third embodiment of the invention; Fig. 4 shows a schematic representation of an observer model for determining a local car speed using a method according to a fourth embodiment of the invention; Fig. 5 shows a schematic representation of an observer model for a method according to the invention for sensor data fusion for individual local car speeds; Fig. 6 shows a schematic diagram of a structure for the method according to the second embodiment of the invention; and Fig.Figure 7 shows a diagram of a calculation of a wheel speed quality.
[0082] Fig. 1 shows a schematic representation of an observer model 1 for determining a local car speed v car using a method according to a first embodiment of the invention.
[0083] By means of the control-engineering observer model 1 functioning as a simulation model of a system of a chassis unit, in particular a bogie of a wagon of a rail vehicle, into which measurement results from sensors 3 arranged on the chassis unit or the bogie ( Fig. 6 ) are input as input variables of the system, state variables are determined. Alternatively, the system can also be one of the wagons of the rail vehicle, in which case the sensors 3 are not necessarily located on the bogie unit or the bogie, but are provided at a suitable location on the wagon.
[0084] The state variables can, for example, be more precise or corrected measurement results from sensors 3. Furthermore, hidden state variables that cannot be measured or can only be measured with great effort can also be determined. The state variables are determined using equations, in particular differential equations, that define the behavior of the state variables in relation to the respective input variables, but also the behavior of the state variables in relation to each other. This allows precise state variables to be determined by sensor fusion, even in the case of individual inaccurate or erroneous measurement results, by considering several relevant measurement results.
[0085] The observer model 1 can optionally contain a Kalman filter 2 ( Fig. 2 ) who can also estimate his own errors when determining the state variables.
[0086] The input variables include a longitudinal acceleration ax,mess of the vehicle measured by an acceleration sensor and a wheel speed or angular velocity ω axis of a wheel axle. Instead of an angular velocity ω axis, a rotational speed n, specifically a wheel speed n wheel or a wheel speed v Odo , determined, for example, from the wheel speed n wheel and the wheel radius, can also be entered.
[0087] One state variable determined in the observer is the local car velocity v car . Alternatively, other state variables, such as acceleration or drift d, can be determined and output if necessary.
[0088] The carriage has a longitudinal direction and comprises an acceleration sensor that detects an acceleration az,mess in a direction perpendicular to the longitudinal direction in a vertical plane.
[0089] During operation, the speed-relevant sensor data available locally in the car are recorded as the respective input variable for observer model 1 and entered into observer model 1. The sensor data contains the data from the acceleration sensor for the longitudinal acceleration ax,mess of the car and the wheel speed or angular velocity ω axis of the wheel axle. Observer model 1 continuously monitors the state of the input variables and determines the state variable, namely the local car speed v car. Continuous in this context means that there is no switching between different operating modes, in which different input variables are evaluated. Alternatively, further speed-relevant sensor data, such as data from a positioning system, e.g.by means of a satellite navigation system, brake pressures or supply voltages or supply currents of drive motors.
[0090] The speed-relevant sensor data is collected by the sensors located on the running gear unit or bogie. Alternatively, the sensor data is collected by sensors located at a suitable location on the car. Furthermore, sensor data can also be collected from sensors other than the acceleration sensor and a pulse generator for the wheel speed or the angular velocity of the axle.
[0091] An acceleration az,mess in a direction perpendicular to the longitudinal direction of the carriage in a vertical plane is measured, and the acceleration sensor data for the acceleration ax,mess in the longitudinal direction are corrected by the acceleration az,mess in the vertical plane perpendicular to the longitudinal direction and input. Alternatively, the data for the acceleration ax,mess in the longitudinal direction are not corrected by the acceleration az,mess in the vertical plane perpendicular to the longitudinal direction, but the acceleration az,mess perpendicular to the longitudinal direction of the carriage in the vertical plane is an input variable of the observer model.
[0092] Alternatively, the rotation of the vehicle's coordinate system by the angle of the gradient and inclination of the elevation profile and thus the correction value of the driving speed to be determined in the vehicle's longitudinal plane can also be determined by other methods.
[0093] Fig. 2 shows a schematic representation of an observer model as a Kalman filter 2 for determining the local car speed v car using a method according to a second embodiment of the invention.
[0094] The observer model in the second embodiment differs from the observer model 1 in the first embodiment in that it is necessarily a Kalman filter 2 or a control-engineering observer model that contains a functionality of the Kalman filter and that is designed as a control-engineering observer model specialized for an acceleration and braking process.
[0095] During operation, a speed of the car v car , an acceleration of the car and a drift, namely a systematic deviation of a measured acceleration from an acceleration related to a movement of the vehicle mass of the car, are determined as state variables, which leads to an accurate observation of the actual driving speed.
[0096] The speed-relevant sensor data contain measurement results of a wheel speed sensor ω axis , and a variance Var(X,t) of a measurement noise of the state variable speed is variably adapted over time.
[0097] Optionally, a value of the variance Var(x,t) is adjusted proportionally to a difference between a requested and actually realized braking or driving value or proportionally to a factor determined from at least one of the requested and actually realized braking and driving values as explained above.
[0098] Fig. 3 shows a schematic representation of the observer model 1 for determining the local car speed v car using a method according to a third embodiment of the invention.
[0099] The method according to the third embodiment differs from the method according to the first embodiment in that a vehicle inclination α about a horizontal axis perpendicular to the longitudinal axis of the vehicle is one of the state variables, and measured variables required for determining the vehicle inclination are input as input variables.
[0100] The measurement required to determine the vehicle inclination α, in addition to the acceleration ax,mess in the longitudinal direction, is an acceleration az,mess in a vertical direction perpendicular to the longitudinal direction to determine the vehicle inclination α. Alternatively, other suitable measurement variables can be used to determine the vehicle inclination, for example, a measurement result from a gyroscope or inclinometer, or by combining a known topography of a route section with a positioning method, for example, by comparing the current position with digital map material.
[0101] Fig. 4 shows a schematic representation of the observer model 1 for determining the local car speed v car using a method according to a fourth embodiment of the invention.
[0102] The method according to the fourth embodiment differs from the methods according to the first or third embodiment in that the input variables additionally contain measurable forces F x introduced into the carriage from the outside.
[0103] Ideally, all known forces are included, with all other forces being recorded as part of a disturbance variable observation approach. In particular, the measurable forces introduced during a traction phase must be included, and during the braking phase the corresponding braking forces acting in the braking system must be taken into account. These are caused, for example, by pneumatic pressures, the use of adhesion-independent brakes (e.g. Mg brakes) or brakes that support the braking force via the wheel on the rail (e.g. wheel brakes), and other devices that drive or brake the wagon (e.g. retarders, electrodynamic brakes), and the variables that can be derived directly or indirectly from them. In this observer model, physical relationships of the driving resistances can also be optionally taken into account, and their residual errors during a rolling phase can be compensated for.
[0104] Fig. 5 shows a schematic representation of an observer model 1, 2 for a method according to the invention with a sensor data fusion for individual local car speeds v car,i to v car,n of a train for determining a train speed v train .
[0105] A plurality of local car speeds v car,i to v car,n of the cars of the train, determined according to a method of the preceding embodiments, are averaged by the observer model 1, 2 to determine the train speed v train.
[0106] The observer model 1, 2 for sensor data fusion, for example, is a conventional observer model 1, but can optionally also be a Kalman filter 2. However, it can also be designed as a simple arithmetic averaging.
[0107] Optionally, a diagnostic function is performed by checking the plausibility of the individual local car speeds v car, i. Implausible local car speeds v car, i are ignored for determining the train speed v train.
[0108] In each case, a Kalman filter 2 is used as the observer model for determining the local car speeds v car,i to v car,n. These Kalman filters 2 each determine a variance of a measurement noise Var car,i (ẋ,t) to Var car,n (ẋ,t) as a measure of confidence for checking the plausibility of the state variable of the local car speed v car,i to v car en,n of the individual cars. The variance of the respective measurement noise Var car,i (ẋ,t) for the individual local car speeds v car is entered as an input variable into the control observer model 1, 2 for sensor data fusion for individual local car speeds v car,i to v car en,n. If a minimum level of confidence for one of the local car speeds v car is undershot, a corresponding wheel speed is devalued accordingly as an input variable in the Kalman filter 2 or is not taken into account.
[0109] The train speed determined by the observer model 1, 2 for sensor data fusion for individual local vehicle speeds v Wagen,i to v Wagen,n is compared with a train speed calculated from other sources in order to implement further diagnostic functions and corrective measures. In this embodiment, a vehicle speed v Verortung determined by means of a positioning system, e.g., a satellite navigation system, is input into the observer model 1, 2 for sensor data fusion for individual local vehicle speeds v Wagen,i to v Wagen,n.
[0110] Fig. 6 shows a schematic diagram of a structure for the method according to the second embodiment of the invention.
[0111] Sensors 3 record respective measurement results, such as longitudinal acceleration, momentum, or forces, and input any erroneous measurement signals as input variables to Kalman filter 2. In Kalman filter 2, the erroneous measurement results are processed using the simulation model, and precise output signals are determined. These precise output signals are then provided for and processed by optimized algorithms in railway operations, such as wheel slide protection or inclination detection.
[0112] Alternatively or additionally, the following variants can be included in one of the methods according to the invention: A wheel diameter is defined as a further state variable, with low process noise since the wheel diameter is approximately constant.
[0113] This allows wheel wear to be determined without further measurements, for example using an "extended Kalman filter".
[0114] Slippage in a wheel-rail contact is defined as a further state variable. The pulse generator then measures the difference between the speed and the slip. A very precise calculation of instantaneous slip is possible, especially in conjunction with precise position measurement.
[0115] A gradient or a gradient are not known in advance, but are determined as an additional disturbance variable from the Kalman filter 2.
[0116] The acceleration is fused from a position signal and a speed signal.
[0117] A simplified 2D simulation model of a car body, bogie, and wheelsets is integrated into the Kalman filter 2 with slip as one of the state variables. This allows a wheel contact force to be defined as one of the state variables.
[0118] All features presented in the description, the following claims, and the drawings may be essential to the invention both individually and in any combination, with the invention being defined by the claims. In particular, each observer model can be configured to determine at least one of the states: speed, acceleration, drift magnitude, and gradient value.
[0119] Fig. 7 shows a diagram of a calculation of a quality of wheel speed.
[0120] A calculation block B is shown, which is designed to determine a quality Q of the wheel speed v Odo based on input values Q1, Q2, Q3, Q4, which lead to the left side of the calculation block B. Finally, the calculation block B outputs the quality Q of the wheel speed v Odo on its right side.
[0121] The input value Q1 comes from a calculation block that represents a base value calculation. Based on one or more parameters P1 resulting from the vehicle's operation, a base value for the quality Q is calculated here, which is then transmitted as input value Q1 to calculation block B. For example, a parameter P1 can simply be information about whether a brake intervention is present, whether the corresponding wheel is driven, or whether it is torque-free.
[0122] Another input value, Q2, comes from a calculation block that specifically processes parameter P2 from the car's anti-skid function. For example, if the car's anti-skid function is triggered, the input value Q2 is changed accordingly, so that calculation block B subsequently outputs a characteristic value corresponding to a lower quality Q.
[0123] Another input value, Q3, comes from a calculation block that processes the wheel speed v Odo and a reference speed v Ref as input variables. For example, by comparing the wheel speed v Odo with the reference speed v Ref, a minimum slip can be determined that would occur compared to the car at zero slip and with ideal power transmission between wheel and rail. Such a minimum slip can then be entered, for example, as input value Q3 into calculation block B. However, the input value Q3 can also be an abstract value.
[0124] Another input value Q4 comes from a calculation block that evaluates moments and / or forces on the wheel and uses this to determine the input value Q4. In this example, the quantities M and P are used as input variables. M represents a variable that corresponds to a drive torque acting on the wheel, or a variable from which the drive torque of the wheel can be determined. For example, this could be a current from an electric motor. The variable P represents a variable that corresponds to a braking torque acting on the wheel, or a variable from which the braking torque of the wheel can be determined. For example, this could be a current from an eddy current brake, a brake pressure, or something similar. From such a variable M, P, a force between the rail and the wheel, which corresponds to a drive or braking force, can also be determined.If the quantities M, P are relatively high, an input value Q4 can be determined, which causes the calculation of a lower quality Q in the calculation block B, compared to a state with lower quantities M, P.
[0125] In a special embodiment, the calculation block B is designed as an addition operator which adds the input values Q1, Q2, Q3, Q4 and thereby forms the quality Q as a sum.
[0126] The processing of the input values Q1, Q2, Q3, Q4 can also be weighted.
[0127] In addition, embodiments for quality calculation are conceivable that only process a part of the input values Q1, Q2, Q3, Q4 shown here. LIST OF REFERENCE SYMBOLS
[0128] 1Observer model 2Kalman filter 3Sensor ax actual acceleration in the longitudinal direction of the car ax,mess measured acceleration in the longitudinal direction of the car az,mess measured acceleration in the vertical direction of the car αGradient value dDrift quantity / Drift nSpeed n Wheel Wheel speed v Odo Wheel speed Var(x,t)Variance of the measurement noise of the state variable Speed Var Car,i (ẋ,t)Variance of the measurement noise of the state variable Speed of the car iv Car Speed of the car v Train Train speed ω Axle Angular velocity of an axle BCalculation block MAnput torque representing quantity PBraking torque representing quantity P1Parameter P2Parameter P3Parameter P4Parameter QQuality Q1Input value Q2Input value Q3Input value Q4Input value v Ref Reference speed
Claims
1. Method for determining a carriage speed (vcarriage) of a carriage with the steps of: - Acquiring sensor data available in the carriage as a respective input variable for an observer model (1, 2); - Inputting the input variables into the observer model (1, 2); - Performing a continuous determining of the carriage speed (vcarriage) by the observer model (1, 2), characterised in that the observer model (1, 2) determines a drift variable (d), which is determined as the state variable of the observer model (1, 2), and wherein the drift variable (d) is defined as a value of a - difference of an actual longitudinal acceleration (ax) of the carriage from a measured longitudinal acceleration (ax,measured), and / or - difference of a longitudinal acceleration, which is determined from a rotational speed (n), preferably from a wheel rotational speed (nwheel), from the measured longitudinal acceleration (ax,measured), and / or - difference of the longitudinal acceleration, which is determined from the rotational speed (n), preferably from the wheel rotational speed (nwheel), from the actual longitudinal acceleration (ax).
2. Method according to claim 1, wherein the sensor data contain the measured longitudinal acceleration (ax,measured) of the carriage and / or the rotational speed (n) of a carriage element, the rotational speed (n) of which is proportional to the wheel rotational speed (nwheel) and / or wherein the input variables contain measurable forces or forces (Fx) introduced by the braking or drive system of the carriage preferably determined indirectly by means of an observer model.
3. Method according to claim 1 or 2, wherein the observer model (1, 2) uses a speed and an acceleration of the carriage as state variables, wherein the speed is preferably the determined carriage speed (vcarriage) and / or the acceleration is a longitudinal acceleration of the carriage.
4. Method according to any one of the preceding claims, wherein at least one part of the sensor data is acquired by sensors arranged on a chassis unit, in particular a bogie.
5. Method according to any one of the preceding claims, wherein the sensor data, in particular speed-related measured values, contain - an acceleration sensor, - an inclination sensor - an inertial measurement unit, - a rotational speed sensor, - a brake pressure sensor, - a braking current sensor, and / or - an engine6. Method according to any one of the preceding claims, wherein a drift dynamic is determined by determining a rate of change over time of the drift variable (d), preferably by the observer model (1, 2), wherein the value of the drift dynamic is preferably compared to predetermined maximum values and further preferably, when the predetermined maximum values are exceeded by the value of the drift dynamic, a faulty retrieval process of the rotational speed (n), in particular of the wheel rotational speed (nwheel), is assumed, and the predetermined maximum values preferably include gradient and inclination changes over a route, installation and calibration errors affecting the acceleration sensors used and normal acceleration and deceleration dynamics of the train.
7. Method according to any one of the preceding claims, wherein a weighting of at least one of the input variables occurs continuously on the basis of an evaluation criterion, in order to adapt the influence of this input variable in the observer model (1, 2).
8. Method according to any one of the preceding claims, wherein a Kalman filter (2) is used as the observer model (1, 2), and a variance of a measuring noise is preferably adapted to at least one of the input variables of an input variable of the observer model (1, 2).
9. Method according to claim 7 or 8, wherein adaptation of the value of the variance of the measuring noise is performed for at least one of the input variables or weighting occurs for at least one of the input variables as a function of a parameter, arising from the carriage's drive and / or braking processes.
10. Method according to any one of the preceding claims, wherein one of the input variables of the observer model (1, 2) is the wheel rotational speed (nwheel) or is determined from a wheel rotational speed (nwheel), wherein a wheel speed (nodo) is preferably determined from the wheel rotational speed (nwheel) and the wheel speed (vodo) is most preferably also one of the input variables of the observer model (1, 2).
11. Method according to any of the preceding claims, with the features according to claim 2, wherein a gradient value (a) is determined, which preferably represents one of the input variables of the observer model (1, 2), wherein the gradient value (a) is preferably measured and / or estimated and / or determined from a database, wherein the gradient value (a) is most preferably used to correct the measured longitudinal acceleration (ax,measured).
12. Method according to any one of the preceding claims, wherein the carriage has a longitudinal direction and a vertical acceleration (az,measured) is determined in a vertical direction of the carriage perpendicular to the longitudinal direction in a vertical plane, and the measured longitudinal acceleration (ax,measured) is corrected by the vertical acceleration (az,measured) and / or the vertical acceleration (az,measured) is an input variable of the observer model (1, 2), wherein a correction of the longitudinal acceleration (ax,measured) preferably occurs through the vertical acceleration (az,measured) in the observer model (1, 2) and / or wherein a vehicle inclination about a horizontal axis perpendicular to the longitudinal axis and perpendicular to the vertical axis is used as one of the state variables of the observer model (1, 2), and measured variables necessary to determine the vehicle inclination are input as input variables into the observer model (1,2), wherein the vertical acceleration (az,measured), in particular, is input.
13. Method according to any of claims 10 to 12, having a further step of: - determining a quality of the determined wheel speed (vodo), wherein - the quality is preferably formed as a characteristic value taking account of at least one of the following variables: - a brake pressure and / or an application force of a brake, which is designed to brake the wheel, - a current and / or a voltage of a device, which is designed to apply a torque to the wheel, - a drive or braking torque applied to the wheel, - a weight force of the carriage, which is determined by load detection - a wheel slip of the wheel, - a reference speed (vRef) and / or the carriage speed (vcarriage), - a coefficient of friction between the wheel and a surface on which the wheel rolls, - the detected acceleration (ameasured), - a drive or braking torque applied to the wheel, - a global vehicle speed, - information on whether anti-skid protection has been applied, - information on whether the brakes have been applied, - information on whether the corresponding wheel is driven.
14. Method according to claim 13, wherein the variance of the measuring noise for at least one of the input variables and / or the weighting for at least one of the input variables is influenced by the quality, in particular the characteristic value, exceeding or falling below a threshold.
15. Method for determining a train speed (vtrain) with the steps of: - Determining at least one carriage speed (vcarriage) of a carriage, which belongs to a train with at least one carriage according to the method of any one of claims 1 to 14, and - Determining a train speed (vtrain) from the at least one carriage speed (vcarriage), preferably by averaging the at least one carriage speed (vcarriage), - preferably performing a plausibility check of the at least one carriage speed (vcarriage), and - preferably disregrading non-plausible carriage speeds (vcarriage) for determining the train speed (vtrain), and / or - preferably disregarding carriage speeds (vcarriage), which have been determined on the basis of wheel speeds (vodo) with a quality below a predetermined quality level.
16. Method according to claim 15, wherein a Kalman filter is used as the observer model for determining the local carriage speed (vcarriage), a variance (Varcarriage,i(x,t)) of a measuring noise in the Kalman filter is determined as a measure of confidence for checking the plausibility of a local carriage speed (vcarriage), and when a minimum measure of confidence of one of the local carriage speeds (vcarriage) is not reached, the corresponding wheel rotational speed (nwheel) or the corresponding wheel speed (vodo) is disregarded as an input variable.
17. Method according to any of claims 15 or 16, with the step of: - Comparing or correcting the determined train speed (vtrain) with a train speed (vtrain) formed from other sources, wherein another source is preferably a radar measurement or a GPS measurement, wherein the comparing or correcting preferably occurs by means of merging using a Kalman filter.
18. Device for determining a carriage speed (vcarriage) or a train speed (vtrain), which is designed to perform the method according to any one of claims 1 to 17.
19. Vehicle, in particular a rail vehicle, in particular having at least one carriage, wherein the vehicle is designed to perform the method according to any one of claims 1 to 17, and / or wherein the vehicle has a device according to claim 18.
20. Computer program product with code means, which are stored on a machine-readable medium and which prompt a data processing device, preferably a device according to claim 18, to perform the method according to any one of claims 1 to 17, when the code means are executed on the data processing device.
