METHOD FOR ORDERING VEHICLES IN A VEHICLE FLEET ACCORDING TO MAINTENANCE REQUIREMENTS, CORRESPONDING COMPUTER PROGRAM AND COMPUTER SYSTEM

Description

[0001] The present invention relates to the maintenance of a fleet of vehicles, in particular railway vehicles, such as trains, metros, trams, etc.

[0002] Systems for monitoring potential failures of a railway vehicle are known based on the generation of events during the operation of this vehicle. More precisely, the monitored vehicle is equipped with a plurality of sensors capable of acquiring data which, once processed, make it possible to generate characteristic events. An event constitutes a new entry in a monitoring database. It is defined by a plurality of attributes, such as for example a vehicle identifier, an event identifier, a time of occurrence of the event, the type to which the event belongs, and different values ​​characterizing it (which may depend on the type to which the event belongs).

[0003] Currently, the events generated for a vehicle are analyzed by maintenance operators in order to carry out a diagnosis on the condition of the vehicle and determine whether it is necessary to carry out a maintenance operation on this vehicle and the nature of this maintenance operation: simple inspection or replacement of such or such faulty component (or about to fail) of the railway vehicle in question.

[0004] However, such monitoring systems generate thousands of events every day and for each vehicle in a fleet.

[0005] Such a mass of information is difficult for an operator to analyze. Such an analysis remains complex, particularly when it comes to establishing meaningful correlations between events. It therefore requires a significant investment in terms of work time and costs.

[0006] Furthermore, it is virtually impossible for an operator to compare vehicles in a fleet to determine which one is in urgent need of maintenance in order to prioritize the required maintenance operations on that vehicle over another. Today, the choice to inspect one vehicle over another is essentially an empirical decision made by the maintenance operator.

[0007] It would therefore be interesting to be able to quickly and simply identify the vehicle(s) in a fleet that have a priority need for maintenance, if only to optimize maintenance tasks and limit the time during which equipment cannot be used.

[0008] Document US 2019 / 197798 A1 presents a computer-implemented method for ordering the vehicles of a fleet of vehicles according to a maintenance need, the method comprising the steps of, for each vehicle of the fleet: determining a time series, the time series comprising, for each time step, an instantaneous value of at least one quantity of interest obtained from monitoring events acquired by means of a system for monitoring the vehicles of the fleet of vehicles; and analyzing, over a predetermined time interval, the time series, taking into account the time series determined for all the vehicles of the fleet of vehicles, considering that at each time step, the state of a vehicle is either a “normal” state or an “abnormal” state requiring maintenance, so as to obtain an optimal sequence of states over the predetermined time interval.

[0009] The aim of the present invention is in particular to meet this need.

[0010] For this purpose, the invention relates to an alternative method, a computer program product and a computer system according to the appended claims.

[0011] The invention and its advantages will be better understood upon reading the following detailed description of a particular embodiment, given solely as a non-limiting example, this description being made with reference to the appended drawings in which: [ Fig 1 ] There figure 1 is a schematic representation of a computer system for implementing the method according to the invention; and, [ Fig 2 ] There figure 2 is a block representation of a preferred embodiment of the scheduling method according to the invention.

[0012] The method according to the invention makes it possible to automatically order the trains of a fleet of trains according to a maintenance requirement.

[0013] A train fleet is a group of trains that are similar or even identical to each other and that are put into circulation on a network in an essentially interchangeable manner. In other words, the question of ordering the trains in a fleet in relation to each other according to a maintenance need makes sense, since these different trains are comparable in their structure (and therefore their wear and tear and potential failures) and in their use (and therefore in their operating conditions).

[0014] The method according to the invention is based on the detection of bursts or bursts by analyzing a time series of a quantity of interest developed from events generated by a state-of-the-art monitoring system.

[0015] In the context of time series analysis of a quantity, burst detection is a technique for identifying values ​​of that quantity that show a sharp increase over a limited period of time compared to a baseline level.

[0016] Many burst detection algorithms are known, particularly in signal processing applications. Generally, a burst detection algorithm identifies periods of time (or bursts) during which a target value is unusually frequent.

[0017] Burst detection algorithms can be adapted for discrete data streams, such as events generated by a train fleet monitoring system. In the following, the embodiment that is presented in detail implements a burst detection algorithm based on logarithmic maximum likelihood.

[0018] According to the invention, burst detection for a particular train is performed by determining the baseline level from the events generated by the monitoring system for all trains in the fleet.

[0019] There figure 1 represents a computer system 10 for implementing the method according to the invention.

[0020] It includes a monitoring database 12 collecting events generated by a monitoring system 20 of the fleet's trains.

[0021] The computer system 10 comprises a computer 14 comprising calculation means, such as a processor, and storage means, such as a memory. The memory stores in particular the instructions of computer programs, in particular a program 30 whose execution allows the implementation of the method according to the invention.

[0022] The program 30 can be broken down into modules, in particular a module 32 for determining a time series of a quantity of interest, a module 34 for analyzing a time series, a module 36 for detecting bursts, and a module 38 for scheduling the trains in the fleet.

[0023] The computer system 10 advantageously comprises a human-machine interface 16, for example a tablet, making it possible to display the results of the implementation of the method according to the invention for the benefit of a maintenance operator.

[0024] Referring to the figure 2 , a preferred embodiment of a method for ordering trains of a fleet of trains according to a maintenance need is presented.

[0025] The method 100 is implemented independently of a prior step 90 of acquiring events.

[0026] The method 100 as such comprises a step 110 of determining a time series of a quantity of interest or characteristic (this step corresponding to the execution of the module 32), a step 120 of analyzing a time series (this step corresponding to the execution of the module 34), a step 130 of burst detection (this step corresponding to the execution of the module 36) and a step 140 of scheduling the trains of the fleet (this step corresponding to the execution of the module 38). Acquisition step 90

[0027] This step, prior to the implementation of the method according to the invention, is carried out in accordance with the prior art.

[0028] It is carried out by the monitoring system 20.

[0029] It consists of acquiring, for each train in the fleet, monitoring data and creating events from the acquired data.

[0030] The data is delivered by sensors fitted to each train.

[0031] Events are associated with a particular train. They are dated. For example, the time step chosen for dating events is that of the day.

[0032] Events are recorded in the monitoring database 12.

[0033] It should be noted that a high number of occurrences of an event may be due to an actual failure of the monitored equipment (the monitoring system may then continue to report this failure until it is repaired). But, a high number of occurrences may also be due to an error in the design of the monitoring system itself, in particular the modeling of the failure that this system uses to generate monitoring events (and then the monitoring system reports the event several times over time). Step 110 of determining a time series of a quantity of interest

[0034] The purpose of step 110 is to construct, for each train, a time series of a quantity of interest from the events contained in the monitoring database 12.

[0035] The quantity of interest is any temporal variable developed from all or part of the monitoring events associated with a train and allowing, by aggregating these

[0036] monitoring events, to quantify a maintenance need at each time step.

[0037] For example, we first define a time window corresponding, for example, to a week of train operation.

[0038] The quantity of interest is then, for example, the total number of monitoring events generated for the train considered within the time window. The quantity of interest thus calculated is dated, for example, with the last day of the time window used.

[0039] Simply repeat this operation by sliding the time window of the chosen time step (here the day), to calculate a quantity of interest for each day of operation of a train.

[0040] The time series for a train is then made up of the sequence of values ​​of the quantity of interest for this train.

[0041] Such a time series is obtained for each of the trains in the fleet.

[0042] Alternatively, instead of the cumulative number of events, the quantity of interest could correspond to the cumulative number of events of a particular type (i.e., associated with a specific failure or a specific component of the train). Alternatively, instead of analyzing only a time series of a quantity of interest, it could be chosen to calculate different quantities of interest and to follow each of these quantities over time. Step 120 of analyzing a time series

[0043] For each train in the fleet, this step consists of analyzing the time series of the quantity of interest obtained in step 110 by applying a burst detection algorithm, then quantifying the maintenance requirement of the train considered based on the bursts detected.

[0044] This step takes into account not only the time series associated with the train considered but also the time series associated with the other trains in the fleet.

[0045] The burst detection algorithm more particularly used in the present embodiment is based on the model that at each time step a train can only be in two possible states, respectively a nominal operating state (“normal state” in the following) and an operating state requiring maintenance (“abnormal state” in the following).

[0046] More precisely, assuming that the events follow a binomial distribution when the train is in the normal state and a binomial distribution (identical to simplify the present description) when the train is in the abnormal state, the burst detection algorithm aims to determine the sequence of states which best corresponds to the observations constituted by the values ​​of the time series of the quantity of interest over the considered time interval.

[0047] Consider, for example, a time interval of one week, subdivided into time steps of one day. Each time step in the interval is indexed by the integer t between 1 and T, where is the maximum number of time steps in the interval of interest (here T is equal to 7).

[0048] Consider a fleet composed of N trains. Each train is indexed by an integer j between 1 and N

[0049] The daily value of the characteristic quantity for train j on day t is noted: gj (t)

[0050] The daily value of the total characteristic quantity for all trains in the fleet on day t is given by: g F t = ∑ j = 1 N g j t

[0051] The integrated value of the characteristic quantity for train j over the time interval is given by: G j = ∑ t = 1 T g j t

[0052] The integrated value of the total characteristic quantity for all trains in the fleet over the time interval is given by: G F = ∑ t = 1 T g F t

[0053] We define the probability of the base level for train j, or reference probability, by: p j ref = G j / G F

[0054] This reference probability is associated with the normal state for train j. It corresponds to the events affecting train j in relation to all the events affecting the trains in the fleet, over the time interval considered.

[0055] The associated burst probability for train j is then defined by: p j burst = α . p j ref , where α is a constant of proportionality.

[0056] This burst probability is associated with the abnormal state for train j.

[0057] It is then a matter of calculating a sequence of states s(t) over the time interval considered which is compatible with the characteristic values ​​gj (t) and g F (t) for train j.

[0058] To do this, a statistical criterion is chosen. For example, in the present embodiment, the statistical criterion chosen is based on a cost function combining a logarithmic likelihood function, advantageously combined with a transition function.

[0059] The instantaneous likelihood function associated with state s(t) at time t (also noted s ) for train j is the binomial likelihood function (t): L s , j t = g F t g j t . p j s g j t . 1 − p j s g F t − g j t

[0060] For numerical reasons, we seek to minimize the logarithm of (t) over the considered time interval. In order to evaluate the likelihood of the normal state (s = “normal”), the probability p j s is taken equal to p j ref . The likelihood of the abnormal state (s = “abnormal”) is assessed by taking p j s equal to p j burst .

[0061] The transition function H s , j s ' t of the state s at time t to state s' at the next instant t+1 allows a cost to be associated with a change of state.

[0062] Preferably, since, by definition, the normal state is the expected state of the train and the abnormal state is an "exceptional" state, the transition function is non-zero only if there is a change of state and the transition is from the normal state to the abnormal state.

[0063] It is conceivable that the transition function is proportional to T, to make it comparable to the likelihood term in the optimization process as T increases (i.e., as the time interval increases).

[0064] Thus, the instantaneous criterion to minimize is as follows: f j t = − ln L s ,j t + H s , j s ′ t

[0065] Over the time interval, the criterion to minimize becomes: F j = ∑ t = 1 T f j t

[0066] The optimal sequence of states s(t) of train j, that is to say the sequence of states ("normal" or "abnormal") of train j which best explains the observation according to the binomial model, is therefore the one which minimizes the cost function for all the time steps of the time interval considered.

[0067] The optimal sequence is determined, for example, using optimization by a “greedy” algorithm, known to those skilled in the art. Step 130 of burst detection

[0068] The optimal sequence of states is finally analyzed to isolate possible bursts. A burst is then characterized by a minimum number of consecutive time steps in the abnormal state. This threshold number is predefined. It can be set to one, a burst being detected as soon as the train is in the abnormal state.

[0069] In addition to its duration, a burst can be characterized by an intensity. The intensity of a burst is, for example, given by the sum of the values ​​of the cost function fj (t) over the time step(s) associated with the burst. Scheduling Step 140

[0070] Once the previous steps have been completed for each train, the way in which the trains in the fleet are ordered depends on the quantity of interest.

[0071] If we consider, as in the previous example, the cumulative number of daily events, we can classify the trains in decreasing order of duration of the bursts detected for each train over the time interval considered. If two bursts have the same duration, then the trains are classified by decreasing intensity of these bursts.

[0072] If, on the other hand, we consider the number of daily events of a particular type, then the bursts are classified by intensity (knowing that the intensity takes into account in a certain way the duration of the burst).

[0073] Once the fleet trains are prioritized, only the IDs of the first X trains (e.g., X=3) are displayed on interface 16.

[0074] Advantageously, by selecting the identifier of a train, the operator accesses the Y most recurring events (for example, Y=5) affecting this train Benefits

[0075] This method helps maintenance operators to select the priority train for maintenance quickly and efficiently.

[0076] What is relevant for ranking trains relative to each other is not the number of event occurrences as such, but rather a significant change in the proportion of events generated by a train compared to the rest of the fleet, in a certain period of time and possibly for a certain type of events.

[0077] If an unusual and abrupt increase occurs for a particular train (in the total number of events, or just for a single type of event), then it can be considered that, for some reason, the behavior of this train is altered. It is important to flag this automatically because it may be symptomatic of something abnormal that should be rectified as quickly as possible.

Claims

1. A computer-implemented method (100) for ordering the vehicles of a fleet of vehicles according to a need for maintenance, the method including the steps of, for each vehicle in the fleet, - determining (110) a time series, the time series including, for each time step, an instantaneous value of at least one quantity of interest obtained from monitoring events acquired by means of a system for monitoring vehicles of the fleet of vehicles; and, - analyzing (120), over a predetermined time interval, the time series, taking into account the time series determined for all the vehicles of the fleet of vehicles, considering that at each time step, the state of a vehicle is either a "normal" state or an "abnormal" state requiring maintenance, so as to obtain an optimal sequence of states over the predetermined time interval, the method being characterized in that it comprises the additional steps of: - for each vehicle in the fleet, detecting (130) presence, if any, of one or a plurality of bursts in the optimum sequence of states; then, - ordering (140) a list of vehicles of the fleet of vehicles according to the properties of the bursts detected for each vehicle.

2. The method according to claim 1, wherein the quantity of interest is the total number of events or the total number of events of a particular type affecting the vehicle considered, acquired during a reference time window.

3. The method according to any of the preceding claims, wherein analyzing the time series includes an optimization of a cost function, associating a likelihood function and preferentially a transition function, so as to determine an instantaneous probability for each time step.

4. The method according to claim 3, wherein the instantaneous probability for a time step is compared with a reference probability calculated from the quantities of interest of all vehicles of the fleet of vehicles, so as to determine the optimum state of the vehicle considered at the time step considered.

5. The method according to any of the preceding claims, wherein detecting the presence of bursts in the optimum sequence of states consists of determining a number of consecutive time steps the vehicle spends in the "abnormal" state in the optimum sequence of states, and, when the number of time steps is greater than a predetermined threshold, considering same as a burst.

6. The method according to any of the preceding claims, wherein a burst is characterized by a duration and / or an intensity and ordering the list of vehicles of the fleet of vehicles according to the properties of the bursts detected for each vehicle is carried out depending on the duration and / or the intensity of said bursts.

7. The method according to any of the preceding claims, wherein the vehicle is a railway vehicle, such as a train, a subway, or a tram.

8. A computer program including software instructions which, when executed by a computer, implement a method according to any of the preceding claims.

9. A computer system (10), characterized in that same is suitable for implementing a method (100) according to any of claims 1 to 6, the computer system being suitable for accessing the content of a monitoring database (23), for reading monitoring events acquired by a monitoring system (20).

10. The system according to claim 9, including: - a module (32) for determining a time series of a quantity of interest; - a module (34) for analyzing the time series of a quantity of interest, for determining an optimum sequence of states; - a module (36) for detecting bursts in the optimum sequence of states; and, - a module (38) for scheduling the vehicles of a fleet of vehicles according to the properties of the bursts detected.