Monitoring system for at least one plurality of homogeneous devices of at least one railway vehicle

Abstract

A monitoring system for a plurality of identical devices (202, 205) of at least one railway vehicle (100) is disclosed, wherein the functional state of each device (202, 205) is represented by respective values ​​(x1, x2; y1, y2) of at least one operating quantity (X(t), Y(t), ...) common to the devices (202, 205). The monitoring system comprises: control means (201, 204) for acquiring values ​​of the operating quantities (X(t), Y(t), ...) at a series of acquisition times, each value representing the functional state of a respective device at the acquisition time; and diagnostic means (210, 210') for receiving the acquired values ​​at the acquisition times and detecting an operational anomaly or a maintenance request from the at least one device as a function of a comparison, for each acquisition time, of the received values ​​with a range of reference values ​​(Mx, My, ...) comprising reference values ​​determined as a function of at least two values ​​acquired at the acquisition times.

Description

Technical Field

[0001] The present invention generally relates to the field of railway vehicles. In particular, the present invention relates to a monitoring system for at least one plurality of homogeneous devices of at least one railway vehicle, particularly for safety applications in the field of railway transportation.

Background Art

[0002] A modern-designed railway vehicle 100 is shown in FIG. 1.

[0003] The railway train 100 is composed of a plurality of vehicles 101, and each vehicle is provided with at least one braking system 102, at least one access door system 103, and at least one air conditioning and heating unit 104.

[0004] Furthermore, the railway train 100 is provided with a plurality of pantographs 105 and a plurality of units 106 for generating, filtering, and drying compressed air.

[0005] The train 100 includes, without particular limitation, for example, a plurality of other systems or components not shown, such as bogie support units, toilets, etc.

[0006] The systems shown above are each controlled by one or more electronic control units not shown. The electronic control units typically communicate with each other via communication means of a serial nature, which is also not shown in FIG. 1.

[0007] Each electronic control unit acquires an electrical signal from a transducer configured to measure the operating quantity characteristics of the system to which it is related, and generates an electrical control signal for controlling the related system. Each control unit executes not only a control function but also a diagnostic function based on the analysis of these electrical signals.

[0008] The diagnostic function usually performs a comparison between the operating quantities (optionally pre-processed) measured or generated by each electronic control unit and the diagnostic comparison parameters pre-loaded in the non-volatile memory of the electronic control unit. The diagnostic parameters characterize the correct operating state, pre-alarm state, and alarm state of the system. The diagnostic parameters are defined at the system design stage.

[0009] When the diagnostic parameters are designed based purely on theoretical calculations or based on laboratory experiments, they often require further modification based on the operating conditions of the train.

[0010] Document WO2016041756 claims various aspects of a method for performing a diagnostic function on a system and / or components mounted on a railway vehicle, in particular a method for updating diagnostic algorithms and / or related parameters mounted on the train from a remote server.

[0011] In daily life, railway operators are hostile to remote access to the operating parameters or diagnostic parameters of on-vehicle systems for reasons of operational security and cyber security.

Summary of the Invention

Problems to be Solved by the Invention

[0012] Therefore, an object of the present invention is to provide a monitoring system for at least one homogeneous device of at least one railway vehicle that is independent of the need for remote updates. Group of the same kind of device.

[0013] The above and other objects and advantages, according to one aspect of the present invention, at least one of at least one railway vehicle having the features defined in claim 1 GroupIt is achieved by a monitoring system for the same type of device. Preferred embodiments of the present invention are defined in the dependent claims, and the content thereof should be understood as an essential part of this specification.

Brief Description of the Drawings

[0014] Next, the functional and structural features of some preferred embodiments of the electronic control system for emergency and service brakes according to the present invention will be described. Refer to the accompanying drawings.

Figure 1

Figure 2

Figure 3a

Figure 3b

Figure 4

Figure 5

Modes for Carrying Out the Invention

[0015] Before detailing multiple embodiments of the present invention, it should be made clear that the present invention is not limited in its application to the structural details and component configurations described in the following specification or shown in the figures. The present invention can take other embodiments and can be actually implemented or constructed in various different ways. It should also be understood that the expressions and technical terms are for the purpose of explanation and should not be construed as limitations. "Include", "comprise", or variations thereof should be understood to include the elements described below and their equivalents, as well as additional elements and their equivalents.

[0016] Hereinafter, at least one of the at least one rail car 101 Group An embodiment of a monitoring system for the same devices 202, 205 is described.

[0017] FIG. 2 illustrates at least one of at least one rail car 101. Group 2 shows an embodiment of a monitoring system for homogeneous devices 202, 205.

[0018] The functional state of each device 202, 205 is represented by a respective value x1, x2; y1, y2 of at least one operating quantity X(t), Y(t), . . . that is common to the multiple devices 202, 205.

[0019] The monitoring device acquires information at a predetermined acquisition period T i A series of acquisition time points T determined according to i , T i+1 , T i+2 , ..., a plurality of values ​​x1(t i , t i+1 , t i+2 , ...), x2(t i , t i+1 , t i+2 , ...);y1(t i , t i+1 , t i+2 , ...), y2(t i , t i+1 , t i+2 , ...) are obtained. i , t i+1 , t i+2 , ...), x2(t i , t i+1 , t i+2 , ...);y1(t i , t i+1 , t i+2 , ...), y2(t i , t i+1 , t i+2,...)) each value is at the acquisition time point T i , T i+1 , T i+2 ,... represents the functional state of each of at least one Group of the devices 202, 205.

[0020] Further, the monitoring device, via the communication means 211, at the acquisition time, a plurality of values x1(t i , t i+1 , t i+2 ,...), x2(t i , t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...), y2(t i , t i+1 , t i+2 ,...) is configured to receive diagnostic means 210, 210'. Also, the diagnostic means 210, 210' receive a plurality of values x1(t i , t i+1 , t i+2 ,...), x2(t i , t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...), y2(t i , t i+1 , t i+2 ,...) as a function of at least one range of reference values, and is configured to detect at least one operation abnormality or maintenance request of at least one of the devices.

[0021] The range of reference values includes reference values Mx, My,... and proximity values ΔMx, ΔMy,... related to the reference values Mx, My,...

[0022] The reference values Mx, My,... are obtained by the diagnostic means 210, 210' at the acquisition time, a plurality of values x1(t i , t i+1 , t i+2 ,...), x2(t i, t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...)、y2(t i , t i+1 , t i+2 ,...) is determined as a function of at least two values.

[0023] The diagnostic means can be set to determine reference values Mx, My,... at each acquisition time point.

[0024] The diagnostic means can be configured to determine neighborhood values ΔMx, ΔMy,... at each acquisition time point.

[0025] The specifications of the same type of devices 202, 205 of at least one railway vehicle 101 will be described in detail below. "A plurality of the same type of devices" refers to a set of at least two devices 202, 205 that perform the same function and are mounted on at least one railway vehicle 101 or railway train 100.

[0026] At least one control means 201, 204 is configured to control the same type of devices 202, 205. There may be only one control means, which controls all the same type of devices, or rather, there may be two or more control means, each of which controls at least one of the respective same type of devices. In the second case, the control means are also of the same type as each other.

[0027] "Control means" can refer to, for example, a control device, a controller, a processor, a control module, a control unit, etc.

[0028] As a non-limiting example, a plurality of the same type of devices 202 can be two devices of at least two door systems 103 of the railway vehicle 101 or the train 100, and the movement of the doors is controlled by the control means 201. In a further non-limiting example, a plurality of the same type of devices 205 can be at least two devices of the electro-pneumatic brake system 102 of the railway vehicle 101 or the train 100 and / or related brake actuators.

[0029] Referring to FIG. 2, the control means 201, 204 can exchange a plurality of signals 203, 206 with the associated homogeneous devices 202 and / or 205, respectively.

[0030] The plurality of signals 203, 206 may include input signals sent from transducers configured to measure the amount of operation characteristic of the associated device, and may include output signals configured to control the associated device.

[0031] As a non-limiting example, when the plurality of homogeneous devices include the devices of the door system 103, the plurality of signals 203, 206 may constitute input signals including the rotational speed and position signals of the electric motor of the door and the door movement end signal, and may constitute output signals including signals for controlling the power of the electric motor for moving the door. Therefore, examples of the plurality of homogeneous devices of the door system may be, for example, a plurality of electric motors for moving the door, or a travel end device.

[0032] As a further non-limiting example, when the plurality of homogeneous devices are the devices of the brake system 102, the signals 203, 206 may constitute input signals including the axle speed signal and signals from various linear or binary pressure sensors configured to monitor the air pressure at various points of the associated brake system, and may constitute output signals including control signals for the pneumatic solenoid valves used to control the pressure. Therefore, examples of the plurality of homogeneous devices of the brake system 102 may be, for example, a plurality of speed sensors, a plurality of linear or binary pressure sensors, or a plurality of pneumatic solenoid valves.

[0033] In one embodiment, at least one operating quantity X(t), Y(t), ··· constitutes a value directly obtained or generated by at least one control means 201, 204, or a value / measurement input manually or remotely from the outside via, for example, a maintenance software tool. In a further embodiment, in order to prepare data for diagnostic analysis, at least one operating quantity X(t), Y(t), ··· may constitute a value obtained from preprocessing performed on at least one operating quantity directly obtained or generated by at least one control means 201, 204.

[0034] In other words, the control means 201, 204 may be configured to execute preprocessing functions 208, 207 on the quantities represented by the signals 203, 206.

[0035] When a plurality of homogeneous devices belong to the brake system 102, the preprocessing functions 207, 208 may include a brake pressure application counter in the brake cylinder over a specific period, a travel distance in a predetermined period obtained by time-integrating the speed signal from the axle, and the number of operations of each pneumatic solenoid valve in a predetermined period.

[0036] When a plurality of homogeneous devices belong to the door system 103, the preprocessing functions 207, 208 may include an identifier of the maximum peak value of the current to the door operating motor, or a counter of the number of door openings of each door in a predetermined period, or the total electrical energy consumed by each electric motor to move the door in a predetermined period.

[0037] The control means 201, 204 may share, via the communication means 211, the quantities represented by the signals 203, 206, which may be preprocessed in some cases, by the preprocessing functions 208, 207 respectively. The communication means 211 may be of a wired type or a wireless type.

[0038] The definition of the operating quantities X(t), Y(t),... common to the plurality of devices 202, 205 will be described in detail below.

[0039] The "common operation amounts X(t), Y(t),... " have the same meaning and the same dimension, and refer to amounts related to at least two devices that form part of a plurality of homogeneous devices 202, 205.

[0040] A first non - limiting example of the operation amount X(t) is represented by the current absorbed by each electric motor of each door system 103 of the same railway vehicle 101 or train 100. x1(t) is the current absorbed by the first electric motor related to the first homogeneous device, i.e., the first door 103, x2(t) is the current absorbed by the second electric motor related to the second homogeneous device, i.e., the second door 103, and xn(t) is the current absorbed by the nth electric motor related to the nth homogeneous device, i.e., the nth door 103.

[0041] A second non - limiting example of the operation amount Y(t), which is different from X(y), is represented by a set of pulse counters for actuating the brake solenoid valves belonging to each brake system 102 of the same vehicle 101 or train 100. y1(t) is the value of the actuation pulse counter of the brake solenoid valve belonging to the first homogeneous device, i.e., the first brake system 102, y2(t) is the value of the actuation pulse counter of the brake solenoid valve belonging to the second homogeneous device, i.e., the second brake system 102, and yn(t) is the value of the actuation pulse counter of the brake solenoid valve belonging to the nth homogeneous device, i.e., the nth brake system 102.

[0042] A third non-limiting example of the amount of operation Z(t) different from X(y) is represented by a set of thickness values of brake pads (or brake shoes) belonging to each brake unit 102 of the same vehicle 101 or train 100, and the thickness values of the brake pads (or brake shoes) are indicators of the wear level of the friction material. In this example, z1(t) is the thickness of the brake pad belonging to the first brake unit 102, z2(t) is the thickness of the brake pad (or brake shoe) belonging to the second similar device, i.e., the second brake unit 102, and zn(t) is the thickness of the brake pad (or brake shoe) belonging to the nth similar device, i.e., the nth brake unit 102.

[0043] One or more diagnostic means 210 may receive the amounts of operation X(t), Y(t),... acquired or generated by one or more control means 201 via the communication means 211 and optionally pre-processed by the pre-processing functions 208, 207.

[0044] One or more diagnostic means 210 may be centralized in the centralized diagnostic unit 213, or one or more diagnostic means 210 may be distributed among a plurality of control means 201, 204 assuming the form shown as 210' in FIG. 2. In other words, the diagnostic means 210 may be an algorithm executed inside the centralized diagnostic unit 213, or the diagnostic means 210' may be one or more execution algorithms distributed among a plurality of control means 201, 204.

[0045] Details of the functions of the diagnostic means are provided below.

[0046] The diagnostic means 210, 210' are a plurality of values x1(t i , t i+1 , t i+2 ,...) and x2(t i , t i+1 , t i+2,...) in order to identify deviations in the behavior of one or more of the values, a plurality of values x1(t belonging to the operation amount X(t) i , t i+1 , t i+2 ,...) and x2(t i , t i+1 , t i+2 ,...) are configured to perform analysis operations.

[0047] When component failure, abnormal operation, or life evaluation occurs, the diagnostic means 210, 210' are configured to activate an alarm display, and in some cases, are encoded to indicate its source and nature, and in some cases, may be associated with one or more items of diagnostic information characteristic of the nature of the alarm.

[0048] After the generation of the alarm, the alarm and any related values are This by the diagnostic means that generated the alarm, diagnosis via the wireless transmission means associated with the diagnostic means, or is transmitted to the ground center for data collection and alarm management using the unit 214 mounted on the train (this unit 214 can process the received alarms and transmit these alarms to the driver and / or the TCMS (Train Control and Monitoring System), and this TCMS can further transmit the above alarms and related values to the ground collection center). .

[0049] The unit 214 can be an independent unit, or can coincide with one or more of the control means 201, 204, or can coincide with the TCMS system.

[0050] Some of the diagnostic means 210, 210' can be operable on the same railway vehicle 101 or train 100, and each of them can be configured to analyze one or more common operation amounts X(t), Y(t),...

[0051] In one embodiment, the diagnostic means 210, 210' are a plurality of values x1(t acquired at the acquisition time i , t i+1 , t i+2 ,...) and x2(t i , t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...) and y2(t i , ti+1 and i+2 be configured to determine reference values Mx, My, ... by taking the average of at least two values of t, t, ...

[0052] In this way, the analysis of the operation amount Y(t) can be based on the comparison of each value of a plurality of values y1(t, t, t, ...), y2(t, t, t, ...) associated with the operation amount Y(t) with the average value My of the operation amount Y(t). i and i+1 and i+2 and i and i+1 and i+2 be obtained based on the comparison of each value of a plurality of values y1(t, t, t, ...), y2(t, t, t, ...) associated with the operation amount Y(t) with the average value My of the operation amount Y(t).

[0053] Figure 3a shows curves y1(t), y2(t),... y(t) representing various values of the operation amount Y(t) as a function of time. n (t).

[0054] Figure 3b shows x1(t), x2(t),... x(t) representing various values of the operation amount X(t) as a function of time. n (t).

[0055] As described above, the diagnostic means can be set to determine the reference values Mx, My,... according to each acquisition time point, that is, the acquisition period T.

[0056] The diagnostic means 210, 210' can calculate the reference value by taking the average value My of the values y1(t, t, t, ...), y2(t, t, t, ...),... yn(t, t, t, ...), and in some cases, by filtering with an appropriate time window Δt for removing statistical noise. i and i+1 and i+2 and i and i+1 and i+2 and i and i+1 and i+2 and

[0057] Subsequently, the diagnostic means 210, 210' compare each value y1(t, t, t, ...), y2(t, t, t, ...),... with the reference value My, and in some cases, by filtering with an appropriate time window Δt for removing statistical noise, the reference value can be calculated. i and i+1 and i+2 andi , t i+1 , t i+2 ,...),... yn(t i , t i+1 , t i+2 ,...) can be configured to be compared with a reference value My, i.e., an average value.

[0058] A plurality of values y1(t i , t i+1 , t i+2 ,...), y2(t i , t i+1 , t i+2 ,...),... yn(t i , t i+1 , t i+2 ,...) When the difference Δy between the individual value of each value and the reference value My, i.e., the average value, exceeds a predetermined neighborhood value ΔM indicating a malfunction state, the diagnostic means is configured to issue an alarm signal.

[0059] Although not limited, the neighborhood value ΔM can be an allowable range centered on the average value My.

[0060] Although not limited, the neighborhood value ΔM can be a certain neighborhood.

[0061] Although not limited, the neighborhood value ΔM can be represented by a reference value for comparison, rather the average value M, and in some cases, a percentage value with positive and negative double signs. This solution offers the advantage of automatically adapting to full-scale temporal variations.

[0062] Although not limited, the neighborhood value ΔM can be expressed as a function of the variance of at least one operating quantity y1(t), y2(t),... yn(t). This solution offers the advantage of automatically adapting to data variability.

[0063] Although not limited, the neighborhood value ΔM can be represented as a set of one or more of the above solutions and a set of times. This solution offers the advantage of automatically adapting not only to the nature of the data but also to time dependencies such as predicted variations due to aging.

[0064] One example is a counter for the number of actuations of a specific solenoid valve present in all homogeneous braking systems 102, for example the braking system associated with a motor bogie.

[0065] The number of actuations of the specific solenoid valve is expected to be on average the same for all braking systems 102 belonging to the same type of bogie. Potential short-term variations are masked by numerical filtering using a time window Δt. A significant deviation from the average value of the cumulative pulses of the pneumatic solenoid valve can indicate, for example, a pneumatic leak in the braking system associated with the pneumatic solenoid valve.

[0066] When a predetermined first quantity of motion X(t) to be analyzed is affected in different ways by local inhomogeneous situations of the train, it is advantageous to identify at least one second quantity of motion Y(t) associated with the local inhomogeneous situations of the train and to normalize the first quantity of motion X(t) to be analyzed according to the second quantity of motion Y(t).

[0067] The diagnostic means can be configured to normalize the first quantity of motion X(t) at each acquisition time.

[0068] As a non-limiting example, the braking system 102 associated with a load bogie can stress its pneumatic solenoid valve more than the braking system 102 associated with a motor bogie, since the braking system 102 associated with a motor bogie mainly uses the traction motor for braking.

[0069] In such a case, the curve for the solenoid valve associated with the load bogie deviates from the curve for the solenoid valve associated with the motor bogie in a short time, so comparing the actuation pulse counters of the specific pneumatic solenoid valve in all braking systems 102 has no meaning regardless of whether they are associated with a load bogie or a motor bogie.

[0070] To solve this drawback, by way of non-limiting example, each control unit means 201, 204 associated with the braking system 102 obtains the maximum pressure value to the brake cylinder generated thereby during each braking, continuously and locally calculates its average value P (note: in the original text, "P with overbar"), and normalizes the value generated by the local counter of the actuation pulses of the same type of solenoid valve according to the average value P, that is, divides the value accumulated by the local counter by the average value P to obtain the number of actuation times per [bar] as a result, which is convenient.

[0071] In this way, the normalized values of the actuation pulse counters of the pneumatic solenoid valves generate compatible values so that they can be reused and compared.

[0072] Further embodiments are always substantially based on a statistical approach. For example, the diagnostic means 210, 210' can be configured as follows. - Calculate the first average value by calculating the average values of a plurality of values x1(t i , t i+1 , t i+2 ,...) and x2(t i , t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...) and y2(t i , t i+1 , t i+2 ,...) obtained at the acquisition time. - Calculate the average values of a plurality of values x1(t i , t i+1 , t i+2 ,...) and x2(t i , t i+1 , t i+2 ,...); y1(t i , t i+1 , t i+2 ,...) and y2(t i , t i+1 , t i+2Identify a value that is at least a predetermined threshold greater than or less than the first average value among (...). - A plurality of values x1(t i 、t i+1 、t i+2 、...)、x2(t i 、t i+1 、t i+2 、...); y1(t i 、t i+1 、t i+2 、...)、y2(t i 、t i+1 、t i+2 Determine reference values Mx, My,... by calculating the average value of values that are not identified as being greater than or less than at least one predetermined threshold value than the first average value among (...).

[0073] In other words, values that are ±S away from the average value My can be identified, and the average value M obtained by subtracting the identified values can be recalculated. Therefore, the outlier value ym(t) that exists beyond the distance ±S from the average value M does not affect the definition of the reference value for calculating the vicinity of the change amount Δy, making the estimated change amount Δy more reliable.

[0074] When the observed operation amount Y(t) depends on a plurality of operation amounts H(t), K(t),... J(t), the normalization process becomes complicated.

[0075] When the observed operation amount Y(t) depends on a plurality of operation amounts H(t), K(t),... J(t), a neural network can be used as a self-learning and self-calibration algorithm mounted on the train. The neural network is set to continuously calculate the reference value for comparing a plurality of values and can advantageously avoid the requirement for remote update.

[0076] In FIG. 4, for example, the circular buffer 401 has a size of R records, and every R-th record is configured to host, for example, one or more quantities of operations X(t), Y(t),... Z(t) to be monitored, and the behavior of the one or more quantities of operations H(t), K(t),... J(t) depends on the behavior of the one or more quantities of operations X(t), Y(t),... Z(t) to be monitored.

[0077] FIG. 5 shows in more detail a preferred format of the records included in the circular buffer 401.

[0078] In a non-limiting example of FIG. 5, the first record is generated at time point aT by the electronic control unit 1 and transmitted to the diagnostic means 210, 210'. The first record includes x1(aT), y1(aT), z1(aT), h1(aT), k1(aT), j1(aT) related to the control means 1.

[0079] The second record is generated at time point aT by the control means 3 and transmitted to the diagnostic means 210, 210'. The second record includes x3(aT), y3(aT), z3(aT), h3(aT), k3(aT), j3(aT) related to the control means 3.

[0080] The third record is generated at time point aT by the control means n and transmitted to the diagnostic means 210, 210'. The third record includes xn(aT), yn(aT), zn(aT), hn(aT), kn(aT), jn(aT) related to the control unit means n.

[0081] The r1-th record is generated at time point bT > aT by the control means 2 and transmitted to the diagnostic means 210, 210'. The r1-th record includes x2(bT), y2(bT), z2(bT), h2(bT), k2(bT), j2(bT) related to the control means 2.

[0082] The r-th record is generated by the control means 1 at the time point bT > aT and transmitted to the diagnostic means 210, 210'. The r-th record includes values x1(bT), y1(bT), z1(bT), h1(bT), k1(bT), j1(bT) related to the control means 1.

[0083] R must be of a size capable of storing the number of records considered necessary to perform at least effective learning of the neural network.

[0084] The neural network 402, whose configuration is prior art, can be set to receive, at the input, one or more amounts of motion H(t), K(t),... J(t) and, at the output, generate one or more amounts of image X'(t), Y'(t),... Z'(t) related to the respective amounts of the same kind X(t), Y(t),... Z(t) to be monitored.

[0085] There may be a learning algorithm 403 acting on the internal coefficients of the neural network 402 by receiving the amounts of motion X(t), Y(t),... Z(t) and the amounts of image X'(t), Y'(t),... Z'(t) to be monitored and minimizing the error existing between each amount of motion X(t), Y(t),... Z(t) to be monitored and the related amounts of image motion X'(t), Y'(t),... Z'(t).

[0086] The learning algorithm can be, but is not limited to, the backpropagation algorithm.

[0087] The diagnostic means 210, 210' can be provided to receive records from the control means 201, 204 via the communication means 211 and store them in a circular buffer.

[0088] If necessary, the transmission of records from the control means 201, 204 to the diagnostic means 210, 210' can be performed synchronously to ensure the temporal consistency of the records transmitted.

[0089] The time value at the time of transmission of a single record by the control means 201, 204 can be one of the operation amounts H(t), K(t),... J(t).

[0090] When the buffer accumulates a record amount Q ≤ R, the diagnostic means 210, 210' can start the learning process of the neural network 402 by activating the learning algorithm 403 that uses the record set Q in order to execute the learning operation.

[0091] The parameter Q can be defined based on the experience related to how much time is required to accumulate a considerable number of records useful for the reliability learning of the neural network 402, not limited to the design stage of the diagnostic means 210, 210'.

[0092] As a non-limiting example, important information for diagnostic use can be stored at a one-hour interval. Thus, for the first six months of service, Q = 2,880U (where U is the number of control means related to the operation amounts X(t), Y(t),... Z(t), H(t), K(t),... J(t)). Therefore, learning based on the service history of six months of service and 16 hours of daily service requires 2,880 records transmitted by each of the control means 201, 204 related to the operation amounts X(t), Y(t),... Z(t), H(t), K(t),... J(t).

[0093] During the accumulation period of the record amount Q that occurs when the vehicle is tested and operated, the devices related to the operation amounts y1, y2,... yn to be monitored must therefore perform their functions correctly, that is, do not operate abnormally or fail.

[0094] When an abnormal operation state related to the value yn(t) occurs during the collection of Q records, an inaccurate data set is generated, and as a result, the learning of the neural network regarding the operation amount Y(t) becomes inaccurate.

[0095] As an example, this drawback can be overcome by evaluating a sample of the quantity to be monitored during the accumulation stage of the Q records using the statistical method described above, eliminating the outliers from the learning process, and generating an alarm report.

[0096] After learning, the image quantities X'(t), Y'(t),... Z'(t) are considered to reliably represent the actual behavior of the associated operation quantities X(t), Y(t),... Z(t) to be monitored.

[0097] When learning is performed, for each new nth incoming record, the diagnostic means 210, 210' can be set to directly pass a plurality of incoming values of the operation quantities H(t), K(t),... J(t) to the neural network 402, and compare each operation quantity X(t), Y(t),... Z(t) to be monitored with the respective values of the associated operation image quantities X'(t), Y'(t),... Z'(t), and activate a comparison algorithm 404 configured to generate associated status reports W1, W2,... Wn.

[0098] The comparison algorithm 404 operates, for example, by verifying that the difference between each value of the operation quantities X(t), Y(t),... Z(t) to be monitored and the respective values of the associated image quantities X'(t), Y'(t),... Z'(t) does not exceed predetermined values ΔM1, ΔM2,... ΔMn respectively associated with X(t), Y(t),... Z(t) to be monitored. The predetermined values ΔM1, ΔM2,... ΔMn are stored in a non-volatile memory section associated with the diagnostic means 210, 210'.

[0099] Also, the comparison algorithm 404 can operate, for example, by verifying that the derivative value of the difference between each value of the operation quantities X(t), Y(t),... Z(t) to be monitored and the respective values of the associated image quantities X'(t), Y'(t),... Z'(t) does not exceed predetermined values ΔD1, ΔD2,... ΔDn respectively associated with X(t), Y(t),... Z(t) to be monitored. The predetermined values ΔD1, ΔD2,... ΔDn are stored in a non-volatile memory section associated with the diagnostic means 210, 210'.

[0100] The diagnostic information W1, W2, ... Wn can be, for example, at least two value ranges indicating the severity of the state of the associated observed operation amount in at least two levels, and although not limited thereto, for example, PASS and FAIL can be assumed.

[0101] The diagnostic means 210, 210' can be configured to take into account the environmental and / or operational variations of the train associated with the life of the train.

[0102] In this case, the diagnostic means 210, 210' continuously store the received records in the circular buffer 401, and periodically activate the learning algorithm 403 to update the internal coefficients of the neural network 402 according to the new operating conditions, and can advantageously and accurately maintain the image quantities X'(t), Y'(t),... Z'(t) continuously in time without requiring external correction or recalibration.

[0103] A further advantage of using the solution described in FIG. 4 is represented by the fact that a neural network of appropriate size comes to significantly represent the history prior to the last data stored in the circular buffer 401. By being able to download the coefficients of the neural network 402 during the holding stage of the related devices 202, 205, the history of the related devices can be obtained with good approximation.

[0104] The limitations of the embodiments disclosed above constitute the presently preferred embodiments, but thereby can be changed without departing from the broader scope defined in the main claims.

[0105] Therefore, the achieved advantage lies in obtaining a solution that solves the non-productive cases of known monitoring systems through a solution with reduced complexity and cost.

[0106] Various aspects and embodiments of a procedure for setting up a monitoring system according to the present invention have been described. It is understood that each embodiment can be combined with any other embodiment. Further, the present invention is not limited to the described embodiments and can be modified within the scope defined by the appended claims.

Claims

1. A monitoring system for at least one group of devices (202, 205) that perform the same function in at least one railway vehicle (101), wherein the functional state of each device (202, 205) is represented by respective values (x1, x2; y1, y2) of at least one operating quantity (X(t), Y(t),...) common to the group of devices (202, 205), the monitoring system being, at a series of acquisition time points (t ｉ , t ｉ , t ｉ＋１ , t ｉ＋２ ,...) determined based on a predetermined acquisition period (T ｉ , t ｉ＋１ , t ｉ＋２ ,...), a plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) of the at least one operating quantity (X(t), Y(t),...) are acquired, and each value of the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) represents the functional state of each device of the at least one group of devices (202, 205) at one acquisition time point (t ｉ , t ｉ＋１ , t ｉ＋２ ,...), and via communication means (211), a plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) and receives the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) at each acquisition time point (t ｉ , t ｉ＋１ , t ｉ＋２ ,...)), as a comparison function for each acquisition time point (t ｉ＋２ ,...)), diagnostic means (210, 210') configured to detect an operation abnormality or a maintenance requirement by at least one device among the at least one group of devices (202, 205); comprising the range of the reference value includes a reference value (Mx, My,...) and neighboring values (ΔMx, ΔMy,...) related to the reference value (Mx, My,...), the reference value (Mx, My,...) is determined as a function of at least two values among the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) by the diagnostic means (210, 210'), and a monitoring system characterized by this.

2. The diagnostic means (210, 210') determines that when at least one of the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) is not within the range of the reference value, the monitoring system according to claim 1, configured to determine the presence of a failure or maintenance requirement by at least one of the at least one group of devices (202, 205).

3. The monitoring system according to claim 1 or claim 2, wherein the diagnostic means is configured to determine the reference value (Mx, My,...) at each acquisition time point.

4. The diagnostic means (210, 210') determines the reference value (Mx, My,...) by calculating the average between at least two of the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) acquired at one acquisition time point, the monitoring system according to any one of claims 1 to 3.

5. The diagnostic means (210, 210') The plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) x2(t ｉ, t ｉ＋１ , t ｉ＋２ ,..., ); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,..., ), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) by calculating the average of the said values, a first average value is determined, the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) that are larger or smaller than the first average value by a predetermined threshold value are specified, the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...), y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...)) that are larger or smaller than the first average value by a predetermined threshold value are not specified, and by calculating the average value of the values, through the determination of a second average value, the reference values (Mx, My,...) are determined, the monitoring system according to any one of claims 1 to 3.

6. The diagnostic means is configured to determine the proximity values (ΔMx, ΔMy,...) as a function of the percentage values of the reference values (Mx, My,...), the monitoring system according to any one of claims 1 to 5.

7. The monitoring system according to any one of claims 1 to 5, wherein the diagnostic means is configured to determine the proximity values (ΔMx, ΔMy,...) as a function of the variance of the at least one operation amount (X(t), Y(t),...).

8. The at least one operation amount includes at least two operation amounts (X(t), Y(t),...), that is, at least a first operation amount (X(t)) and a second operation amount (Y(t)), The monitoring system according to any one of claims 1 to 7, wherein the diagnostic means is configured to normalize the first operation amount (X(t)) as a function of the at least one second operation amount (Y(t)).

9. The monitoring system according to any one of claims 1 to 8, wherein the diagnostic means (210, 210') is configured to determine the reference values (Mx, My,...) by a neural network (402).

10. The diagnostic means is configured to perform a first learning of the neural network (402) by a learning algorithm (403) associated with the neural network (402), The first learning is performed using a predetermined amount Q of the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) and x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) and y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) previously acquired by the at least one control means (201, 204). The monitoring system according to claim 9.

11. After performing the first learning, the diagnostic means is configured to perform further learning by the learning algorithm (403) in the time continuity of the learning time determined according to a predetermined learning period. wherein the learning algorithm uses a predetermined amount Q of the plurality of values (x1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) x2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...); y1(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) y2(t ｉ , t ｉ＋１ , t ｉ＋２ ,...) obtained by the at least one control means (201, 204) between the current sampling time and the previous sampling time. The monitoring system according to claim 10.

12. The monitoring system according to claim 10, wherein the learning algorithm (403) is a backpropagation algorithm.

13. The monitoring system according to any one of claims 1 to 12, wherein the diagnostic means is integrated into an electronic diagnostic unit (213).

14. The monitoring system according to any one of claims 1 to 13, wherein the diagnostic means is distributed over a plurality of control means (201, 204).

15. The monitoring system according to any one of claims 1 to 14, wherein the at least one operation amount (X(t), Y(t),...) includes a value directly obtained or generated by the at least one control means (201, 204) or a value manually or remotely input from the outside.

16. The monitoring system according to any one of claims 1 to 14, wherein the at least one operation amount (X(t), Y(t),...) includes a value obtained from preprocessing executed by the at least one control means (201, 204) with respect to the at least one operation amount directly acquired or generated by the at least one control means (201, 204).

17. The monitoring system according to any one of claims 1 to 16, wherein the diagnostic means is configured to provide a diagnostic index divided according to at least two severities.

18. The monitoring system according to any one of claims 1 to 17, wherein the diagnostic means is configured to transmit a diagnostic instruction to a further unit or module mounted on a train including the at least one railway vehicle.

19. The monitoring system according to any one of claims 1 to 18, wherein the diagnostic means is configured to transmit a diagnostic instruction via a wireless connection.

20. The monitoring system according to any one of claims 1 to 19, wherein the communication means (211) is a serial communication system or a wireless communication system.

Images