RAILWAY, SUBWAY OR TRAMWAY NETWORK"

DESCRIPTION :

The present invention relates to an apparatus and a method for monitoring a switch of a railway, subway or tramway network, particularly a switch having a remotely controlled electromechanical drive .

As is known, switches are very complex units , especially those of high-speed railway lines , because they must be manufactured with very strict tolerances in order to reduce the mechanical stress caused by trains running thereon at high speed . However, the need for keeping mechanical tolerances as small as possible makes such switches more subj ect to mal functions that might likely cause accidents , e . g . a derailment caused by a switch stuck at hal f travel , a point blade stuck because of a broken tie rod, or the like .

In order to prevent such mal functions from occurring, switches must be inspected by at least one skilled operator walking along a railway, subway or tramway line when it is not in operation ( e . g . at night ) , so that the operator can reach the switches without running any personal risks and conduct manoeuvring tests on said switches without j eopardi zing the network' s safety .

Nevertheless , such an approach is unsatis factory as far as fault prevention is concerned, because too much time pas ses between inspections ; as a matter of fact , the operator walking along the l ine very often f inds that some switches are already faulty, resulting in the traf fic along the l ine having to be temporarily stopped until all the necessary repair work has been carried out on the switches .

The present invention aims at solving these and other problems by providing a method for monitoring a switch o f a railway, subway or tramway network .

Furthermore , the present invention aims at solving these and other problems by providing also an apparatus for monitoring a switch of a railway, subway or tramway network .

The basic idea of the present invention is to use a neural network for determining a monitoring datum which represents an operating state of a switch, wherein said neural network is trained in a manner such that , when inputting to said neural network at least one training datum representing a quantity concerning a manoeuvre of a second switch in an abnormal operating condition, said neural network will output a second monitoring datum indicating the occurrence of said abnormal operating condition, and wherein said at least one monitoring datum is transmitted to a supervision computer via transmission means .

This makes it possible to determine the occurrence of abnormal operating conditions as soon as the switch is first manoeuvred, without requiring the presence of an operator along the line . It will thus be possible to immediately take the necessary steps ( e . g . to avoid actuating the switch) and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur .

Further advantageous features of the present invention will be set out in the appended claims .

These features as well as further advantages of the present invention will become more apparent in the l ight of the following description of a preferred embodiment thereof as shown in the annexed drawings , which are provided herein merely by way o f non-limiting example , wherein :

- Fig . 1 shows a railway, subway or tramway monitoring system comprising an apparatus for monitoring a switch of a railway, subway or tramway network according to the invention;

- Fig . 2 shows a block diagram of the apparatus of Fig . 1 ;

- Fig . 3 shows a flow chart of a method for monitoring a switch of a railway, subway or tramway network according to the invention .

In this description, any reference to "an embodiment" will indicate that a particular configuration, structure or feature is comprised in at least one embodiment of the invention . Therefore , expressions such as " in an embodiment" and the like , which may be found in di f ferent parts of this description, will not necessarily refer to the same embodiment . Moreover, any particular configuration, structure or feature may be combined as deemed appropriate in one or more embodiments . The references below are therefore used only for simplicity' s sake , and shall not limit the protection scope or extension of the various embodiments .

With reference to Fig . 1 , the following will describe a system S for monitoring a switch D of a railway, subway or tramway network R, wherein said switch comprises a left-hand point blade AS , a right-hand point blade AD, a left-hand stock rail CAS whereon said left-hand point blade AS abuts , a right-hand stock rail CAD whereon said right-hand point blade AD abuts , at least one tie rod TR that mutually constrains said point blades AS , AD and a point machine CM, a frame T constrained to the le ft-hand stock rail CAS and/or to the right-hand stock rail CAD and supporting said switch D, wherein said point machine CM contains at least one actuator configured for actuating said at least one tie rod TR upon receiving a command (preferably remotely transmitted) , thereby moving the point blades AS , AD of the switch D .

The system S comprises the following elements :

- an apparatus 1 for monitoring said switch D ( e . g . an embedded device , an industrial PC, a development board, or the like ) , preferably positioned inside or near a point machine of said switch D;

- sensor means (which will be further described below) configured for detecting at least one operating datum representing a quantity concerning a manoeuvre of said switch D when said switch D is in an operating condition, i.e. when moving or attempting to move the point blades AS, AD;

- a supervision computer 3 (e.g. a remote physical server and/or a remote and/or "in-cloud" virtualized server) in communication with said apparatus 1 and configured for receiving a signal SM generated by said apparatus 1, and for notifying an operator when a monitoring datum is encoded in said signal SM which indicates that the apparatus 1 has detected an abnormal operating condition of the switch D.

The switch D is considered to be in an abnormal operating condition when it is faulty or when it is in an operating condition in which, while it can still manoeuvre the point blades AS, AD, it cannot do so optimally, and such an operating condition may lead over time to failure of said switch D (e.g. breakage of said at least one tie rod TR, or damage to one or more windings of a motor contained in the point machine CM, etc.) , thus preventing traffic on at least a portion of the line.

It must be pointed out that, although the switch represented in Fig. 1 is a simple-type one, the invention is also applicable to any type of switch comprising an electromechanical drive, without however departing from the teachings of the present invention .

Also with reference to Fig. 2, the apparatus 1 preferably comprises the following components:

- execution means 11, e.g. one or more CPUs, GPUs, DSPs, FPGAs and/or the like, which preferably implement, whether in hardware and/or software form, a neural network trained by inputting to said neural network at least one training datum representing a quantity concerning a (second) manoeuvre of a second switch in an abnormal operating condition, and forced to output a second monitoring datum indicating the occurrence of said abnormal operating condition; - volatile memory means 12, e.g. a random access memory RAM, in signal communication with the execution means 11. When the apparatus 1 is in an operating condition, said volatile memory means 12 store the monitoring data acquired by the sensor means ;

- non-volatile memory means 13, preferably one or more magnetic disks (hard disks) or a Flash memory or another type of memory, in signal communication with the execution means 11 and with the volatile memory means 12, and wherein said nonvolatile memory means 13 contain at least one set of instructions implementing a method for monitoring the switch D of a railway, subway or tramway network according to the invention. Moreover, said non-volatile memory means 13 may also contain information that makes it possible to configure and/or operate the neural network (e.g. a set of internal weights, numbers of levels in the various layers, or the like) ;

- input and/or output (I/O) means 14 which can be connected to the sensor means and which are configured for acquiring at least one operating datum which represents a quantity concerning a manoeuvre of said switch, and which may include, as will be described below, the intensity and/or voltage of an electric current absorbed by the actuator, the acceleration of a point blade, etc. These input/output means 14 may comprise, for example, a USB™, IEEE 1394, RS232, RS485, IEEE 1284 adapter or the like;

- transmission means 15, preferably wired and/or wireless ones (e.g. a network interface such as 803.2 (also known as Ethernet) or 802.11 (also known as WiFi™) or 802.16 (also known as WiMax™) or a data network interface such as GSM/GPRS/UMTS/LTE/5G, TETRA™, LoRa™, XBee™, ZigBee™ or the like) , configured for transmitting a signal SM, in which said at least one monitoring datum is encoded, to the supervision computer 3 , wherein said monitoring datum represents an operating state of said switch D, i . e . indicates whether said switch D is operating correctly or requires maintenance because it i s moving with di f f iculty or it is stuck, or the like ;

- a communication bus 17 allowing information to be exchanged among the execution means 11 , the volatile memory means 12 , the non-volatile memory means 13 , the input/output means 14 , and the transmission means 15 .

Al so with re ference to Fig . 3 , when the apparatus 1 i s in an operating condition, said apparatus 1 executes a set o f instructions implementing the method according to the invention, which comprises the following phases :

- an acquisition phase Pl , wherein at least one operating datum is acquired, via the input means 14 , which represents a quantity concerning a manoeuvre of said switch D;

- a processing phase P2 , wherein, by means of the neural network and on the basis of said at least one operating datum, at least one monitoring datum is determined which represents the operating state of said switch D;

- a transmission phase P3 , wherein the signal SM, in which said at least one monitoring datum is encoded, is transmitted, via transmission means 15 , to the supervision computer 3 .

This makes it possible to determine the occurrence of abnormal operating conditions as soon as the switch is first manoeuvred, without requiring the presence of an operator along the line . It will thus be possible to immediately take the necessary steps and/or carry out preventive maintenance , in order to reduce the probability that faults or even accidents might occur .

The operating data collected during the data acquisition phase Pl may be transmitted to entities close to the switch D, e . g . to portable data processing and displaying systems used by service personnel ( e . g . smartphones , tablets or laptops ) or to data collection devices (e.g. microcontrollers comprising communication means, such as a LoRa™ communication interface or the like) positioned on travelling vehicles (e.g. train carriages circulating along the line) and/or flying vehicles (e.g. remotely controlled vehicles performing surveillance tasks over the line) . Furthermore, the operating data may be transmitted to remote entities (e.g. one or more servers) located in a data processing centre, i.e. to the supervision computer 3.

In more detail, the apparatus 1 is preferably configured for transmitting (during the transmission phase P3) , via the transmission means 15, said at least one operating datum to said supervision computer 3, wherein said supervision computer 3 is configured for executing the following steps:

- receiving, via communication means comprised in said supervision computer 3 (and compatible with the transmission means 15 of the apparatus 1) , said at least one operating datum;

- generating a predictive maintenance plan for said switch D on the basis of a historic model (e.g. generated by means of laboratory stress tests, so that said historic model can numerically model the ageing of the parts making up the switch D) and said at least one operating datum, e.g. by calculating a probability of a failure of the switch D on the basis of an operating datum and using, as a model, a historic sequence of operating data of another switch that is similar to the switch D.

It must be pointed out that "predictive maintenance plan" refers to the definitions contained in the UNI 13306 2018, UNI 10147 2003 and UNI 9910 specifications, i.e. a maintenance plan developed by predicting at least one instant of time when the life cycle of a component will end (e.g. when it will fail) on the basis of a historic model and data concerning said component and/or a similar one .

It will thus be possible to carry out predictive maintenance in order to reduce the probability that faults or even accidents might occur .

The neural network is preferably of the feed- forward type , e . g . a Convolutional Neural Network ( CNN) , a Fully Connected Deep Neural Network ( FC DNN) , or a Hierarchical Temporal Memory (HTM) ; such neural network preferably has a number o f inputs equalling the quantity of operating data that have been acquired by the input means 14 and that must undergo the categori zation process . Also , said neural network must be trained beforehand by means of a training process , which is preferably carried out by using a workstation comprising a CPU having more computational capacity than the execution means 11 , wherein said CPU i s preferably configured for executing a set of instructions implementing a training algorithm, preferably a Stochastic Gradient Descent ( SGD) algorithm .

This makes it possible to determine the occurrence of abnormal operating conditions as soon as the switch D i s first manoeuvred, without requiring the presence of an operator along the line . It will thus be possible to immediately take the necessary steps and/or carry out preventive maintenance , in order to reduce the probability that faults or even accidents might occur .

In more detail , the method according to the invention preferably comprises also a training phase , wherein, prior to executing the processing phase P2 , said neural network is trained in a manner such that , when inputting to said neural network said at least one training datum, the latter will output the desired ( second) monitoring datum . This will allow the monitoring system S to be flexibly adapted in the course of its service li fe , e . g . by increasing the number of abnormal operating conditions that said monitoring system S will be able to autonomously recognise .

It will thus be possible to immediately take action and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur.

The sensor means that detect the operating data may comprise one or more of the following elements:

- electric current sensors (e.g. a voltmeter and/or an ammeter) configured for detecting a voltage and/or an intensity of an electric current absorbed by the switch D, in particular by the point machine CM, when said switch is actuated, i.e. when it is making a manoeuvre;

- an inertial sensor (e.g. a gyroscope/accelerometer , preferably of the solid-state type, and/or an inclinometer, and/or the like) so positioned as to detect at least a speed and/or an acceleration and/or an inclination of a portion of said switch (e.g. on the tie rods, on a point blade, on the stock rails, in the point machine, or the like) , wherein said speed and/or acceleration and/or inclination is due to the movement of the point blades AS, AD of the switch D and/or of the rails to which said switch D is constrained through the frame C, to vibration of the rail portion near the stock rails CAS, CAD, produced when the point blades AS, AD end their travel against said stock rails CAS, CAD, to vibration emitted by the actuator located inside the point machine CM, or the like ;

- at least one extensometer (e.g. a resistor-type extensometer, or the like) capable of detecting mechanical deformations of a portion of said switch D being monitored; for example, said extensometer may be positioned near the stock rails CAD, CAS of said switch D;

- at least one temperature sensor (e.g. a thermistor, a thermocouple, or the like) so positioned as to detect the temperature within the point machine CM, of one of the stock rails CAD, CAS, of one of the point blades AS, AD, of the environment, or the like; - an ambient humidity sensor (preferably an electronic Wheatstone-bridge hygrometer ) so positioned as to detect a humidity within the point machine CM or in the environment ;

- a microphone 2 ( e . g . a piezoelectric microphone ) so positioned as to detect a sound produced by at least one element of the switch D, e . g . the noise produced by the operation of the actuator inside the point machine and/or the sound produced in the environment by the point blades AS , AD as they abut on the stock rails CAS , CAD, and vice versa .

Therefore , the operating data preferably represent an intensity and/or a voltage of an electric current absorbed by said switch D during the manoeuvre and/or a sound emitted by said switch D during the manoeuvre and/or an acceleration, a temperature , an extension of a portion of said switch D during the manoeuvre and/or a humidity within the point machine CM of said switch D .

The following will describe a number of technical characteristics that permit detecting a set o f abnormal operating conditions of the switch D . This set o f conditions should however be considered as a non-limiting example .

During the training phase , said at least one training datum represents an intensity and/or a voltage of an electric current absorbed during the ( second) manoeuvre of said second switch when it is poorly lubricated, and said second monitoring datum indicates that said second switch is moving with di f ficulty .

This makes it possible to automatically detect an abnormal operating condition caused by insuf ficient lubrication of the actuator and/or of bearings allowing the movement of the point blades AS , AD; in fact , because of increased friction, the electric power absorbed during the actuation will exceed a range which is supposed to be normal . It will thus be pos sible to immediately take the necessary steps and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents a temperature of a portion of the second switch during the ( second) manoeuvre when it is poorly lubricated, and said second monitoring datum indicates that said second switch i s moving with di f ficulty .

This makes it possible to automatically detect an abnormal operating condition caused by excessive friction undergone by a portion of the switch D, e . g . because of insuf ficient lubrication of an actuator and/or of bearings allowing the movement o f the point blades ; in fact , due to greater friction, more heat wi ll be dissipated than during a normal actuation . It wi ll thus be possible to immediately take the necessary steps and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents an acceleration of a portion of the second switch during the ( second) manoeuvre when it is poorly lubricated, and said second monitoring datum indicates that said second switch i s moving with di f ficulty .

This makes it possible to automatically detect an abnormal operating condition caused by insuf ficient lubrication of the actuator and/or of bearings allowing the movement of the point blades AD, AS , since the acceleration of the point blade during the actuation will be lower than a range that is supposed to be normal , combined with an increased intensity of high- frequency accelerations ( e . g . in the range of 50- 60 Hz ) induced by electric windings included in the actuator, which will encounter excessive resistance while driving the switch D . It will thus be possible to immediately take the necessary steps and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur . In combination with or as an alternative to the above , during the training phase , said at least one training datum represents a sound emitted by said second switch during the ( second) manoeuvre when it is poorly lubricated, and said second monitoring datum indicates that said second switch i s moving with di f ficulty .

This makes it possible to automatically detect an abnormal operating condition caused by insuf ficient lubrication of the actuator and/or of bearings allowing the movement of the point blades AD, AS , since during the manoeuvre a greater portion of energy will be di ssipated as noi se , and hence the microphone 2 will detect a signal representing sounds generated by squeaking, creaking or vibrating mechanical parts , such as a tie rod TR, the point machine CM, etc . It will thus be possible to immediately take the necessary steps and/or carry out preventive maintenance in order to reduce the probability that faults or even accidents might occur .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents a sound emitted by said second switch during the ( second) manoeuvre when at least one tie rod comprised in said second switch is broken or the actuator contained in the point machine CM is out of calibration due to a thermal dri ft phenomenon, and said second monitoring datum indicates that said second switch is faulty .

This makes it possible to automatically detect an abnormal operating condition caused by breakage of said at least one tie rod TR or by an imperfect actuation thereof by the actuator, since during the actuation no striking noise will be detected, produced by one of the point blades AS , AD of the switch D abutting on one of the stock rails CAS , CAD, or that sound wil l be detected either too late or too soon . It will thus be possible to immediately take action in order to reduce the risk of accidents . In combination with or as an alternative to the above , during the training phase , said at least one training datum represents a bending of a tie rod comprised in said second switch when said second switch is at least partly stuck, and said second monitoring datum indicates that said second switch i s moving with di f ficulty .

This makes it possible to automatically detect an abnormal operating condition caused by bending of said at least one tie rod TR, since during the actuation at least one of the point blades AS , AD encounters an obstacle along its travel , e . g . a foreign body caught between said point blade AS , AD and its stock rail CAS , CAD . It will thus be possible to immediately take action in order to reduce the risk of accidents .

In combination with the above , during the training phase , said at least one training datum comprises also an ambient temperature and/or humidity value . This makes it possible to cancel the thermal/environmental drift phenomena to which extensometers are especially subj ect , thus increasing the level of precision of the monitoring activity ( i . e . reducing the risk o f false positives or negatives ) . This will result in a more ef fective preventive maintenance , reducing the risk of accidents .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents an intensity and/or a voltage of an electric current absorbed by said second switch during the ( second) manoeuvre when at least one tie rod comprised in said second switch is broken, and said second monitoring datum indicates that said second switch is faulty .

This makes it possible to automatically detect an abnormal operating condition caused by a mal function or breakage of said at least one tie rod TR of the switch D; in fact , the electric power absorbed during the manoeuvre will be either higher ( in the event of a malfunction) or lower ( in case of breakage ) than a range which is supposed to be normal . It will thus be possible to immediately take action in order to reduce the risk o f accidents .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents an acceleration of a portion of the second switch during the ( second) manoeuvre when at least one tie rod comprised in said second switch is broken, and said second monitoring datum indicates that said second switch is faulty .

This makes it possible to automatically detect an abnormal operating condition caused by breakage of said at least one tie rod TR, since the acceleration o f the point blade during the actuation wil l be either lower or higher than a range which is supposed to be normal , depending on the position o f the inertial sensor . I t will thus be pos sible to immediately take action in order to reduce the risk of accidents .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents an intensity and/or a voltage of an electric current absorbed by said second switch during the ( second) manoeuvre when said second switch is stuck, and said second monitoring datum indicates that said second switch is stuck .

This makes it possible to automatically detect an abnormal operating condition caused by sticking of the switch D, since the intensity and/or voltage of the electric current absorbed by said switch D will substantially equal the characteristic values of a motor comprised in the actuator contained in the point machine CM when the rotor of said motor is stuck . It will thus be possible to immediately take action in order to reduce the risk of accidents .

In combination with or as an alternative to the above , during the training phase , said at least one training datum represents a sound emitted by said second switch during the ( second) manoeuvre when said second switch is stuck, and said second monitoring datum indicates that said second switch is stuck . This makes it possible to automatically detect an abnormal operating condition caused by sticking of the switch D, since the sound detected by the microphone will have a frequency within a range , preferably 49 Hz to 61 Hz , resulting from vibrations induced by the electric windings of the actuator, which wi ll cause said at least one tie rod TR, the point machine CM, etc . to vibrate .

It wi ll thus be poss ible to immediately take action in order to reduce the risk of accidents .

In combination with or as an alternative to the above , for the purpose of increasing the level of precision of the signalling provided by the apparatus 1 through the monitoring data, the sensor means preferably comprise also video acquisition means ( e . g . a camera ) configured for acquiring at least one image of said network R ( e . g . a portion of the railway line before or after the switch D) . More generally, the input means 14 o f the apparatus 1 are preferably also configured for acquiring a carriage presence signal ( generated by the sensor means ) having a content which is influenced by a carriage moving towards or away from the switch D, and wherein said execution means 11 are also configured for executing the following steps :

- detecting, on the basis of said presence signal , the carriage moving towards or away from said switch ( D) ;

- determining said at least one monitoring datum only when said carriage is detected .

In more detail , i f the sensor means comprise video acqui sition means , the presence signal preferably comprises a video stream representing ( a portion of ) said network R, preferably that portion of the network R which is entering or exiting said switch D, so that the execution means 11 can detect the presence of the carriage by executing a set of instructions implementing a video acquisition algorithm .

In this manner, the operation of the apparatus 1 is optimi zed and the risk of false positives is reduced, because the monitoring datum can only be determined when a carriage is entering or exiting the switch D, i . e . when said switch is being stressed and the data received by the apparatus 1 are most signi ficant . It follows that this mode of operation, based on images and/or vibration detections and/or sound detections and/or the like always aimed at detecting a train approaching the switch D, makes it poss ible to reduce the probability that faults or even accidents might occur .

In combination with or as an alternative to the above , the apparatus 1 preferably comprises also the following elements :

- electric accumulator means ( e . g . a lithium-ion battery, a supercapacitor, or the like ) configured for supplying power to said apparatus 1 or part thereof ;

- generator means ( e . g . a photovoltaic panel and/or an aeolian and/or piezoelectric and/or thermoelectric and/or vibrational generator and/or the like ) connected to said accumulator means and configured for converting luminous , mechanical or thermal energy into electric energy .

This makes it possible to position said apparatus 1 at a larger number o f points along the network R, in that said apparatus 1 does not need a low-voltage power supply . The increased positioning capability of the apparatus 1 further reduces the risk of accidents .

Of course , the example described so far may be subj ect to many variations .

Some of the possible variants of the invention have been described above , but it will be clear to those skilled in the art that other embodiments may also be implemented in practice , wherein several elements may be replaced with other technically equivalent elements . The present invention is not , therefore , limited to the above-described illustrative examples , but may be subj ect to various modi fications , improvements , or replacements of equivalent parts and elements without however departing from the basic inventive idea, as speci fied in the following claims .