WO2024089585A1

WO2024089585A1 - Process, system and computer program for monitoring a device to be monitored such as a milling machine for diaphragms, a drilling machine for piles and further heavy construction equipments

Info

Publication number: WO2024089585A1
Application number: PCT/IB2023/060686
Authority: WO
Inventors: Ivano Guerra
Original assignee: Soilmec S.P.A.
Priority date: 2022-10-24
Filing date: 2023-10-23
Publication date: 2024-05-02

Abstract

The process for monitoring the operation of a device to be monitored (100, 100') according to the invention comprises the following steps: A) providing a virtual simulator (200) programmed or otherwise configured for simulating, through a supervised learning artificial intelligence algorithm, the operation of the real device to be monitored (100, 100') under nominal conditions; B) through input sensors detecting the input physical quantities (VINi) of the device to be monitored (100, 100') and through index sensors detecting the index physical quantities (VURj) of the device to be monitored (100, 100'); C) determining a simulated value (VUNj) of the index physical quantities by the virtual simulator (200) based on the plurality of input physical quantities (VINi); D) comparing said simulated value of the index physical quantities (VUNj) with the value of the index physical quantities (VURj) detected by the one or more index sensors. The invention further concerns a monitoring system and a computer program.

Description

PROCESS, SYSTEM AND COMPUTER PROGRAM FOR MONITORING A DEVICE TO BE MONITORED SUCH AS A MILLING MACHINE FOR DIAPHRAGMS, A DRILLING MACHINE FOR PILES AND FURTHER HEAVY CONSTRUCTION EQUIPMENTS

Field of the invention

[1] The present invention concerns a process, a system and a computer program for detecting in advance anomalous operating conditions, and in particular fault predictors, in a device such as for example a hydromilling machine, a bucket or other operating machine.

Prior art

[2] In the field of the foundation machines, such as for example drilling machines for making piles or equipment for digging diaphragm walls by means of a hydro-milling machine or bucket, there may be more or less prolonged machine downtimes due to faults in turn caused by the particularly heavy working conditions to which this type of machines is subjected.

[3] In order to minimize such machine downtimes, which entails high costs of non-product ion, it is currently known to carry out regular interventions of routine maintenance and preventive replacement of mechanical, hydraulic or electrical components of critical importance for the operation of said equipment.

[4] If, on the one hand, this reduces the costs connected to non-product ion, on the other hand, the owners or operators of the machines must bear the costs for the spare parts and labour hours for the preventive maintenance interventions.

[5] In addition, components are often replaced which, although no longer in optimal operating condition, could still be used for tens or in some cases hundreds of hours of work before causing failures.

[6] To overcome the above drawbacks, fault detection methods based on the knowledge of the physical behaviour of individual components of the operating machine are currently known.

[7] These known detection methods are based on mathematical models of behaviour of the components that analyse input data coming from sensors and compare them with the theoretical operating values of the component.

[8] In the known predictive methods vibrations detected in the vicinity of a component are for example analysed in frequency; the temperatures detected and the other quantities detected can also be analysed with time filters to verify if an anomalous condition persists longer than a reference time interval so as to reduce false alarms .

[9] If the measured values exceed the threshold values for a certain period of time, an appropriate control system emits alarm signals.

[10] These methods often require long development times to model the physical behaviour of the component or to filter the false alarms, and they often manage to detect the fault with too short advance with respect to when the fault occurs or even when a fault has already occurred, thus not leaving enough time to carry out the intervention of preventive replacement of the faulty or otherwise malfunctioning component.

[11] The author of the present inventing believes that a limitation of these known fault prediction systems lies in the fact that in a complex machine they lead to the adoption of a plurality of thresholds, each of which relating to a particular component or subgroup, and signal anomalies simply when only one of these thresholds is reached or exceeded, monitoring independently of each other the behaviours of the various components.

[12] An object of the present invention is to obviate the aforementioned drawbacks of the state of the art by providing a system and a process capable of detecting, with more advance than the known systems and processes, operating conditions that are a sign of faults or malfunctions such that a machine downtime is required. Summary of the invention

[13] In a first aspect of the present invention, this object is achieved by a process having the features according to claim 1.

[14] In a second aspect of the present invention, this object is achieved with a system having the features according to claim 13.

[15] In a third aspect of the present invention, this object is achieved with a computer program having the features according to claim 16.

[16] Such a computer program may be run entirely on one or more logic units installed on board the device to be monitored, entirely on one or more logic units outside the device to be monitored, such as for example on a remote computer connected in the cloud to the device to be monitored, or it may be run partly on one or more logic units installed on board the device to be monitored and partly on one or more logic units outside that device .

[17] Further features of the invention are the subjectmatter of the dependent claims.

[18] The advantages achievable with the present invention will become more apparent, to the person skilled in the art, from the following detailed description of some particular non-limiting examples of embodiment, illustrated with reference to the following schematic figures.

List of Figures

Figure 1 shows a side view, according to the direction of observation Y, of a milling machine for diaphragm walls monitored by means a process and the monitoring system according to a particular embodiment of the invention;

Figure 2 shows a perspective view and a functional diagram of the milling machine of Figure 2;

Figure 3 shows an example functional diagram of a Perceptron belonging to the virtual simulator usable to monitor the milling machine of Figure 1 according to a particular embodiment of a monitoring process and system according to the invention;

Figure 4 shows an example of a neural network usable in a monitoring process and system according to a particular embodiment of the present invention and comprising multiple Perceptrons of the type of Figure 3;

Figure 5 shows a diagram of some components of the milling machine of Figure 1 monitored or that can possibly be monitored via the neural network of Figure 4;

Figure 6 shows a block diagram of the process for detecting nominal or non-nominal operating conditions according to a particular embodiment of the present invention;

Figure 7 shows a graphical example of a residual error obtained with a process and with a monitoring system according to a particular embodiment of the invention; Figure 8 shows an enlarged detail of the graph of Figure 7;

Figure 9A shows a graph of the alarms caused by residual errors obtained with a process and of a monitoring system according to a particular embodiment of the invention; Figure 9B shows a graph of the drift as a function of the time of the residual errors from which the graph of Figure 9A was also obtained;

Figure 9C shows a graph of the first derivative as a function of the time of the residual errors from which the graph of Figure 9A was also obtained;

Figure 10 shows a perspective view of a drilling machine for piles or other foundations monitored by means of a process and the monitoring system according to a particular embodiment of the invention.

Detailed description

[19] Figures 1-9C relate to a system, a process and a computer program according to a particular embodiment of the invention.

[20] The process, control system and computer program according to the present invention is particularly suitable for monitoring the operation and detecting in advance the premonitory conditions of faults or other malfunctions and machine downtimes for example of milling machines for digging diaphragm walls of the type for example described in patent applications EP2378181A1, EP2924174A1, EP3467209A1, EP2573275A1 or EP2586962A1, drilling machine for piles or other foundations of the type for example described in patent applications EP2339108A1, EP3372777A1, EP3721177A1,

EP2468960A1 or EP3293351A1.

RECTIFIED SHEET (RULE 91) ISA/EP [21] A further example of the aforementioned type of milling machines for digging diaphragm walls, also called "milling machines" in the current technical jargon, is shown in Figure 1 and indicated with the overall reference 100.

[22] The milling machine 100 preferably comprises a base machine 2 and a disintegrating or digging tool 3 operatively connected to the base machine 2.

[23] The base machine 2 may comprise a tracked undercarriage 4, a turret 5 rotating with respect to the tracked undercarriage 4 and an arm 6, tiltable and hinged to the turret 5, which supports the digging tool 3 by means of a suspension element 7 which is driven forward through a winch.

[24] The suspension element 7 may be flexible, can be directly wound or unwound through a winch 8.

[25] The winch 8 is preferably installed on the base machine 2, inside or above the body, or fixed near the winders 44 and 45, or still connected directly to the arm 6.

[26] On or in any case on the rotating turret 5 there is preferably fixed a driver' s cab 102 configured for containing the driver of the milling machine 100 or other device to be monitored 100, 100' and any control unit 31 (Figure 2) .

[27] Within the rotating turret 5 of the base machine 2 there is preferably housed a main power motor 17, preferably single and of endothermic type - such as for example a diesel motor or - but which in alternative constructions could also be electric.

[28] The main power motor 17 is preferably configured for providing the mechanical power necessary to implement all the hydraulic functions of the digging equipment 1, thus both the main service functions and the main digging functions.

[29] A coupler 18 provided with a plurality of output shafts, to which the power received from the motor 17 is distributed, can be connected to the output shaft of the power motor 17.

[30] A plurality of pumps that can be divided into a first group of pumps 19, and a second group of pumps 20, to command the digging functions, can be connected to the output shafts of the coupler 18.

[31] The system preferably comprises at least one pump or a group of pumps, at least one actuator, at least one distributor or control valve for commanding the actuators, at least one heat exchanger, at least one main oil storage tank and at least the pipings necessary for connecting the aforesaid components.

[32] In particular, at least the winch 8 or the means directly involved in the extraction of the tool 3 from the digging (for example one or more command actuators of the winders) can be connected to the first group of pumps 19.

[33] Any winders of the hydraulic pipes 44 or of the sludge pipe 45 can be fed independently of the winch 8, or be fed with a diverter valve from the same line.

[34] All the supplies of the toothed drums 10, 11 and of the suction pump 14 can be connected to the second group of pumps 20.

[35] All the actuators installed on the digging tool 3 that command and operate the digging functions and that are at least partially introduced into the digging and immersed in the stabilizing fluid can be connected to the second group of pumps 20.

[36] These actuators that command the digging functions can be connected to the base machine 2 via hydraulic supply and discharge lines 12A, 12B, 13A, 13B, 14A and 14B, also called delivery and return lines, which provide the hydraulic power.

[37] The group of pumps 19 for the service functions is preferably configured for sucking the oil from the main tank 25 and send it to the distributor 21, to which the service function actuators are connected.

[38] The group of pumps 19, the main tank 25 of the first hydraulic circuit S and the distributor 21 are preferably installed on the base machine 2.

[39] There can be many actuators connected to the hydraulic circuit, whether they are of rotary or linear type .

[40] Among these, for example, the rotary motors to command the movement of the tracks of the undercarriage 4, the rotation of the turret 5, the rotation of the tool movement winch 8 3, the rotation of the manoeuvring winches of the support arm 6 and the rotation of the winders for the hydraulic pipings and for the sludge pipe .

[41] For simplicity' s sake Figure 2 shows only two service actuators 8 and 41.

[42] The first actuator can coincide with the rotary motor of the winch 8, which can wind or unwind the suspension element 7 causing the tool 3 to rise or fall in the digging of the diaphragm wall.

[43] The group of pumps 20 for the digging functions can be configured for sucking the oil from the main tank and it sends it to the distributor 22, to which the actuators of the digging functions are connected.

[44] The group of pumps 20 for the digging functions, the main tank and the distributor 22 are preferably installed on the base machine 2.

[45] With reference to figure 2, the actuators of the digging functions 12, 13 e 14 installed on the digging tool 3 can be connected to the distributor 22.

[46] The actuators 12 and 13 are preferably rotary motors that drive the rotation of the toothed drums 10 and 11 of the milling machine.

[47] The actuator 14 can be a suction pump of the digging sludges, which are sent to the surface via a dedicated flexible piping, called a "sludge pipe" .

[48] The rotary actuators 12, 13 and 14 of the digging tool 3 can each be advantageously connected to a third drain line 12C, 13C and 14C at low pressure for disposal of the lubricating oil or of the excess oil.

[49] The pipings 12A, 12B, 13A, 13B, 14A, 14B and 33C preferably follow the movement of the tool 3 in the digging and are at least partially immersed in the stabilizing fluid.

[50] It follows that, in the presence of loose or damaged fittings, in the presence of cracks in the pipes or of an excessive pressure difference between the pressure of the external liquid and the internal pressure of the components installed on board the digging tool, the stabilizing sludge of the digging pushed by the hydrostatic pressure can penetrate into the pipes contaminating the oil or penetrate into the components damaging them, or vice versa the oil could escape contaminating the external stabilizing sludge.

[51] To overcome this drawback, compensation devices are advantageously arranged on the frame 9 and connected to the main actuators so as to restore internally to them the same external pressure, which grows linearly with the depth and with the density of the fluid.

[52] The hydraulic motors 12, 13 driving the digging wheels 10,11 are preferably installed in a "housing chamber of the motors or motor reducers" which is completely filled with hydraulic oil.

[53] This chamber is preferably separated from the external environment by the gaskets, interposed between the housing of the rotor and the rotating drums.

[54] When the digging tool 3 is immersed in the digging fluid, the aforesaid gaskets separate the housing chamber of the motor 12, 13 with respect to the fluid present in the digging.

[55] To prevent the digging fluid from entering the chamber of the motor 12, 13 it is preferable to keep this chamber pressurized and adapt its pressure to the changing pressure of the external environment, i.e. the digging fluid.

[56] At least one compensation device 46, is then connected to the housing chamber of the motor 12, 13 and which is preferably driven by the pressure present in the digging fluid - at the depth where both the compensator 46 and the digging module are located at a certain instant - and acts on the internal pressure of the housing chamber of the motor, adapting it accordingly .

[57] Preferably a respective compensator 46 is connected to each motor 12, 13.

[58] Although these compensation devices 46 are simple, of mechanical type and provided with a membrane for the direct transduction of the pressure, it can sometimes happen that they are not precise and that, despite the presence of external pressure commands to correct the errors, even momentary pressure imbalances can occur, which in the long run damage the seal of the relative moving members .

[59] The probability of sludge penetrating a component of the hydraulic circuit is all the greater as the hydrostatic pressure is greater than the internal pressure of this component of the hydraulic circuit, whereby this problem becomes increasingly burdensome as the digging depth increases and it especially affects the low pressure return lines.

[60] The penetration of contaminating particles, such as sand, water or sludges, into the hydraulic circuit can also take place through the digging actuators, in particular the rotary ones 12 and 13, where the sealing gaskets between the rotating parts are in direct contact with the stabilizing sludge.

[61] These rotating gaskets in fact separate the external environment, in which the pressure of the digging fluid is present, from the housing chamber of the motor and of the reducers in which the pressure of the compensation circuit caused by the compensator 46 is present.

[62] An incorrect or non-operat ion of the compensator 46 could therefore lead to the presence of an insufficient compensation pressure value in the housing chamber of the motor and of the reducers, and this low pressure value could allow the entry of the digging fluid into the chamber causing damages to the motor and the reducers .

[63] These gaskets are lubricated using a part of the oil of the system.

[64] The milling machine 100 may also comprise, as already mentioned in part, one or more of the following components or subsystems:

- a sludge and debris suction system 110 arrange to suck the sludge, water and other liquids or sewage present at the bottom of the digging and in which the wheels or the toothed drums 10, 11 of the milling module 106 can -as it mostly happens- be at least partially immersed while the milling module 106 is digging;

-one or more of the aforementioned hydraulic compensators 46;

-one or more electrical panels 118, each of which comprises switches -for example power ones- and other electrical circuits that power or control for example the other actuators, pumps and other hydraulic components or the main heat motor 17 of the milling machine 100.

[65] The winch group 108 may for example comprise a hydraulic motor configured for driving a winch 8 which by means of a suspension element 7 supports and moves the digging tool 3.

[66] The sludge and debris suction system 110 preferably comprises one or more of the following:

-a digging sludge suction pump 14;

- one or more dedicated flexible pipings 1100 currently called "sludge pipes" 1100 and fluidically connected to the sludge suction pump 14;

-one or more winders 44, 45 arranged for winding and unwinding -for example on reels or other drums- the sludge pipes 1100 to keep the pump 14 constantly in fluidic connection with the sludge pipes 1100 as the height of the frame 1066 of the milling module 106 varies .

[67] The pump 14 is configured for sucking the sludges and the debris present at the bottom of the digging and in which the wheels or other milling tools 1060 are generally immersed during operation, and it evacuates them through the sludge pipes 1100.

[68] The hydraulic compensator 46, a possible example of which is described in the aforementioned patent application EP 2924174 Al, serves to regulate the pressure of the hydraulic circuits inside the machine 100, in particular of the hydraulic circuits that during operation are immersed in a liquid or other external fluid at relatively high pressure such as for example the hydraulic circuits that drive the wheels or milling drums 1060 or otherwise present in the frame 1066 of the milling module 106.

[69] Such liquids or other external fluids may for example be the liquid mixture of sludge and debris and/or the stabilizing fluid, such as for example a sludge based on bentonite or polymers, present at least in the deepest part of the digging.

[70] The hydraulic compensator 46 is configured for regulating the pressure of the hydraulic circuits internal to the machine 100 by adapting it based on the one of the aforementioned liquids or other external fluids and making it not much lower and more preferably equal to or greater than the pressure of the liquids or other external fluids, so as to reduce the entry of external contaminating particles into the internal hydraulic circuits of the machine 100.

[71] In particular the circuit of the hydraulic motor 14 that drives the sludge suction pump 14 is advantageously and fluidically connected to a hydraulic compensator 46. [72] The process according to one aspect of the present invention, for monitoring the operation of a device to be monitored such as for example the milling machine 100, a drilling machine for piles or other foundations or other operating machine, comprises the following steps : -providing a virtual simulator programmed or otherwise configured for simulating the operation of the real device at least under nominal conditions through a supervised learning artificial intelligence algorithm; -providing, on the real milling machine 100, drilling machine 100' or other device to be monitored, one or more input sensors, each of which is configured for detecting one or more input physical quantities VIN of the device to be controlled 100, and one or more index sensors, each of which is configured for detecting one or more second index physical quantities VURj;

-determining a simulated value VUNj of the index physical quantities by the virtual simulator based on the input physical quantities detected by the one or more input sensors ;

-comparing said simulated value VUNj of the index physical quantities with the value of the index physical quantities VURj measured more directly by the one or more index sensors (Figure 6) , where:

-i is an integer variable between 1 and N and N can be for example between 1-500 or 1-1000;

-j is an integer variable between 1 and M, where M is preferably less than N and e for example comprised between 1-500 or 1-1000 but may also be equal to or greater than N;

-every physical quantity VIN , VUNj, VURj can for example vary as a function of the time and thus be VIN = VIN (t) ; VUNj = VUNj (t) ; VURj = VURj (t) .

[73] Preferably this process comprises the step of evaluating whether the machine 100 is operating under nominal or anomalous conditions or potentially at risk based on the comparison of the one or more index physical quantities determined by the virtual simulator with the corresponding one or more index physical quantities detected by the one or more index sensors.

[74] Preferably the virtual simulator comprises the simulation - for example carried out by a computer program - of a neural network 200, preferably of feedforward type, formed by one or more so-called Perceptrons- (Figure 3, 4) ; Figure 3 shows an example of a single Perceptron, known per se.

[75] Preferably such a neural network 200 is of deterministic type in the sense that its one or more outputs is determined deterministically based on its inputs and the weights of the synaptic connections of the one or more Perceptrons are determined deterministically and not for example probabilistically or according to fuzzy logic.

[76] Preferably said neural network 200 comprises a plurality of Perceptrons connected so as to form several layers in cascade, from upstream to downstream (Figure 4) .

[77] Preferably in the neural network 200 the output signals of the Perceptrons of one layer more upstream are used as inputs of all the Perceptrons at least of the layer immediately downstream (Figure 4) . [78] Preferably said neural network 200 comprises three layers of Perceptrons, of which the output signals of the Perceptrons of the first layer, more upstream, are provided as inputs to all the Perceptrons of the intermediate layer between the first layer, more upstream, and the third layer, more downstream.

[79] Preferably each layer of the neural network 200 may comprise a number of Perceptrons or other types of neurons comprised between 1-10.000, or between 1-1000, between 3-500, between 10-400, between 50-300 Perceptrons or other types of neurons.

[80] The neural network 200 may however comprise a different number of layers, for example a single layer, two single layers or four, five, six or more layers.

[81] The most downstream layer of all of the neural network 200 could for example comprise at least as many neurons as there are components of the milling machine or other device to be monitored 100 of which it is desired to predict the faults or malfunctions: for example with reference to Figure 5 the most downstream layer of all of the neural network 200 could comprise at least 12 Perceptron or other neurons or a number of neurons multiple of 12, if more physical quantities of a plurality of components of the milling machine for the diaphragm walls 100, of the drilling machine 100' or of other devices to be monitored were to be simulated via the neural network 200.

[82] Preferably the input sensors cannot also act as index sensors.

[83] This means that the same input sensor cannot act, at least at the same time, also as an index sensor; however, this does not exclude that an input sensor can be of the same type as an index sensor and that both are capable of measuring, for example, two temperatures, two pressures, two flow rates, two speeds.

[84] The physical quantities inputted to and outputted from the neural network 200, the outputs of the input sensors and index physical quantities detected by the one or more index sensors may be constant or time-varying signals, acquired or provided for example as continuous or sampled - i.e. discrete - functions over time.

[85] The aforementioned comparison of the one or more index physical quantities determined by the virtual simulator with the corresponding one or more index physical quantities detected by means of the one or more index sensors is preferably performed by comparing simulated values acquired by the index sensors simultaneously with each other, for example by comparing values sampled at the same time instant.

[86] The outputs of the input sensors and the index physical quantities detected by means of the one or more index sensors may possibly be filtered and undergo further signal processing before being processed respectively by the neural network 200 or by the residual error calculation process described below.

[87] By "to simulate the operation of the real device at least under nominal conditions" in the present description is intended to ensure that the virtual simulator mimics the behaviour of the device 100 or one or more of its subgroups or components under design or optimal, suboptimal conditions or in any case under conditions that do not allow to predict or in any case indicate malfunctions or faults, or in any case are conditions sufficiently similar to the ideal operating conditions .

[88] These nominal operating conditions can be defined quantitatively or quantitatively depending on the subgroup or component considered.

[89] For example, a condition that, based on historical data acquired on other specimens of devices 100 of the same or similar model, is not associated with events of faults or malfunctions, which have actually occurred, after an appropriate reference period of time, can be considered as a nominal operating condition.

[90] The learning, i.e. training, of the neural network or other supervised learning artificial intelligence system, is preferably conducted based on these definitions of normal operation of the device to be monitored, whether it is an entire operating machine 100, 100' or only one or more of its components and subgroups .

[91] This training phase of the neural network 200 is preferably conducted based on the historical data - such as, for example, recordings of operating events - of operating machines 100, 100' or other real devices to be monitored 100, 100' .

[92] This historical data can be obtained for example from devices to be monitored 100, 100' that are identical or in any case similar to the specimen that is to be monitored by the trained neural network 200.

[93] Advantageously, according to a particular embodiment thereof, the process according to the invention comprises the step of calculating, or in any case determining, the derivatives -for example first or second ones- and/or the integral over time of one or more index physical quantities simulated by the neural network -in other words the derivatives and/or the integral over time of the outputs of the neural network 200, realizing a data augmentation that allows a better analysis of the behaviour of the device to be monitored 100, 100' .

[94] Advantageously, according to a particular embodiment thereof of the process according to the invention, the comparison between the value of one or more index physical quantity simulated by the virtual simulator 200 and the value of that physical quantity actually detected -for example directly through one or more index sensors, for example the aforementioned sensor that detects the pressure inside the compensator 46- is carried out by means of the following steps:

5.1) determining - for example by calculating it - a difference between the simulated index physical quantity VUNj and that VURj directly measured by the corresponding one or more index sensors;

5.2) preferably determining an absolute value of this difference, for example the simple modulus or the square of the aforementioned arithmetic difference; this absolute value is conventionally indicated, in this description, as "residual error" ERSj and its value is the greater the greater is the deviation between the value detected through the index sensors on the milling machine or other real device to be monitored 100, 100' and its value simulated through the neural network or other virtual simulator 200 (block 300 of the block diagram in Figure 6) ;

5.3) evaluating an anomalous operating condition, indicating a possible future fault or other malfunction, based on the residual error, possibly based on its trend and/or frequency over time (block 400 of the block diagram in Figure 6) .

[95] The step S.3) can be implemented for example by calculating the mean squared error - that is, standard deviation o - over time of the residual error of one or more index physical quantities and verifying whether it is less than or greater than a predetermined threshold.

[96] This predetermined threshold can be, for example, equal to three or five times this standard deviation (Figure 6, 7) .

[97] Each residual error ERSk can for example vary as a function of the time and thus be ERSk = ERSk (t) , where j is an integer variable between 1 and Q, where Q can be equal to, greater than or less than the number N of input physical quantities VINN or the number M of index physical quantities VUNj or directly detected VURj .

[98] Preferably but not necessarily Q and M are less than N.

[99] The residual errors can be calculated as a mean for example on each work cycle or working day; in this case the axis of the abscissas of the graph of Figure 7, 8, 9A-9C corresponds to the time and the axis of the ordinates to the mean on each work cycle or working day of the residual error of a predetermined index physical quantity .

[100] If the residual error exceeds a certain first or second predetermined threshold, also called "lower threshold" and "upper threshold" respectively in the present description, with a frequency over time equal to or greater than a third and fourth predetermined threshold, according to a particular embodiment of the process according to the invention, a first and a second non-nominal operating condition can be detected respectively, diagnosing de facto an anomalous operation potentially indicating future faults or malfunctions of the corresponding component, subgroup or of the entire milling machine 100 or other device to be monitored 100, 100' .

[101] The lower and upper threshold and the corresponding third and fourth threshold in frequency may be associated with an anomaly of operation that is respectively mild and severe, and preferably not associated with the occurrence of a real fault.

[102] Reaching or exceeding the upper threshold and possibly also the corresponding fourth threshold in frequency may possibly lead to the start of revision or preventive maintenance procedures to the milling machine or other device to be monitored 100, 100', with significant cost savings and machine downtimes.

[103] These lower and upper thresholds are preferably the daily average threshold of a residual error.

[104] For example with reference to Figure 7 exceeding the lower threshold once, and the upper threshold four times over four consecutive days or working cycles can be interpreted as an anomaly of operation respectively mild and severe, i.e. respectively mild and severe alarm.

[105] Advantageously, according to a particular embodiment thereof, the process according to the invention comprises the step of determining the trend or drift over time of the one or more residual errors corresponding to one or more respective index physical quantities .

[106] For this purpose, the process according to the invention preferably comprises the step of determining a derivative - more preferably the first derivative - with respect to the time of the one or more residual errors and verifying whether said derivative is equal to or greater than a fifth predetermined threshold (Figure 9B) .

[107] This derivative may indicate a drift of the milling machine 100 or other device to be monitored 100 from a nominal operating condition and of a serious operating anomaly .

[108] For example in the historical data series shown in Figure 9C the derivative before the residual error has exceeded the fifth predetermined threshold on 01 March and on 07 March a fault occurred, such as for example the breakage of a component of the milling machine 100.

[109] The drift from this nominal operating condition is shown for example by the trend over time of the mild or severe anomalies (Figure 8B) .

[110] Advantageously, according to a particular embodiment thereof, the process according to the invention comprises the step of emitting a message or other serious alarm signal when the derivative -for example the first derivative- of a residual error reaches or exceeds the aforementioned fifth predetermined threshold .

[111] The neural network or other virtual simulator 200 can receive in input a number of input physical quantities, for example comprised between 1-1000, or between 1-500, 5-300, 10-200, 50-100.

[112] The neural network or other virtual simulator can calculate or otherwise determine -for example simultaneously or at the same iteration of the virtual simulation algorithm, if it operates with discretized time- in output a number of index physical quantities for example comprised between 1-500 or 2-400, 5-300, 10-

200, 50-100, 60-80.

[113] The input physical quantities VIN -where i is an integer variable between 1 and N and N can be understood, as mentioned, for example between 1-1000 and any physical quantity VIN can for example vary as a function of the time and thus be VIN = VIN (t) - provided in input to the neural network or other virtual simulator 200 can for example be one or more physical variables of the milling machine 100 or other device to be monitored 100, 100 ' cho sen for example from the following list:

- a temperature of a component for example hydraulic, electrical or electronic;

-a supply, internal or delivery pressure of a hydraulic component such as a pump, cylinder, rotary motor or other linear or rotary motor or actuator, a filter, a distributor valve, pressure limiter, check valve or other type of valve;

-a pressure along a conduit traversed by or otherwise containing hydraulic oil or other fluid of a hydraulic or pneumatic circuit

-time -for example seconds, minutes, hours or days- of operation or of inactivity of a component;

-speed of rotation of a thermal motor -such as for example the main power motor 17— or of a hydraulic motor or other rotary actuator such as for example the rotary motors that command the movement of the tracks of the undercarriage 4, the rotation of the turret 5, the rotary motor 108 that drives the winch 8 or more generally the rotation of the winch 8 for moving the tool 3, the rotation of the manoeuvring winches of the support arm 6, the rotation of the winders for the hydraulic pipings and for the sludge pipe, the rotary motors 12, 13 that drive the rotation of the toothed drums 10, 11;

- linear speed of translation of a hydraulic cylinder, such as for example the hydraulic cylinders 16 that drive any correction flaps 15, the hydraulic cylinders that incline the support of the toothed drums 10, 11 or other milling wheels or of other hydraulic, electric or pneumatic linear actuator;

- depth at which the digging tool 3 is located;

-speed of descent or of ascent of the digging tool 3;

- resisting torque at the output shaft of a thermal, hydraulic or electric motor;

-resisting force on the stem or other output slider of a hydraulic cylinder, linear electric motor or other linear actuator;

-temperature of the external environment;

-temperature of an electrical panel;

-temperature of the oil or other fluid of the hydraulic circuit at high or low pressure;

- speed or flow rate of the hydraulic oil or other fluid in a predetermined point of the hydraulic circuit, such as for example a point inside a rotary or linear motor or pump, in a supply or return, i.e. delivery or discharge conduit from a rotary or linear motor or pump or still in a particular point of a piping, fitting, coupling, manifold or other conduit;

-pressure or difference in delivery, internal or output pressure between the motors that drive the winches 8;

-weight of the digging tool 3 during the execution of a digging; tensioning of the rope or other suspension element 7 for example during the execution of a digging;

-fuel consumption and/or oil pressure of the main power motor 17;

-inclination of the arm 6;

-inclinations of the base machine 2 and/or of the digging tool 3 with respect to the two reference axes horizontal or parallel to the ground X and Y;

-signals inputted to or outputted from an electrical sensor, such as for example the output signal of the clogging sensor 30 of the filters;

-alarm signals or other signals emitted by electronic control units, such as for example the control unit of the main power motor 17.

[114] Such supply, discharge or delivery pressures, internal pressures, rotation speed or linear translation, internal temperatures, operating or activity times, torques or resisting forces may be for example those acting on the aforementioned rotary or linear motors, linear actuators and/or on the aforementioned pumps.

[115] The input physical quantities VIN provided in input to the neural network or other virtual simulator 200 may also advantageously comprise the derivatives - for example the first, second or higher-order derivatives- and the integrals with respect to the time of the input physical quantities directly detected by the input sensors, for example in particular the derivatives and the integrals with respect to the time of the temperatures, pressures, flow rates and other physical quantities mentioned above, achieving an effective data augmentation.

[116] The first index physical quantities VUNj provided in output to the neural network or other virtual simulator 200 and/or the second index physical quantities VURj detected by sensors to compare them with the first index physical quantities VUNj determined by the virtual simulator 200 can advantageously be one or more chosen for example from the following list:

-a pressure inside one of the hydraulic compensators 46, for example to the one fluidically connected to one of the motors 12 and 13 that drives the cutting wheel 10 or 11, or to the one fluidically connected to the sludge pump 14;

-the degree of contamination, ageing, of deterioration or quality of the oil or other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) ;

-the temperature of the oil or other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) , for example the temperature of the hydraulic oil in the general hydraulic circuit, i.e. in the circuit connecting the various motors, actuators and hydraulic pumps ;

-the temperature of the carpentry box of the rotary if the device to be monitored is a drilling machine;

-the voltage of the electric batteries that power the main power motor 17.

[117] As a consequence of the previously described drawbacks of the pressure compensator 46, it is in fact advantageous to apply a pressure transducer, i.e. a particular example of an index sensor, hydraulically or in any case fluidically connected to the compensators 46, which can measure the compensation pressure present inside the compensator 46.

[118] This pressure is preferably also the pressure present in the "housing chamber of the motor" .

[119] By comparing the internal pressure of the compensator 46 when it is at a certain depth, with the pressure of the digging fluid present in the digging measured at the same depth (by means of a further pressure sensor fixed to the frame 9) an important indication on the correct operation of the compensator is obtained.

[120] More particularly, this value of the pressure inside the compensator can be the value of the compensation pressure that is being generated or that one has in all environments, housings of actuators, motors, pumps and of mechanisms, devices and sections of hydraulic system fluidically connected to the compensator, whereby this value is indicative of the correct operation of all the aforesaid components.

[121] In particular the internal pressure of the compensator 46 should preferably always be very close to the pressure measured in the digging.

[122] The pressure value inside the compensator 46, measured by the pressure transducer, i.e. index sensor, can then be processed with the residual error calculation algorithm described above, which compares this measured value with the value of "pressure of the compensator" 46 under nominal operating conditions determined by the neural network or other supervised learning virtual simulator .

[123] Consequently, in addition, the milling machine 100, drilling machine 100' or other device to be monitored can be provided with one or more of the following index sensors:

-a pressure sensor configured for detecting the pressure inside a compensator 46;

-a sensor of the degree of contamination, ageing, deterioration or quality of the oil or of other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) ;

-a temperature sensor of the oil or other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) , fo r example the temperature of the hydraulic oil in the general hydraulic circuit, i.e. in the circuit connecting the various motors, actuators and hydraulic pumps ;

-a rotary body temperature sensor if the device to be monitored is a drilling machine.

[124] More generally, the index physical quantities in output to the neural network or other virtual simulator 200 and/or detected through sensors to compare them with those determined by the virtual simulator 200 can be one or more chosen for example from the following list:

- a temperature of a component for example hydraulic, electrical or electronic;

-depth at which the digging tool 3 is located;

-speed of descent or of ascent of the digging tool 3;

-temperature of the external environment;

-temperature of an electrical panel;

-weight of the digging tool 3 during the execution of a digging;

- tensioning of the rope or other suspension element 7 for example during the execution of a digging;

-fuel consumption and/or oil pressure of the main power motor 17;

-inclination of the arm 6;

[125] Such supply, discharge or delivery pressures, internal pressures, rotation speed or linear translation, internal temperatures, operating or activity times, torques or resisting forces may be for example those acting on the aforementioned rotary or linear motors, linear actuators and/or on the aforementioned pumps.

[126] The neural network or other virtual simulator 200 described above may be implemented through one or more logic units, each comprising for example a microprocessor circuit.

[127] Such logic units may for example be part of an electronic computer mounted on board the milling machine 100 or other device to be monitored 100 or of a remote electronic computer, located for example tens, hundreds or thousands of metres or kilometres from the milling machine 100 or other device to be monitored 100 and connected to the latter for example via a cloud.

[128] The neural network or other virtual simulator 200 described above may be implemented, for example, by a computer program which, run on the aforesaid one or more logic units, implements at least some steps of the process described above.

[129] The neural network 200 manages to simulate in a very faithful and precise way the index physical quantities of a milling machine, drilling machine or other device to be simulated 100, following a generally shorter and less challenging learning phase of the construction of virtual simulators of another type, in particular of those known, based for example on the simulation of the individual components of a machine or other complex system, thus allowing a considerable saving of time and labour.

[130] Through the same or similar neural architecture, it is also possible to faithfully simulate milling machines, drilling or other operating machines 100, even very different ones, simply by changing the learning phase and the example situations provided during such learning .

[131] In addition, the neural simulator 200 can be continuously improved and modified at relatively low costs by retraining it with new examples of operation that can be continuously collected, via a data transmission system for example known per se, from a fleet of milling machines 100 or other operating machines operating on construction sites or otherwise in operation in the field.

[132] It can be considered that the greater precision and sensitivity of the predictions and simulations of the neural simulator 200 are due, among other things, to its ability to process and exploit a large amount of physical quantities and other input signals, even several hundreds or thousands, exploiting synergies, concomitances and correlations otherwise unnoticed among the various operating parameters of a complex system such as an operating machine; as well as to the ease of exploiting real data, coming from the field, for the learning of the neural network.

[133] Through the process and monitoring system described above, it was possible to develop a very reliable predictive maintenance system capable of more effectively optimizing the maintenance interventions, reducing both the replacements of components still in good state of preservation and machine downtimes due to unexpected faults.

[134] The monitoring process and system described above are able to predict the behaviour of a foundation machine 100 by identifying the occurrence of failures or faults more in advance than known predicting methods.

[135] In addition to optimizing the maintenance interventions, this also allows to use critical or higher-value components, i.e. more expensive to replace, for a longer period of time than what is cautiously indicated in the current preventive maintenance intervals with a considerable overall saving for the user .

[136] The embodiments described above are susceptible to numerous modifications and variants, without departing from the scope of the present invention.

[137] For example, the above teachings can also be applied to devices to be monitored 100 other than milling machines for diaphragm walls, and for example also to drilling machines for piles or other foundations, a possible example of which is shown in Figure 10 and indicated with the overall reference 100' .

[138] Every reference in this description to "an embodiment", "an example of embodiment" means that a particular characteristic or structure described in relation to such embodiment is comprised in at least one embodiment of the invention and in particular in a particular variant of the invention as defined in a main claim .

[139] The fact that such expressions appear in various passages of the description does not imply that they are necessarily referred solely to the same embodiment.

[140] In addition, when a feature, element or structure is described in relation to a particular embodiment, it is observed that it is within the competence of the person skilled in the art to apply such feature, element or structure to other embodiments.

[141] Numerical references which only differ in terms of different superscripts 21' , 21", 21¹¹¹ unless specified otherwise indicate different variants of an element with the same name.

[142] Furthermore, all of the details can be replaced by technically equivalent elements.

[143] For example, the materials used, as well as the dimensions thereof, can be of any type according to the technical requirements. [144] It must be understood that an expression of the type "A comprises B, C, D" or "A is formed by B, C, D" also comprises and describes the particular case in which "A consists of B, C, D" . [145] The expression "A comprises a B element" unless otherwise specified is to be understood as "A comprises one or more elements B".

[146] References to a "first, second, third, ... n-th entity" have the sole purpose of distinguishing them from each other but the indication of the n-th entity does not necessarily imply the existence of the first, second ... (n-l)th entity.

[147] The examples and lists of possible variants of the present application are to be construed as non- exhaustive lists.

Claims

1) Process for monitoring the operation of a device to be monitored (100, 100' ) comprising the following steps:

- providing a virtual simulator (200) programmed or otherwise configured for simulating, through a supervised learning artificial intelligence algorithm, the operation of the real device to be monitored (100, 100' ) at least under nominal conditions;

- providing one or more input sensors, each being configured for detecting a plurality of input physical quantities (VIN ) of the device to be monitored (100, 100 ' ) and one or more index sensors, each being configured for detecting one or more index physical quantities (VURj) of the device to be monitored (100, 100' ) ;

- determining a simulated value (VUNj) of the index physical quantities by the virtual simulator (200) based on the plurality of input physical quantities (VIN ) detected by the one or more input sensors;

- comparing said simulated value of the index physical quantities (VUNj) with the value of the index physical quantities (VURj) detected by the one or more index sensors .

2) Process according to claim 1, comprising the step of evaluating whether the device to be monitored (100, 100' ) operates or not under nominal conditions based on comparing

A) the one or more index physical quantities determined by the virtual simulator with

B) the corresponding one or more index physical quantities detected by the one or more index sensors.

3) Process according to claim 1 or 2, wherein the virtual simulator (200) comprises a neural network.

4) Process according to one or more preceding claims, wherein the step of comparing the value of one or more index physical quantities simulated by the virtual simulator (200) with the value of such physical quantity detected by the one or more index sensors comprises in turn the following steps:

5.1) determining a difference between the simulated index physical quantity and the physical quantity directly measured by the corresponding one or more index sensors ;

5.2) determining an absolute value of such difference, also referred to as "residual error" in the present disclosure ;

5.3) evaluating a condition of anomalous operation based on said residual error.

5) Process according to claim 4, wherein step S.3) comprises the step of evaluating a condition of nominal or non-nominal operation based on the first, second or higher-order derivative of the residual error with respect to the time.

6) Process according to one or more preceding claims, comprising the step of determining the time frequency of the non-nominal operation episodes.

7) Process according to one or more preceding claims, comprising the step of determining and/or verifying the trend over time of a residual error.

8) Process according to one or more preceding claims, wherein the input physical quantities of said plurality are selected from the following list:

- an internal or external temperature or internal or external pressure of a component of the device to be monitored (100, 100') ;

- a resisting torque or force counteracting the driving force of a motor, pump or other rotary or linear actuator of the device to be monitored (100, 100') ;

- a rotation speed or linear actuation speed of a motor, pump or other rotary or linear actuator of the device to be monitored (100, 100') ;

- an operating or inactivity time of a component of the device to be monitored (100, 100') ;

- the depth at which a digging tool (3) of the device to be monitored (100, 100') is located;

- the temperature of the environment outside the device to be monitored (100, 100') ;

- the temperature, pressure, flow rate or flowing speed, degree of contamination, degree of ageing, of deterioration or quality of the oil or other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) ;

- a pressure, flow rate or flow speed or temperature at or near an inlet and/or outlet or rather delivery port of a pump, rotary or linear motor or other rotary or linear actuator of the device to be monitored (100, 100' ) ;

- the weight of a digging tool (3) forming part of the device to be monitored (100, 100') and configured for milling or otherwise digging a ground;

- an inclination with respect to the vertical axis or to a horizontal direction or plane of a component of the device to be monitored (100, 100') or of the device to be monitored (100, 100') as a whole;

- one or more signals inputted to an electrical sensor of the device to be monitored (100, 100') or outputted from said sensor, such as for example the output signal of a clogging sensor (30) of a filter for purifying hydraulic oil or other fluid of a hydraulic circuit;

- an alarm signal or other signal emitted by an electronic control unit forming part of the device to be monitored (100, 100' ) .

9) Process according to one or more preceding claims, wherein the index physical quantities simulated by the virtual simulator (200) and detected by the one or more index sensors are selected from the following list:

- an internal or external temperature or internal or external pressure of a component of the device to be monitored (100, 100' ) ;

- a resisting torque or force counteracting the driving force of a motor, pump or other rotary or linear actuator of the device to be monitored (100, 100' ) ;

- a rotation speed or linear actuation speed of a motor, pump or other rotary or linear actuator of the device to be monitored (100, 100' ) ;

- an operating or inactivity time of a component of the device to be monitored (100, 100' ) ;

- the depth at which a digging tool (3) of the device to be monitored (100, 100' ) is located;

- the temperature of the environment outside the device to be monitored (100, 100' ) ;

- the temperature, pressure, flow rate or flowing speed, degree of contamination, degree of ageing, deterioration or quality of the oil or other fluid of a hydraulic circuit of the device to be monitored (100, 100' ) ;

- a pressure, flow rate or flow speed or temperature at or near an inlet, supply, or delivery port, or rather return, outlet, or discharge port, of a pump, rotary or linear motor or other rotary or linear actuator of the device to be monitored (100, 100' ) ;

- the weight of a digging tool (3) forming part of the device to be monitored (100, 100' ) and configured for milling or otherwise digging a ground;

- an inclination with respect to the vertical axis or to a horizontal direction or plane of a component of the device to be monitored (100, 100' ) or of the device to be monitored (100, 100' ) as a whole;

- one or more signals inputted to an electrical sensor of the device to be monitored (100, 100' ) or outputted from said sensor, such as for example the output signal of a clogging sensor (30) of a filter for purifying hydraulic oil or other fluid of a hydraulic circuit;

10) Process at least according to claim 8 and 9, wherein an input sensor does not operates, at least in the same instant or measurement interval, also as an index sensor.

11) P rocess according to one or more preceding claims, wherein the device to be monitored (100, 100' ) is selected from the following list: a milling machine (100) for digging diaphragm walls in a soil, a drilling machine (100' ) fo r piles or other foundations, an operating machine .

12) Process according to one or more preceding claims, wherein said value of the index physical quantities (VURj) detected by the one or more index sensors at least before being compared with said simulated value of the index physical quantities (VUNj) is not provided as entering the virtual simulator (200) or otherwise processed by said simulator.

13) Monitoring system for monitoring the operation of a device to be monitored (100, 100' ) comprising:

- a virtual simulator (200) programmed or otherwise configured for simulating the operation of the device to be monitored (100, 100' ) real at least under nominal conditions by a supervised learning artificial intelligence algorithm;

- one or more input sensors, each being configured for detecting a plurality of input physical quantities of the device to be monitored (100, 100' ) ;

- one or more index sensors, each being configured for detecting one or more index physical quantities of the device to be monitored (100, 100' ) ; the system being programmed or otherwise arranged for:

- determining a simulated value of the index physical quantities by the virtual simulator (200) based on the plurality of input physical quantities detected by the one or more input sensors;

- comparing said simulated value of the index physical quantities with the value of the index physical quantities detected by the one or more index sensors.

14) System according to claim 13, comprising a logic unit programmed or otherwise configured for carrying out the virtual simulator (200) .

15) Monitored system, comprising a monitoring system according to claim 13 or 14 and the device to be monitored (100, 100' ) .

16) Computer program comprising a computer programming code and configured for implementing a process according to one or more claims 1-12 performing at least the following steps: - carrying out a virtual simulator (200) programmed or otherwise configured for simulating the operation of the device to be monitored (100, 100' ) real at least under nominal conditions by a supervised learning artificial intelligence algorithm;

- determining a simulated value of the index physical quantities by the virtual simulator (200) based on the plurality of input physical quantities detected by the one or more input sensors; - comparing said simulated value of the index physical quantities with the value of the index physical quantities detected by the one or more index sensors.