Method and device for predicting drift of hardware

Abstract

The method (70) for predicting drift of a physical quantity comprises a step (71) of an operator intervening on hardware causing a variation in the value of this physical quantity and a step (72, 75) of regularly measuring this physical quantity. The method additionally comprises: - a first step (73) of analysing the values for a first predetermined duration following the intervention, - a calibration step (74) that delivers at least one limit value of this physical quantity, on the basis of the result of this first analysis, - a second step (76) of analysing the values for a second sliding duration, - a step (77) of extrapolating a future value of this physical quantity, on the basis of the result of this second analysis, and - a step (78, 79) of determining whether the future extrapolated value crosses one of the limit values within a predetermined future time interval and, if so, a step (80) of triggering an alert.

Description

[0001] DESCRIPTION

[0002] TITLE: METHOD AND DEVICE FOR PREDICTING DRIFT OF EQUIPMENT

[0003] Technical field of the invention

[0004] The present invention relates to a method and a device for predicting drift in equipment. It applies, in particular, to regulation equipment (for example, flow, pressure, temperature or level) subject to drift (for example, valves, regulators, pumps, compressors) and whose measurements of the regulated quantities are available "continuously" or at fairly short time intervals with respect to the drift dynamics of the equipment. More particularly, the present invention applies to a gas pressure reduction station.

[0005] Prior art

[0006] Anomaly detection algorithms operate with a learning phase that requires a large amount of data, both measurements of physical quantities (e.g., pressure) and anomalies (e.g., poor regulations), to establish a link between these data. In practice, however, the number of anomalies collected is too low and the materials considered too heterogeneous to allow sufficient learning for this type of algorithm.

[0007] Time series forecasting algorithms are designed to predict periodic (e.g., seasonal) phenomena. However, the drifts observed on industrial equipment do not correspond to such patterns. In addition, external interventions such as maintenance actions carried out by operators (e.g., bias correction, setting modification) create discontinuities in the time series, which limit their learning potential.

[0008] In the particular case of gas pressure relief stations, operators regularly intervene on these stations, in particular to correct deviations and / or change settings, which has the following disadvantages:

[0009] - potentially excessive deviations in equipment between two external interventions, with certain issues regarding the availability of installations, safety and / or the environment;

[0010] - external interventions on equipment when this is not always necessary and

[0011] - external on-call interventions, for example at night or on public holidays, which are more expensive.

[0012] Document US 2022 / 197271 is known, which describes a method for predicting drift of a physical quantity regulated by regulation equipment. This method comprises a step of statistical analysis of measurements of this physical quantity following a readjustment by the regulation equipment.

[0013] Statement of the invention

[0014] The present invention aims to remedy all or part of these drawbacks. To this end, according to a first aspect, the invention aims at a method for predicting drift of a physical quantity, comprising a step of intervention by an operator on equipment causing a variation in the value of this physical quantity, a step of regular measurement of this physical quantity providing a set of values ​​of this physical quantity, a method which further comprises:

[0015] - a first stage of analysis of the values ​​of this physical quantity during a first predetermined duration following the intervention stage,

[0016] - a calibration step providing at least one limit value of this physical quantity, depending on the result of this first analysis,

[0017] - a second stage of analysis of the values ​​of this physical quantity during a second sliding period after the intervention stage,

[0018] - a step of extrapolation of a future value of this physical quantity, based on the result of this second analysis, and

[0019] - a step of determining whether the future extrapolated value crosses one of the limit values ​​in a predetermined future time interval, and if so, a step of triggering an alert.

[0020] Thanks to these provisions, without using long-term learning, the process automatically determines an abnormality criterion for a material drift, after an external intervention on the material, then a prediction of the occurrence of this abnormality. This process thus makes it possible to decide on a new preventive intervention to avoid an excessively significant drift from occurring. It also makes it possible to intervene on the material only when a drift is predicted. Finally, it avoids, or at least limits, the interventions of on-call operators, which are more costly.

[0021] Note that the first period over which the first analysis is performed is predetermined by default. However, it is possible to extend it depending on the results, for example if the variance is too large, in order to obtain a more "robust" calibration. Similarly, the second rolling period can be dynamic and adapt according to at least one indicator of "stability" of the trends. As for the triggered alert, it can take the form of emails sent automatically to the operational teams concerned.

[0022] In embodiments, during the first analysis step, an average and a standard deviation of the values ​​of the physical quantity are calculated and, during the calibration step, at least one limit value is equal to the calculated average to which a product of the measured standard deviation and a predetermined number is added or subtracted. The average thus calculated may also be increased or decreased by a certain percentage, in particular when limit values ​​of the physical quantity can be determined by a margin of error around the adjustment setpoint.

[0023] The product of the standard deviation and a predetermined number is added to the average in the case of the maximum limit value and subtracted from the average in the case of the minimum limit value. Thanks to these provisions, the detection criterion implemented depends on the initial stability of the values ​​of the physical quantity monitored.

[0024] In embodiments, during the second analysis step, a linear regression of the values ​​of the physical quantity is performed. In embodiments, during the second analysis step, a non-linear regression is performed, for example implementing an exponential smoothing or AutoRegressive Integrated Moving Average (ARIMA) type regression model.

[0025] In embodiments, during the second analysis step, only a selection of values ​​of the physical quantity are processed during the second sliding period after the intervention step.

[0026] Thanks to each of these provisions, a measurement of the trend evolution of the physical quantity is obtained, without consuming significant computing power. A single server can thus implement the method on a large number of equipment subject to drift. In practice, the method of the invention can be implemented once a day, so as not to overload the communication networks. In addition, if there are alerts, they can all be sent together, for example at the start of the day, which facilitates the organization of the day's maintenance schedules.

[0027] In embodiments, the method comprises, prior to the intervention, a step of providing values ​​of the physical quantity measured on equipment similar to the equipment on which the intervention is carried out and / or on the equipment on which the intervention is carried out, to a machine learning model trained to extrapolate a value of the physical quantity based on the measurements carried out on this equipment and, during the second analysis and extrapolation steps, this trained machine learning model is implemented to obtain the future value.

[0028] This benefits from the extrapolation accuracy obtained using a trained machine learning model.

[0029] In embodiments, the method comprises, prior to the intervention, a step of providing values ​​of the physical quantity measured on equipment similar to the equipment on which the intervention is carried out and / or on the equipment on which the intervention is carried out, to a machine learning model trained to extrapolate a value of the physical quantity as a function of the measurements carried out on this equipment and, during the first analysis and calibration steps, this trained machine learning model is implemented to obtain at least one limit value.

[0030] This allows for the determination of limit values ​​and / or a forecast based on artificial intelligence. It is noted that it is thus possible to improve forecasts by carrying out training with other available data, for example gas consumption or meteorological data.

[0031] In embodiments, learning from previous regressions and alerts confirmed or denied by maintenance operators makes it possible to define a confidence indicator for future alerts.

[0032] In embodiments, during the extrapolation step, a future value is extrapolated for the end of a third duration subsequent to the second duration.

[0033] A horizon is thus set, equal to this third duration. It is at this horizon, that is to say at a predetermined deadline, that we determine whether the physical quantity risks becoming abnormal, that is to say crossing a limit value. For example, this horizon is a number of hours or days. Note that this horizon can vary depending on the availability of a maintenance operator, so that a forecast of abnormal drift can trigger preventive intervention by an operator.

[0034] In embodiments, the third duration depends on the day and / or time of the second duration.

[0035] For example, this horizon is a maintenance operator's usual working day. This horizon is thus set to cross the last two days of the week and public holidays in order to be on a working day. The operator can thus intervene before these days which would impose an on-call duty.

[0036] In embodiments, during the extrapolation step, a duration subsequent to the second duration is measured before crossing a limit value.

[0037] Thus, the expected duration before the drift becomes abnormal is evaluated and makes it possible to schedule an operator intervention well in advance. This duration represents a degree of urgency of the maintenance operation. In particular, if several pieces of equipment are subject to a drift forecast according to the method that is the subject of the invention, the order of the next operator interventions can be determined according to this degree of urgency and, possibly, other criteria, such as the seriousness of the drift according to the equipment involved (for example, whether or not there is a backup regulation line or according to the importance of the flow rates concerned), the accessibility of the equipment, the proximity to other equipment or the availability of a maintenance operator.

[0038] According to a second aspect, the present invention relates to a device for predicting the drift of a physical quantity following an intervention by an operator on equipment causing a variation in the value of this physical quantity, which comprises a means for regularly measuring this physical quantity providing a set of values ​​of this physical quantity, characterized in that it further comprises:

[0039] - a first means of analyzing the values ​​of this physical quantity during a first predetermined duration following the intervention,

[0040] - a means of calibration providing at least one limit value of this physical quantity, depending on the result of this first analysis,

[0041] - a second means of analyzing the values ​​of this physical quantity during a second sliding period after the intervention,

[0042] - a means of extrapolating a future value of this physical quantity, based on the result of this second analysis, and

[0043] - a means of triggering an alert if the future extrapolated value crosses one of the limit values ​​within a predetermined future time interval.

[0044] The advantages, aims and particular characteristics of this device being similar to those of the method which is the subject of the invention, they are not recalled here.

[0045] In embodiments, the hardware is control hardware configured to control a flow rate, pressure, temperature, or level.

[0046] In embodiments, the hardware is control hardware configured to control a physical quantity of a valve, an expander, a pump or a compressor. The inventor has determined that the invention is particularly well-suited to predicting these physical quantities and / or controlling this control hardware.

[0047] According to a third aspect, the present invention relates to a gas pressure relief station comprising a device which is the subject of the invention, in which the equipment regulates the pressure at the outlet of a regulating valve.

[0048] The inventor noted the effectiveness of implementing the invention for such a relaxation station.

[0049] Brief description of the figures

[0050] Other advantages, aims and particular characteristics of the invention will emerge from the following non-limiting description of at least one particular embodiment of the method and device which are the subject of the invention, with reference to the appended drawings, in which:

[0051] [Fig 1] schematically represents a particular embodiment of a device which is the subject of the invention installed in a gas pressure reduction and / or delivery station,

[0052] [Fig 2] represents a change in gas pressure at the outlet of a gas pressure reduction and / or delivery station and a determination of limit values ​​for the change in gas pressure at the outlet of a gas pressure reduction and / or delivery station,

[0053] [Fig 3] represents a first forecast of the evolution of gas pressure at the outlet of a gas pressure reduction and / or delivery station,

[0054] [Fig 4] represents a second forecast of the evolution of gas pressure at the outlet of a gas pressure reduction and / or delivery station,

[0055] [Fig 5] represents a third forecast of gas pressure evolution at the outlet of a gas pressure reduction and / or delivery station,

[0056] [Fig 6] represents, in the form of a flowchart, steps of a particular embodiment of the invention applied to the regulation of pressure at the outlet of a pressure reduction station,

[0057] [Fig 7] represents a supervision interface for a network of pressure relief and / or gas delivery stations, and

[0058] [Fig 8] represents, in the form of a flowchart, steps of a particular embodiment of the method which is the subject of the invention.

[0059] Description of the embodiments

[0060] This description is given without limitation, each characteristic of an embodiment being able to be combined with any other characteristic of any other embodiment in an advantageous manner.

[0061] Since the transport of natural gas and / or biomethane is carried out at high pressure (for example 70 bar) and the distribution in town at low pressure (for example 5 bar), a pressure reducing station is intended to reduce the gas from high pressure to low pressure. At each supply point of the network as well as at each delivery point, metering systems are used to measure the quantities of natural gas and / or biomethane which pass through these points. Figure 1 shows an installation 10 comprising a pressure reducing station 13. This pressure reducing station 13 has the function of lowering the pressure of the gas distributed by a general network 11 to pressure levels usable by the different types of customers, for example domestic, tertiary or industrial, connected to a distribution network 12. The pressure reducing and metering station ("PDC") is an installation used to control the pressure and metering of natural gas.In addition to the upstream 11 and downstream 12 gas pipes, this station 13 comprises various components: pressure regulator equipment 14, pressure measurement sensor 15 downstream of the regulator 14, meter (not shown), volume corrector (not shown), safety valves (not shown), valves (not shown), filter (not shown), data processing device 16 and data transmission device 17.

[0062] The pressure regulator 14 is subject to a drift which has the effect of varying the downstream pressure in the downstream gas pipe 12. However, this downstream pressure must remain within a range of values, between a minimum value ensuring the operation of the customer's installations and a maximum value avoiding the risks of leakage at the level of these installations or of rupture of the pipe 12. A maintenance operation, on site or remotely controlled, makes it possible to correct this drift by modifying the values ​​of operating parameters of the pressure regulator 14. The present invention makes it possible to anticipate the need for such maintenance operations.

[0063] Figure 2 shows a time graph 20 of pressure downstream of a pressure regulator. The successive values ​​of the pressure are represented by curve 21. These successive values ​​are obtained either by a continuous measurement of the downstream pressure, or by regular measurements of this pressure. At a time 23, an intervention is carried out by an operator on the regulator, which causes a discontinuity in curve 21. In the case shown in Figure 2, this is an increase in the downstream pressure.

[0064] During a first predetermined duration 22 following the instant 23 of the intervention, extending until the instant 24, a first analysis of the measured values ​​of the pressure is carried out. This first duration is also called a “calibration window”. This first analysis is, for example, statistical and determines provides an average value 25, a standard deviation of the variations of this pressure and a maximum amplitude of these variations.

[0065] A lower limit value 26 and an upper limit value 27 are then set, depending on the result of the first analysis. According to a first example, the lower limit value 26 is equal to the average value reduced by a predetermined percentage, for example five percent. According to a second example, the lower limit value 26 is equal to the average value reduced by a predetermined number of times the standard deviation, for example three times. Alternatively, the average thus calculated is increased or decreased by a certain percentage, in particular when limit values ​​of the physical quantity can be determined by a margin of error around the adjustment setpoint. According to a third example, the lower limit value 26 is equal to the average value reduced by a predetermined proportion of the amplitude, for example half the amplitude and by a predetermined percentage of the average value or by a predetermined number of times the standard deviation.Of course, any other combination of the results of the first analysis can be used to set the lower limit value 26. Similarly, the upper limit value 27 is determined in one of these ways by replacing the decrease with an increase.

[0066] As illustrated in Figures 3 and 4, a second analysis of the pressure values ​​is also carried out during a second sliding duration, 28 in Figure 3 and 31 in Figure 4, subsequent to the intervention step. This second analysis is, for example, statistical and provides a time function, 29 in Figure 3 and 32 in Figure 4, which can extend after the end of the second sliding duration. This second sliding duration 28 or 31 is also called a “sliding window”. It ends at the current time. It should be noted that the second duration can overlap the first duration, as long as the duration between the intervention step and the current time is less than the sum of the first duration and the second duration.

[0067] The pressure values ​​subsequent to this sliding window are therefore not known when processing the pressure values ​​in this sliding window. This time function for modeling the evolution of the pressure therefore allows extrapolation into the future of the pressure value. For example, the second analysis performs a regression, for example linear, as in the embodiment illustrated in Figures 3 and 4. The time modeling function is therefore a linear function, the curve 29 or 32 of which is a straight line. According to another example, the second analysis performs a non-linear regression, for example it implements a regression model of the exponential smoothing or Auto Regressive Integrated Moving Average (ARIMA) type. In embodiments, this analysis step is performed on a selection of values ​​of the physical quantity during the second sliding duration 28 or 31 subsequent to the intervention step.For example, we deal with daily percentiles and the second analysis only deals with the extreme values ​​of each day.

[0068] Based on this time function, a future value, 30 in Figure 3 and 33 in Figure 4, of the pressure is extrapolated. In the embodiment illustrated in Figures 3 and 4, the future value 30 or 33 is the value of the time function 29 or 32 after a third predetermined duration 36 starting at the end of the sliding window. This third predetermined duration is also called a "horizon". In other embodiments, this future value is equal to one of the limit values ​​26 and 27 and the duration 37 between the end of the sliding window and the instant when the time function 29 or 32 is equal to one of the limit values ​​26 and 27 is determined.

[0069] If the future extrapolated value crosses one of the limit values ​​26 or 27 in a predetermined future time interval, the device 13 triggers an alert, locally or remotely. In the case of using a horizon 36, this condition is met if the future value 30 or 33 is greater than the upper limit value 27 or less than the lower limit value 26. In the case of determining the duration 37 before crossing a limit value 26 or 27, this condition is met if this duration is less than the duration of the predetermined future time interval.

[0070] In the variant illustrated in Figure 5, instead of or in addition to the time function 32, the same processing is carried out on at least one of:

[0071] - time function 34, with the same slope as function 32 but shifted so that all pressure values ​​in the sliding window are greater than the value of function 34, - time function 35, with the same slope as function 32 but shifted so that all pressure values ​​in the sliding window are less than the value of function 35.

[0072] This variant makes it possible to take into account the amplitude of variation of the pressure observed in the sliding window 31. In variants, time functions offset from the function 32 are applied by adding and subtracting half the amplitude of variation of the pressure in the sliding window. Other variants are based on a predetermined number of times of the standard deviation of the values ​​of the pressure in the sliding window 31, and / or an amplitude or a standard deviation calculated over a duration other than that of the sliding window 31.

[0073] As illustrated in Figure 6, an embodiment 40 of the method for predicting drift of a physical quantity controlled by equipment, for example regulation equipment, comprises different steps starting at the end of an intervention 41 on the equipment. During a step 42, a horizon is determined. In embodiments, this horizon is fixed, for example equal to one or more days. In embodiments, this horizon varies according to the intervention time of an operator, depending on his place of work or residence or instantaneous geographical position and the travel time necessary for him to reach the equipment subject to the intervention.In embodiments, this horizon varies over time, for example, it depends on the day and / or the time of the sliding window, so that this horizon, greater than a minimum value, for example one day, does not end outside intervention time slots (for example between eight a.m. and six p.m.), nor on a Saturday or a Sunday, nor on a public holiday.

[0074] During a step 43, a regular measurement of the pressure downstream of the regulating device is carried out. This regular measurement provides a set of values ​​of this pressure.

[0075] During a step 44, a first step of analyzing the values ​​of this pressure is carried out during a first predetermined duration following the intervention, i.e. the calibration window.

[0076] It is noted that the first analysis duration is predetermined by default, but that it is possible to extend it depending on the results, for example if the variance is too large, in order to obtain a more "robust" calibration. In variants, step 44 therefore comprises a determination of the variance of the pressure value and, if this variance is greater than a predetermined value, an extension of the first duration, the first analysis then being carried out over the first extended duration.

[0077] During a step 45, a calibration is carried out providing at least one limit value of this pressure, depending on the result of the first analysis.

[0078] During a step 46, the regular measurement of the pressure downstream of the regulating device continues.

[0079] During a step 47, a second analysis of the pressure values ​​is carried out during a sliding window subsequent to the intervention step.

[0080] It is noted that the second analysis duration is predetermined by default, but that it is possible to extend it according to an instability indicator, in order to obtain a more "robust" second analysis. In variants, step 47 therefore comprises a determination of the instability of the trend in the evolution of the pressure value and, if this instability is greater than a predetermined value, an extension of the second duration, the second analysis then being carried out over the first extended duration.

[0081] Note that the second duration can overlap the first duration, as long as the duration between the intervention step and the current time is less than the sum of the first duration and the second duration. Since steps 44 and 47 are independent, step 47 can then end before step 44 and step 45.

[0082] During a step 48, an extrapolation of the future value of the pressure on the horizon is carried out, based on the result of this second analysis.

[0083] During a step 49, the future value at the horizon is compared with each limit value. During a step 50, it is determined whether one of the limit values ​​is crossed by the extrapolation at the horizon. If the result of step 50 is negative, we return to step 46. If the result of step 50 is positive, that is to say if the future extrapolated value crosses one of the limit values ​​in the horizon (predetermined future time interval), during a step 51, a remote alert is triggered, and possibly locally, and we return to step 46. At the same time, during a step 52, an analysis of the alert is carried out remotely. For example, we determine whether meteorological conditions can explain this alert and whether the weather forecasts predict a return to normal conditions.Finally, during a step 53, an intervention decision is made, for example a scheduling of interventions on several pieces of equipment, based on the shortest path to carry out the interventions. Then, we return to step 41, as soon as this intervention is carried out.

[0084] In particular, if several pieces of equipment are subject to a drift forecast according to the method which is the subject of the invention, the order of the next operator interventions can be determined according to this degree of urgency and, possibly, other criteria, such as the seriousness of the drift according to the equipment involved (for example, whether or not there is a backup regulation line or according to the importance of the flow rates concerned), the accessibility of the equipment, the proximity to other equipment or the availability of a maintenance operator.

[0085] In the embodiment illustrated in Figure 6, during the first analysis step 44, an average and a standard deviation of the values ​​of the physical quantity are calculated and, during the calibration step 45, at least one limit value is equal to the calculated average added to the product of the measured standard deviation by a predetermined number.

[0086] In the embodiment illustrated in Figure 6, during the second analysis step 47, a regression, for example linear, of the values ​​of the physical quantity is carried out.

[0087] In the embodiment illustrated in Figure 6, during the extrapolation step 48, a future value is extrapolated for a horizon, i.e. a third duration following the second duration. Preferably, the horizon depends on the day and / or the time of the sliding window. Alternatively, during the extrapolation step 48, a duration following the sliding window is measured before crossing a limit value.

[0088] In the variants described below, a machine learning model is used. This makes it possible to improve the determination of limit values ​​and / or extrapolation by carrying out training with pressure values, but also other available data, for example gas consumption on the network or meteorological data.

[0089] In these variants, the method comprises, prior to the intervention, a step of providing values ​​of the physical quantity measured on equipment similar to the equipment on which the intervention is carried out and / or on the equipment on which the intervention is carried out, to a machine learning model trained to extrapolate a value of the physical quantity based on the measurements carried out on this equipment.

[0090] Of course, the extrapolation is all the more reliable if the measured values ​​used for learning were measured on the equipment 14 to which the method applies and / or on similar equipment with a similar operating context. Similarly, the extrapolation is more reliable if the measured values ​​are measured after an intervention by an operator and are supplemented with values ​​measured before this intervention. In the latter case, when implementing the trained machine learning model, it is provided, in addition to the pressure measurements taken after the intervention, with pressure measurements taken before the intervention.

[0091] In a first variant, during the first analysis 44 and calibration 45 steps, this trained machine learning model is implemented to obtain at least one limit value.

[0092] In a second variant, possibly combined with the first variant, during the second analysis 47 and extrapolation 48 steps, this trained machine learning model is implemented.

[0093] Examples of such machine learning models include a recurrent neural network (RNN) or a convolutional neural network (CNN). These neural networks are particularly suited to cases where additional information, such as consumption and weather data, is incorporated.

[0094] In embodiments, learning from previous alerts, confirmed or denied by maintenance operators, makes it possible to define a confidence indicator for future alerts, according to known techniques.

[0095] As illustrated in Figure 7, an interface 60 for viewing alerts may take the form of a map 61 on which the positions of the equipment subject to interventions, 62 to 66, are represented. Associated with the representation, by an outer circle, of each of these pieces of equipment is a color representation of the duration before crossing a limit value obtained by the calibration of this equipment. In Figure 7, the darker the central circle, the shorter this duration. A digital representation 67 of this duration is also displayed below the representation of each piece of equipment. This double representation of the urgency of an intervention allows an operator to prepare a tour of interventions on the equipment 62 to 66 represented on the map 61, for example according to the emergencies and the access routes to this equipment.

[0096] In Figure 8, we observe a particular embodiment 70 of a method which is the subject of the invention, more generally applied to a physical quantity controlled by equipment, for example regulation equipment. During a step 71, an operator carries out an intervention on this equipment causing a variation in the value of this physical quantity. During a step 72, a regular measurement of this physical quantity is carried out providing a set of values ​​of this physical quantity.

[0097] During a step 73, a first analysis of the values ​​of this physical quantity is carried out for a first predetermined duration following the intervention step.

[0098] During a step 74, a calibration is carried out providing at least one limit value of this physical quantity, depending on the result of this first analysis.

[0099] During a step 75, the regular measurement of the value of the physical quantity is continued.

[0100] During a step 76, a second step of analysis of the values ​​of this physical quantity is carried out during a second sliding period subsequent to the intervention step.

[0101] Note that the second duration may overlap with the first duration, as long as the duration between the intervention step and the current time is less than the sum of the first duration and the second duration. Since steps 73 and 76 are independent, step 76 may then end before step 73 and step 74.

[0102] During a step 77, an extrapolation of a future value of this physical quantity is carried out, based on the result of this second analysis.

[0103] In step 78, the future values ​​are compared with the limit values.

[0104] During a step 79, it is determined whether one of the limit values ​​is crossed. If the result of step 79 is positive, that is to say if the future extrapolated value crosses one of the limit values ​​in the horizon (predetermined future time interval), during a step 80, an alert is triggered and a return is made to step 75. At the same time, during a step 81, an analysis of the alert is carried out remotely. Finally, during a step 82, a decision to intervene is made. Then, a return is made to step 71, as soon as this intervention is carried out.

[0105] Of course, all the technical characteristics, embodiments and variants set out with regard to figures 1 to 6 apply to the method 60 illustrated in figure 7.

[0106] Returning to Figure 1, the equipment is a regulating equipment, and more particularly a pressure regulator 14. The sensor 15 continuously measures the pressure downstream of the pressure regulator 14. The data transmission device 17 transmits to a remote server

[0107] 18 the pressure values ​​measured by the sensor 15 and / or the results of the processing carried out by the data processing device 16.

[0108] More generally, a device 19 which is the subject of the invention applies to industrial equipment. In particular, this equipment can be configured to regulate a measurable physical quantity, for example a flow rate, a pressure, a temperature or a level. For example, this equipment is configured to regulate a physical quantity of a control valve, a pressure reducer, a pump or a compressor, in particular of gas.

[0109] The data processing device 16 and the remote server 18, which are part of the device

[0110] 19 object of the invention, are jointly configured to carry out the steps of the method object of the invention and constitute, by the implementation of at least one computer program, at least:

[0111] - a means 15 for regular measurement of this physical quantity providing a set of values ​​of this physical quantity, characterized in that it further comprises: - a first means, 16 and / or 18, for analyzing the values ​​of this physical quantity during a first predetermined duration 22 following the intervention,

[0112] - a means, 16 and / or 18, of calibration providing at least one limit value, 26 and / or 27, of this physical quantity, depending on the result of this first analysis,

[0113] - a second means, 16 and / or 18, of analyzing the values ​​of this physical quantity during a second sliding period, 28 and / or 31 after the intervention,

[0114] - a means, 16 and / or 18, of extrapolating a future value, 30 and / or 33, of this physical quantity, depending on the result of this second analysis, and

[0115] - a means, 16 and / or 18, of triggering an alert if the future extrapolated value crosses one of the limit values ​​in a predetermined future time interval 36.

Claims

CLAIMS 1. Method (40, 70) for predicting drift of a physical quantity, comprising a step (41, 71) of intervention by an operator on equipment (14) causing a variation in the value of this physical quantity, a step (43, 46, 72, 75) of regular measurement of this physical quantity providing a set of values ​​of this physical quantity, characterized in that it further comprises: - a first step (44, 73) of analyzing the values ​​of this physical quantity during a first predetermined duration (22) following the intervention step, - a calibration step (45, 74) providing at least one limit value (26, 27) of this physical quantity, depending on the result of this first analysis, - a second step (47, 76) of analyzing the values ​​of this physical quantity during a second sliding duration (28, 31) subsequent to the intervention step, - a step (48, 77) of extrapolation of a future value (30, 33) of this physical quantity, as a function of the result of this second analysis, and - a step (49, 50, 78, 79) of determining whether the future extrapolated value crosses one of the limit values ​​in a predetermined future time interval (36), and if so, a step (51, 80) of triggering an alert.

2. Method (40, 70) according to claim 1, in which, during the first analysis step (44, 73), an average and a standard deviation of the values ​​of the physical quantity are calculated and, during the calibration step (45, 74), at least one limit value (26, 27) is equal to the calculated average to which a product of the measured standard deviation by a predetermined number is added or subtracted.

3. Method (40, 70) according to one of claims 1 or 2, in which, during the second analysis step (47, 76), a linear regression of the values ​​of the physical quantity is carried out.

4. Method (40, 70) according to one of claims 1 or 2, in which, during the second analysis step (47, 76), a non-linear regression is carried out, for example implementing a regression model of the exponential smoothing or Auto Regressive Integrated Moving Average (ARIMA) type.

5. Method (40, 70) according to one of claims 1 to 4, in which, during the second analysis step (47, 76), only a selection of values ​​of the physical quantity is processed during the second sliding duration (28, 31) subsequent to the intervention step.

6. Method (40, 70) according to one of claims 1 or 2, which comprises, prior to the intervention, a step of providing values ​​of the physical quantity measured on equipment similar to the material on which the intervention is carried out and / or on the material on which the intervention is carried out, to a machine learning model trained to extrapolate a value of the physical quantity based on the measurements carried out on this material and, during the second analysis (47, 76) and extrapolation (48, 77) steps, this trained machine learning model is implemented to obtain the future value (30, 33).

7. Method (40, 70) according to one of claims 1 to 6, which comprises, prior to the intervention (41, 71), a step of providing values ​​of the physical quantity measured on equipment similar to the equipment on which the intervention is carried out and / or on the equipment on which the intervention is carried out, to a machine learning model trained to extrapolate a value of the physical quantity as a function of the measurements carried out on this equipment and, during the first analysis (44, 73) and calibration (45, 74) steps, this trained machine learning model is implemented to obtain at least one limit value.

8. Method (40, 70) according to one of claims 6 or 7, which comprises a step of providing, by an operator, confirmation or denial of the alert, the trained machine learning model being configured to determine a confidence index for future alerts.

9. Method (40, 70) according to one of claims 1 to 8, in which, during the extrapolation step (48, 77), a future value is extrapolated for the end of a third duration (36) following the second duration (28).

10. Method (40, 70) according to claim 9, wherein the third duration (36) depends on the day and / or time of the second duration (28).

11. Method (40, 70) according to one of claims 1 to 10, in which, during the extrapolation step (48, 77), a duration (37) following the second duration (28) is measured before crossing a limit value.

12. Device (19) for predicting drift of a physical quantity following an intervention by an operator on equipment (14) causing a variation in the value of this physical quantity, which comprises a means (15) for regular measurement of this physical quantity providing a set of values ​​of this physical quantity, characterized in that it further comprises: - a first means (16, 18) for analyzing the values ​​of this physical quantity during a first predetermined duration (22) following the intervention, - a calibration means (16, 18) providing at least one limit value (26, 27) of this physical quantity, depending on the result of this first analysis, - a second means (16, 18) for analyzing the values ​​of this physical quantity during a second sliding duration (28, 31) subsequent to the intervention, - a means (16, 18) of extrapolating a future value (30, 33) of this physical quantity, as a function of the result of this second analysis, and - means (16, 18) for triggering an alert if the future extrapolated value crosses one of the limit values ​​in a predetermined future time interval (36).

13. The device (19) of claim 12, wherein the hardware is control hardware configured to control a flow rate, pressure, temperature or level.

14. Device (19) according to one of claims 12 or 13, in which the equipment is a regulating equipment configured to regulate a physical quantity of a valve, a regulator, a pump or a compressor.

15. Gas pressure relief station (13) comprising a device (19) according to one of claims 12 to 14, in which the equipment (14) regulates the pressure at the outlet of a regulating valve.

Images