Method for monitoring an installation

Abstract

Method for Monitoring an Installation. The invention relates to a method for monitoring an installation (10), comprising the following steps: - receiving initial data, - preprocessing the initial data, - homomorphic encryption of the preprocessed initial data, - homomorphic processing of the encrypted preprocessed initial data, the homomorphic processing yielding a result, and - sending the result to a monitoring device (22) for the installation (10). Each initial data point has a respective significant portion, the preprocessing comprising transforming each initial data point into a respective natural number, the respective natural number preserving the respective significant portion. The invention further relates to an associated computer program product. Figure for the abstract: Fig 1

Description

Title of the invention: Method for monitoring an installation

[0001] The present invention relates to a method for monitoring an installation.

[0002] An installation, for example industrial, has different operating parameters at a given time.

[0003] These parameters may be observed in order to monitor the operation of the installation.

[0004] For example, in the case of an industrial installation having fluid conduits, it is, for example, possible to monitor the pressure of the fluid circulating in the conduits at different locations. This makes it possible in particular to detect a leak, and to plan maintenance at the leak level.

[0005] Certain aspects of the installation are however likely to be more complex than such an example, for example in the case of predictive maintenance for which indications of a future problem must be identified, of carrying out a carbon assessment of the installation or of monitoring the energy consumption of the installation and possibly optimizing such consumption.

[0006] However, the parameters useful for studying these different aspects are likely to be confidential. Thus, it is not possible to subcontract the monitoring of the installation to an external third party for certain complex monitoring aspects.

[0007] An aim of the invention is therefore to propose a method for monitoring an installation for which part of the monitoring is likely to be subcontracted to an external third party.

[0008] To this end, the invention relates to a method for monitoring an installation, comprising the following steps:

[0009] - reception of initial data,

[0010] - preprocessing of the initial data,

[0011] - homomorphic encryption of the preprocessed initial data,

[0012] - homomorphic processing on the encrypted pre-processed initial data, the processing homomorphic giving a result, and

[0013] - sending the result to a monitoring device of the installation,

[0014] wherein each initial data item has a respective significant part, the preprocessing comprising transforming each initial data item into a respective natural number, the respective natural number retaining the respective significant part.

[0015] Homomorphic encryption allows encrypted data to be shared with an external third party. Preprocessing allows data to be obtained that allows faster homomorphic processing, while retaining the respective significant part of each data item. Thus, the result is representative of the appearance of the installation at monitor, so that the monitoring carried out by the installation monitoring device is relevant to the installation.

[0016] The monitoring method may furthermore have one or more of the following characteristics, considered individually or in all technically possible combinations:

[0017] - the method comprises a step of measuring and / or determining the data initials before receipt of the initial data;

[0018] - each initial data has a timestamp, the corresponding timestamp at the time of measurement or determination of the corresponding initial data, the corresponding timestamp being associated with each preprocessed initial data then with each preprocessed and encrypted initial data;

[0019] - the initial pre-processed encrypted data of the homomorphic processing present a timestamp included in a time interval less than an intrinsic time constant of the installation;

[0020] - the method comprises a step of decrypting the result and a step exploitation of the result in the monitoring of the installation;

[0021] - the method comprises a step of post-processing the decrypted result;

[0022] - the initial data preprocessing step includes truncation or rounding of at least one of the initial data, more particularly of each of the initial data, to its respective significant part and / or a change of calculation basis;

[0023] - at least one of the initial data is intended to be included in an interval predetermined, the predetermined interval comprising a lower value, the step of preprocessing the initial data being such that the lower value of said initial data is transformed into zero by the preprocessing; and / or

[0024] - the method comprises a reinforcement learning step comprising the following sub-steps:

[0025] - sub-step of providing examples of model data,

[0026] - carrying out homomorphic encryption on the model data,

[0027] - implementation of homomorphic processing on encrypted model data, the homomorphic treatment giving a typical result,

[0028] - carrying out at least one pre-processing test of the model data,

[0029] - realization of homomorphic encryption of the model data preprocessed by the au less one pretreatment test,

[0030] - implementation of homomorphic processing on preprocessed model data and encrypted, the homomorphic treatment giving an approximate result,

[0031] - comparison of the typical result and the approximate result with respect to a margin of predetermined error, and

[0032] - validation or not of at least one pre-treatment test depending on the comparison of the typical result and the approximate result against the predetermined margin of error.

[0033] The invention further relates to a computer program product comprising the software instructions which, when implemented by computer equipment, implement the monitoring method according to any one of the preceding claims.

[0034] Other characteristics and advantages of the invention will appear on reading the following description of embodiments of the invention, given by way of example only and with reference to the drawings, in which:

[0035] [Fig-1] [Fig. 1] is a schematic view of an installation and equipment for monitoring the installation, according to a first embodiment of the invention,

[0036] [Fig.2] [Fig.2] is a schematic block view of an exemplary monitoring method according to an embodiment of the invention, and

[0037] [Fig.3] [Fig.3] is a schematic view of an installation and equipment for monitoring the installation, according to a second embodiment of the invention.

[0038] An installation 10 according to a first example is shown in [Fig.l].

[0039] The installation 10 is, for example, an industrial installation, for example in the energy industry.

[0040] The installation is, for example, an electricity production installation.

[0041] The installation comprises, for example, a control module 12 of the installation 10.

[0042] The control module 12 is capable of controlling at least one piece of equipment of the installation 10, for example with regard to at least one operating parameter.

[0043] The installation further comprises, for example, at least one sensor 14 capable of measuring at least one parameter of the installation.

[0044] The at least one parameter is, for example, the temperature at a given location, the pressure of a fluid in a pipe, a rotation speed of a turbine and / or any other measurable element relevant for monitoring the installation.

[0045] The at least one sensor 14 associates each measurement with the time of the measurement relative to an internal clock.

[0046] A calculation module 16 is connected to the installation 10 and, more particularly, to the control module 12 and / or to at least one sensor 14.

[0047] The calculation module 16 is in the confidentiality sphere of the installation 10.

[0048] The control module 12 communicates, for example, the at least one operating parameter to the calculation module 16. Advantageously, the control module 12 also communicates, for each operating parameter, the time corresponding to said operating parameter, here using a clock of the control module 12.

[0049] The at least one sensor 14 communicates, for example, the at least one measured parameter to the calculation module 16, as well as advantageously the associated measurement time.

[0050] The calculation module 16 is capable of carrying out calculations on data, here on at least one operating parameter and / or on at least one measured parameter.

[0051] The calculations include, for example, preprocessing of the data and homomorphic encryption of the preprocessed data.

[0052] In the example shown, the calculation module 16 is furthermore capable of receiving calculation instructions from an external module 18.

[0053] The external module 18 does not, for example, belong to the confidentiality sphere of the installation 10. Alternatively, the external module 18 is in said confidentiality sphere.

[0054] The calculation module 16 is connected to a processing module 20.

[0055] The calculation module 16 is capable of sending data, for example preprocessed and encrypted, to the processing module 20, which is able to receive them.

[0056] The processing module 20 is here external to the installation 10.

[0057] In particular, the processing module 20 is, for example, not within the confidentiality sphere of the installation 10.

[0058] The processing module 20 is adapted to carry out processing, more particularly, homomorphic processing on data, in particular comprising received data, and to give a result of the homomorphic processing.

[0059] Homomorphic processing corresponds, for example, to a calculation relating to a carbon balance or a calculation relating to energy consumption and the optimization of said consumption.

[0060] The processing module 20 is capable of sending said result to a monitoring device of the installation 22.

[0061] The processing module 20 comprises, for example, at least one server.

[0062] The monitoring device of the installation 22 is, for example, included in the installation 10.

[0063] Alternatively, it is external to the installation.

[0064] The installation monitoring device 22 is, for example, capable of deciphering the result transmitted by the processing module 20 and of exploiting the result.

[0065] In addition, the monitoring device of the installation 22 is here capable of post-processing the decrypted result.

[0066] The exploitation of the result corresponds to the use of the result within the framework of the operation of the installation.

[0067] More particularly, in the case cited, the exploitation of the result includes, for example, the production of a complete carbon balance or a consumption balance energy consumption of the installation and possibly optimization of said consumption and / or an adaptation of the operation of the installation depending on the result, for example by modifying an operating parameter.

[0068] The control module 12, the calculation module 16 and the external module 18 are each produced in the form of software, or a software brick, executable by a processor.

[0069] In a variant not shown, the control module 12, the calculation module 16 and the external module 18 are each produced in the form of a programmable logic component, such as an FPGA (Field Programmable Gate Array), or in the form of a dedicated integrated circuit, such as an ASIC (Application Specific Integrated Circuit).

[0070] When the control module is produced in the form of one or more software programs, that is to say in the form of a computer program, it is also capable of being recorded on a medium, not shown, that is readable by a computer. The computer-readable medium is, for example, a medium capable of storing electronic instructions and of being coupled to a bus of a computer system. By way of example, the readable medium is an optical disk, a magneto-optical disk, a ROM memory, a RAM memory, any type of non-volatile memory (for example EPROM, EEPROM, FLASH, NVRAM), a magnetic card or an optical card. A computer program comprising software instructions is then stored on the readable medium.

[0071] A method for monitoring an installation will now be described with regard to [Fig.2], in the case shown in [Fig.l].

[0072] The monitoring method comprises the following steps:

[0073] - reception of initial data 110,

[0074] - preprocessing of the initial data 120,

[0075] - homomorphic encryption of the preprocessed initial data 130,

[0076] - homomorphic processing 140 on the encrypted pre-processed initial data, the homomorphic processing giving a result, and

[0077] - sending the result 150 to a monitoring device of the installation.

[0078] In the example shown, the steps of receiving initial data 110, preprocessing the initial data 120 and homomorphic encryption of the preprocessed initial data 130 are, for example, implemented by the calculation module 16.

[0079] The initial data comprises, for example, the values ​​of the at least one calculation parameter of the control module 12 and / or of the at least one parameter measured by the at least one sensor 14.

[0080] The method comprises, for example, a step of measuring and / or determining the initial data before receiving the initial data.

[0081] Said step is carried out within the installation 10, for example, at least partially by the control module 12 and / or the at least one sensor 14.

[0082] Each initial data item has a timestamp, the timestamp corresponding to the time of measurement or determination of the corresponding initial data item.

[0083] The timestamp is, for example, received with the corresponding initial data.

[0084] By way of example, it is considered, for example, that the installation comprises a hydraulic turbine capable of converting the displacement energy of a fluid into rotational energy intended to be converted into electrical energy.

[0085] The initial data include, for example, the rotational speed of the hydraulic turbine and the area of ​​the passage section of said turbine.

[0086] The rotation speed is, for example, measured by a sensor at the turbine. Each rotation speed measurement is associated with the time of the measurement, corresponding to the time stamp.

[0087] The rotation speed measurements are transmitted to the calculation module 16.

[0088] The passage section of the turbine is, for example, adjustable by the control module 12. Thus, the surface area of ​​the passage section at each instant can be determined as a function of the commands sent.

[0089] By way of illustration, the control module 12 transmits to the calculation module 16 the commands relating to the passage section of the turbine.

[0090] The calculation module 16 then calculates the surface area of ​​the passage section of the turbine at each instant when it has a rotation speed measurement and associates with each calculated surface area value the timestamp of the corresponding measurement.

[0091] Thus, the calculation module 16 receives the initial data comprising the rotation speed of the hydraulic turbine and the surface area of ​​the passage section of said turbine, here with the associated time stamps.

[0092] The initial data are, for example, real numbers.

[0093] Each initial data has a respective significant part.

[0094] The respective significant part of each initial data item is determined beforehand, for example using instructions provided by the external module 18, as will be described later.

[0095] The preprocessing comprises the transformation of each initial data into a respective natural integer, the respective natural integer retaining the respective significant part.

[0096] The instructions for implementing the preprocessing are determined by the external module 18 and communicated to the calculation module 16 which implements them.

[0097] Advantageously, the corresponding timestamp is associated with each preprocessed initial data item, that is to say that the preprocessing comprises the association with each preprocessed data item of the timestamp of the corresponding initial data item.

[0098] The step of preprocessing at least one of the initial data comprises a truncation or rounding of said initial data to its respective significant part and / or a change of calculation basis.

[0099] Additionally or alternatively, at least one of the initial data is, for example, intended to be included in a predetermined interval, the predetermined interval comprising a lower value, the step of preprocessing said initial data being such that the preprocessing of the lower value is equal to zero. The possible values ​​of said data greater than the lower value are replaced, during the preprocessing state, by the natural integers starting from 1 by incrementing at each increment of the significant part.

[0100] By way of illustration, in the example cited above, an initial datum is the rotation speed of a hydraulic turbine. The sensor performs, for example, the measurement in revolutions per minute to the nearest hundredth.

[0101] It is, for example, previously determined that the significant part of the initial data of the rotation speed corresponds to the rotation speed truncated to the tenth, that is to say that the significant part of a rotation speed of 1376.12 revolutions per minute is 137 tens of revolutions per minute.

[0102] Alternatively, we consider rounding to the nearest ten. The significant part would then be 138 tens of revolutions per minute.

[0103] In one embodiment, the preprocessing then replaces, in the example given, the initial data of the rotation speed by 137 or 138 depending on the choice made.

[0104] Alternatively, it is further provided that the rotation speed is between 1200 and 1800 revolutions per minute.

[0105] The pretreatment is then such that 1200 rpm is replaced by 0, 1210 rpm by 1, 1220 rpm by 2, etc., up to 1800 rpm by 60.

[0106] The preprocessing then replaces an initial data of the rotation speed having 137 tens as significant part by 17.

[0107] Alternatively, it is envisaged that the significant part of the initial data of the rotation speed corresponds to the rotation speed rounded or truncated to the nearest tenth of a revolution, that is to say that the significant part of a rotation speed of 1376.12 revolutions per minute is 13761 tenths of revolutions per minute.

[0108] In one embodiment, the preprocessing then replaces, in the given example, the initial rotation speed data with 13761.

[0109] The pre-processed initial data is then encrypted using homomorphic encryption, here within the installation 10, more particularly by the calculation module 16.

[0110] Encryption is, for example, an asymmetric cryptography encryption technique.

[0111] More particularly, the encryption is carried out by a public key, here by the calculation module 16.

[0112] The public key may be transmitted, for example, to the processing module 20 and / or to the external module 18.

[0113] The decryption of the data encrypted by said encryption is carried out using another key, called the private key.

[0114] The private key is not communicated, in particular to the external module 18 and / or to the processing module 20. More particularly, the private key remains within the installation 10.

[0115] Advantageously, the corresponding timestamp is associated with each preprocessed and encrypted initial data.

[0116] The timestamp is, for example, not encrypted or is encrypted by non-homomorphic encryption. This is particularly the case when no homomorphic calculation takes place on the timestamp during the homomorphic processing step.

[0117] The pre-processed and encrypted initial data are then transmitted from the calculation module 16 to the processing module 20.

[0118] The processing module 20 here carries out the homomorphic processing step 140.

[0119] The initial preprocessed encrypted data of the homomorphic processing have a timestamp included in a time interval less than an intrinsic time constant of the installation. Said time constant is, for example, between 50 ms, for example for very fast processes such as rotating machines, and one hour, for example for slower processes such as filling storage tanks.

[0120] Homomorphic processing gives a result.

[0121] Said result is then transmitted to the monitoring device 22 of the installation.

[0122] Within the installation 10, here the monitoring device 22, a step 160 of decryption of the result is, for example, carried out, more particularly using the private key mentioned previously.

[0123] The method comprises, for example, a post-processing step 170 of the decrypted result.

[0124] The post-processing step comprises post-processing of the result adapted to reverse the modifications due to the pre-processing step on the initial data.

[0125] Post-processing may or may not include the implementation of a final part of the algorithm to overcome certain limitations of homomorphic calculation.

[0126] The instructions for implementing the post-processing are determined by the external module 18. They are here communicated to the monitoring device 22.

[0127] The method finally comprises a step 180 of exploitation of the result in the monitoring of the installation 10.

[0128] As mentioned previously, the exploitation of the result corresponds to the use of the result within the framework of the operation of the installation.

[0129] The exploitation of the result includes, for example, the production of a complete carbon balance or a balance sheet on the energy consumption of the installation and possibly optimization of said consumption.

[0130] Alternatively or additionally, the exploitation of the result comprises an adaptation of the operation of the installation according to the result, for example by modifying at least one operating parameter.

[0131] Here is now described an example of determining the respective significant part and possibly instructions for implementing the preprocessing.

[0132] This includes a reinforcement learning step comprising the following substeps:

[0133] - sub-step of providing examples of model data,

[0134] - carrying out homomorphic encryption on the model data,

[0135] - implementation of homomorphic processing on encrypted model data, the homomorphic treatment giving a typical result,

[0136] - carrying out at least one pre-processing test of the model data,

[0137] - realization of homomorphic encryption of the model data preprocessed by the au less one pretreatment test,

[0138] - implementation of homomorphic processing on preprocessed model data and encrypted, the homomorphic treatment giving an approximate result,

[0139] - comparison of the typical result and the approximate result with respect to a margin of predetermined error, and

[0140] - validation or not of at least one pre-treatment test depending on the comparison of the typical result and the approximate result against the predetermined margin of error.

[0141] The model data here comprises a plurality of sets of values ​​likely to correspond to the initial data.

[0142] In the example cited above, the model data comprises a plurality of rotational speed values ​​and flow section areas, each speed value being associated with a flow section area having the same timestamp, i.e. having a timestamp included in a time interval less than the intrinsic time constant.

[0143] The model data includes, for example, here are, past data values ​​provided by the installation 10 corresponding to the initial data.

[0144] In the example cited above, the model data includes, for example, old values ​​of the rotation speed and the flow section area of ​​the installation's turbine, for example from a month ago.

[0145] Additionally or alternatively, the model data are examples of random or predetermined values ​​for the initial data.

[0146] Additionally or alternatively, the model data is generated from a dynamic modeling of the installation.

[0147] At least one pre-processing test is determined, here by the external module 18, as well as the associated pre-processing instructions.

[0148] More particularly, a way of determining the significant part is chosen for each model data. In the above-mentioned example, it is for example decided that the rotation speed of the turbine is rounded to the nearest hundred revolutions per minute.

[0149] The preprocessing associated with said significant part is also determined, here by the external module 18.

[0150] In the above-mentioned example, this is, for example, rounding the data to the nearest hundred: 1584.25 revolutions per minute would then become 16 according to the pre-processing.

[0151] Alternatively, provided that the rotation speed is between 1200 and 1800 revolutions per minute, the pretreatment test is then such that 1200 revolutions per minute is replaced by 0, 1300 revolutions per minute by 1, 1400 revolutions per minute by 2, etc., up to 1800 revolutions per minute by 6. 1584.25 revolutions per minute would then become 4 depending on the pretreatment.

[0152] The model data are then preprocessed according to each preprocessing test.

[0153] The model data, as well as the preprocessed model data according to the or each preprocessing test, are encrypted with homomorphic encryption, here according to the public key.

[0154] The homomorphic processing is implemented on the encrypted model data and on the encrypted preprocessed model data for the or each preprocessing test.

[0155] Homomorphic processing on encrypted model data gives a model result.

[0156] The homomorphic processing on the preprocessed model data encrypted for the or each preprocessing test gives a respective test result.

[0157] The respective test result is, for example, post-processed to reverse the changes due to the pre-processing step on the initial data.

[0158] Each test result is compared to the model result to determine the corresponding margin of error, here for all of the plurality of sets of values ​​included in the model data.

[0159] The margin of error is compared to an acceptable error threshold, such that the pre-processing test(s) having a respective margin of error greater than, or strictly greater than, the acceptable error threshold are rejected.

[0160] In the above-mentioned example, we obtain, for example, a margin of error strictly greater than the acceptable error threshold in the or each pre-treatment considered in which the rotation speed is rounded to the nearest hundred. This or these pre-treatments are therefore rejected.

[0161] The pre-treatment test(s) having a respective margin of error strictly lower, or less than, the acceptable error threshold are validated.

[0162] In the case where no pretreatment test is validated, these steps are, for example, repeated with other pretreatment tests until at least one validated pretreatment test is obtained.

[0163] In the above-mentioned example, we will, for example, consider a pre-treatment in which the rotation speed is rounded to the nearest ten.

[0164] In the case where only one pretreatment test is validated, then it is selected. Alternatively, these steps are repeated with other pretreatment tests to validate other tests.

[0165] In the case where several pre-processing tests are validated, then the pre-processing test is selected for which the model data pre-processed by this pre-processing allows the fastest execution of the homomorphic calculation.

[0166] The selected preprocessing test is applied as preprocessing to the initial data preprocessing step.

[0167] In one embodiment, constraints are defined for the prior determination of the pre-treatment test(s).

[0168] In particular, the significant part of each data item is such that the data interacting with each other during the homomorphic processing have respective significant parts that are consistent with the mathematical rules relating to the significant digits of a number.

[0169] For example, in the case of an addition between two data, the sum must not have more decimals than the number which has the fewest. Thus, if during homomorphic processing, two data are added, the respective significant value of the two data will be such that they have the same decimals (for example, by rounding them to the nearest ten)

[0170] The pre-processing test(s) then take into account said constraints on the significant part of each data item.

[0171] This avoids considering pre-processing tests in which figures of a data item are retained in the significant part, whereas they will not be taken into account after addition to less precise data. Thus, this limits the data size to what is useful for homomorphic processing.

[0172] Preprocessing thus makes it possible to limit the size of the data, so that, during homomorphic processing, the calculation time is reduced compared to the case where there is no preprocessing. This thus makes it possible to obtain a result more quickly and to be able to carry out calculations on an installation during operation. In particular, it is, on the one hand, possible to have the result quickly so as to quickly exploit this result, for example by adapting the operation of the installation. On the other hand, it is possible to carry out the homomorphic processing often, for example at regular intervals, without causing calculation failure due to lack of computing power.

[0173] Encryption provides the possibility of subcontracting the processing, which is particularly advantageous in the case where the installation is incapable of carrying out said processing.

[0174] In an alternative embodiment, the assembly is devoid of an external module capable of sending calculation instructions to the calculation module, the calculations carried out by the calculation module according to the embodiment previously described being carried out for example directly by the calculation module or by the processing module which then transmits the instructions to the calculation module.

[0175] An alternative embodiment will now be described with reference to [Fig.3].

[0176] Only the differences with the embodiment of [Fig.l] will now be described.

[0177] The embodiment differs in particular in that there is a plurality of installations 210, 230, for example two installations.

[0178] There may be more than two installations 210, 230.

[0179] Each installation 210, 230 is, for example, such as the installation 10 described with regard to the embodiment of [Fig.l].

[0180] Each installation 210, 230 is likely to comprise a control module 212, 232 and / or at least one sensor 214, 234 as described previously.

[0181] The external module 218 is connected to each of the calculation modules 216, 236 of the installations 210, 230.

[0182] Each calculation module 216, 236 is connected to the processing module 220.

[0183] The processing module 220 is capable of receiving initial data from each of the installations 210, 230.

[0184] Each of the installations 210, 230 is, for example, capable of transmitting a respective portion of the initial data, for example here a first portion by the installation 210 and a second portion by the installation 230.

[0185] The external module 218 is capable of transmitting instructions relating to preprocessing to each of the calculation modules 216, 236.

[0186] For each installation 210, 230, the respective portion of the initial data is preprocessed, for example by the corresponding calculation module 216, 236.

[0187] Then, each respective portion of the preprocessed initial data is encrypted according to a respective homomorphic encryption.

[0188] More particularly, each installation 210, 230 implements a homomorphic encryption different from that of the other installations. In particular, each installation 210, 230 implements a homomorphic encryption with a respective public key, corresponding to a decryption with a respective private key, the respective public key and the respective private key of each installation being different from the public and private keys of the other installations.

[0189] Each respective pre-processed and encrypted portion is then communicated to the processing module 220 which implements the homomorphic processing, which gives a result.

[0190] The result is communicated to the monitoring devices 222, 242 of each installation 210, 230.

[0191] In one embodiment, the entire result is communicated to each monitoring device 222, 242 of each installation 210, 230.

[0192] Alternatively, only a respective portion of the result is communicated to each monitoring device 222, 242.

[0193] This thus allows a plurality of installations 210, 230, whose data interact for example in the operation of the installations, to be able to analyze their data without providing the values ​​to the external installations.

[0194] Since the private key of each installation is unknown to the other installations, confidentiality remains guaranteed.

[0195] In an alternative embodiment, the assembly is devoid of an external module capable of sending calculation instructions to the calculation module, the calculations carried out by the calculation module being carried out for example directly by the processing module.

[0196] The invention further relates to a computer program product comprising the software instructions which, when implemented by computer equipment, implement the monitoring method as described above.

[0197] Data preprocessing allows for faster homomorphic processing, so that it is possible to perform such homomorphic processing on encrypted data in cases where this would not be possible on the initial data, particularly for reasons of computation time.

[0198] This makes it possible in particular to adapt the operation of an installation quickly based on the analysis of recent past data or to analyze data at regular intervals.

Claims

Claims

1. Method for monitoring an installation (10), comprising the following steps: - receiving initial data (110), - preprocessing the initial data (120), - homomorphic encryption of the preprocessed initial data (130), - homomorphic processing on the encrypted preprocessed initial data (140), the homomorphic processing giving a result, and - sending the result (150) to a monitoring device (22) of the installation (10), characterized in that each initial data item has a respective significant part, the preprocessing (120) comprising the transformation of each initial data item into a respective natural integer, the respective natural integer retaining the respective significant part, and in that the method further comprises a reinforcement learning step comprising the following substeps: - substep of providing examples of model data, - carrying out the homomorphic encryption on the model data,- implementing the homomorphic processing on the encrypted model data, the homomorphic processing giving a typical result, - carrying out at least one pre-processing test of the model data, - carrying out the homomorphic encryption of the pre-processed model data by the at least one pre-processing test, - implementing the homomorphic processing on the pre-processed and encrypted model data, the homomorphic processing giving an approximate result, - comparing the typical result and the approximate result with respect to a predetermined margin of error, and - validating or not the at least one pre-processing test depending on the comparison of the typical result and the approximate result with respect to the predetermined margin of error.,

2. A monitoring method according to claim 1, wherein if several pre-treatment tests are validated, the method comprises a step of selecting the pre-treatment test for which the model data preprocessed by this preprocessing allow the fastest execution of the homomorphic calculation.

3. A monitoring method according to claim 1 or claim 2, wherein the selected pre-processing test is applied as a pre-processing to the step of pre-processing the initial data.

4. A monitoring method according to any preceding claim, comprising a step of determining at least one pre-treatment test.

5. Monitoring method according to the preceding claim, wherein constraints are defined for the determination of said at least one pre-treatment test.

6. Monitoring method according to any one of claims 1 to 5, comprising a step of decrypting the result (160) and a step of exploiting the result (180) in the monitoring of the installation (10).

7. Monitoring method according to claim 6, comprising a step of post-processing the decrypted result (170).

8. Monitoring method according to any one of claims 1 to 7, wherein the step of preprocessing the initial data (120) comprises a truncation or rounding of at least one of the initial data, more particularly of each of the initial data, to its respective significant part and / or a change of calculation basis.

9. A monitoring method according to any one of claims 1 to 8, wherein at least one of the initial data is intended to be included in a predetermined interval, the predetermined interval comprising a lower value, the step of preprocessing the initial data (120) being such that the lower value of said initial data is transformed into zero by the preprocessing.

10. A computer program product comprising software instructions which, when implemented by computer equipment, implement the monitoring method according to any one of the preceding claims.

Images