Method for monitoring a corrugated board production plant

Abstract

The method provides for detecting at least one operating parameter of a functional unit of the device and calculating the current value of at least a first statistical function of the operating parameter in a current time window, the current value of the first statistical function defining a first coordinate of a point of current operation of the functional unit. A step is also provided for verifying whether the point of current operation is within a range of permissible values ​​of the first statistical function, the values ​​contained in the range of permissible values ​​corresponding to the correct operation of the functional unit. If the point of current operation is outside the range of permissible values, then determining the position of the point of current operation relative to the range of permissible values ​​and providing a statistical diagnosis of the cause of the deviation of the current value from the range of permissible values ​​based on the position.

Description

Technical Field

[0001] The present invention relates to corrugated cardboard production equipment and, more particularly, to a predictive diagnostic method for functional components or units of corrugated cardboard production equipment. Background Art

[0002] Corrugated board is made from multiple sheets of flat paper fed from a mother roll. It typically consists of at least one fluted sheet and two flat sheets (the so-called liners), with the fluted sheets positioned between them. The liners are bonded to the fluted sheets at the tops of the flutes, or crests. Generally, a corrugated board can include more than one fluted sheet. Typically, a flat sheet is positioned between each pair of fluted sheets.

[0003] Corrugated board production equipment generally includes one or more uncoilers for unwinding continuous rolls or sheets of flat paper, and one or more single-facers. Each single-facer converts a continuous flat paper sheet into a continuous corrugated paper sheet and connects the continuous corrugated paper sheet to a continuous flat paper sheet (the so-called liner). The composite continuous sheet emerging from the single-facer is fed to a duplexer, where a second liner is adhered to the composite continuous sheet. Generally speaking, the equipment may include one or more single-facers to feed one or more sheets consisting of a continuous corrugated paper sheet and a liner to the duplexer. The equipment also includes a section that processes the corrugated paper sheet from the duplexer, which is called the dry end to distinguish it from the section that includes the machine from the uncoiler to the duplexer (the so-called wet end). The dry end section typically includes a longitudinal cutting and creasing station, where the corrugated paper sheet is cut into continuous longitudinal strips.

[0004] The continuous longitudinal strip is further processed in order to produce a series of separate sheets, or so-called folds, ie strips folded in a zigzag manner according to the transverse cutting and creasing lines.

[0005] This equipment includes a number of functional units of various types. For example, an electric motor, a pump, a steam system, a glue feeder, a pressurized air system, etc. are provided.

[0006] Functional units may be subject to wear and tear and may become damaged. Interventions for the maintenance, repair or replacement of functional units of a production line can be very expensive, as the line may need to be stopped for extended periods. Downtimes result in production losses, impacting overall production costs. Considering that the profit margins on the material produced (corrugated cardboard) are very narrow, the increased costs caused by production losses due to repair or maintenance downtime can be very heavy for the user. Moreover, stopping the wet-end section results in a large amount of waste and long restart times, as the corrugated cardboard still in the machine must be completely discharged and rejected, and the hot section (single-facer, double-facer) needs to be brought back to the correct temperature before production can be restarted.

[0007] WO2019048437 discloses a novel method for monitoring the operation of corrugated cardboard production equipment. The method provides for detecting at least one operating parameter of a functional unit of the equipment, such as the current drawn by a motor. The current value of a statistical function of the operating parameter is then calculated within a current time window. The maximum and minimum values ​​of the same statistical function are calculated based on historical data for the operating parameter in question. By comparing the current value of the statistical function with the maximum and minimum values, a predictive diagnostic message is obtained. This innovative method represents an important support for equipment users and manufacturers, as it allows for targeted and timely maintenance interventions.

[0008] The ongoing search for improvements in equipment management requires finding more efficient methods that make predictive diagnostics more effective and maintenance and / or repair interventions more timely and targeted. Summary of the Invention

[0009] The present invention is based on the now discovered knowledge that the statistical functions described and used in the method disclosed in WO2019048437 can provide further useful information for predictive diagnostics of corrugated board production equipment.

[0010] In essence, a method for monitoring the operation of a corrugated board production plant having a plurality of functional units is provided. At least one operating parameter is detected for one or more of the functional units. By "operating parameter" is meant any measurable parameter relating to the operation of the functional unit. For example, an operating parameter of an electric actuator may be the absorbed current or the driving torque. An operating parameter of a hydraulic or pneumatic component may be a working fluid (liquid fluid or gaseous fluid, respectively). The method further comprises a step of calculating a current value of at least a first statistical function of the operating parameter in a current time window. The current value of the first statistical function defines a first coordinate of the currently operated point of the functional unit. The method further comprises a step of verifying whether the currently operated point is within a range of permissible values ​​of the first statistical function. A value contained in this range of permissible values ​​corresponds to a correct operation of the functional unit.

[0011] If the current operating point is outside the range of permissible values, the position of the current operating point relative to the range of permissible values ​​is determined. Based on this position, a statistical diagnosis of the reason why the current value deviates from the range of permissible values ​​(i.e., the reason why the functional unit is operating abnormally) is provided.

[0012] For example, in the case of an electric motor, the statistical function may represent the average value of the current drawn. Abnormal operation may cause the average value of the current drawn to be too high.

[0013] The method is based on the surprising finding that the position of the current operating point (i.e. its coordinates in a one-, two- or multi-dimensional space relative to the range of permissible current values ​​(whose coordinates can be represented by a corresponding statistical function)) not only indicates that a fault is imminent, but also provides an indication about the nature of the fault and therefore provides an indication about possible measures to be taken (i.e. the solution to be adopted) so that the functional unit operates correctly again (i.e. the current operating point is within the range of permissible values).

[0014] Even though it is possible in principle to use only one statistical function and thus a one-dimensional range of permissible values, it is preferred to use at least two statistical functions and thus detect the position of the operating point in a two-dimensional space. In this space, the range of permissible values ​​is defined by a surface, for example a rectangular shape, but this is not mandatory.

[0015] More than two statistical functions may also be used.

[0016] Further advantageous embodiments and features of the method according to the invention are described below and in the appended claims, which form an integral part of the present description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be better understood through the following description and the accompanying drawings, which illustrate non-limiting embodiments of the present invention. More specifically, in the drawings:

[0018] Figure 1A , 1B, 1C and 1D show parts of a corrugated board production plant, which are arranged in sequence along a board feeding path;

[0019] Figure 2 (A), 2(B), 2(C), and 2(D) show example graphs illustrating the predictive diagnostic methods disclosed herein;

[0020] Figure 3 (A) and 3(B) illustrate methods for limiting the range of allowable values ​​to an envelope of multiple ranges of allowable values ​​calculated in multiple learning cycles or stages;

[0021] Figure 4 shows an enlargement of the feeding area of ​​single-face corrugated cardboard on the bridge, wherein the associated motor is used to actuate the corrugated cardboard feeding belt;

[0022] Figures 5 to 10 yes Figure 4 A diagram of abnormal operation of the motor;

[0023] Figure 11 A diagram showing a system for accessing a database containing information about causes of abnormal operations of functional units of equipment; and

[0024] Figure 12The overview diagram is shown. DETAILED DESCRIPTION

[0025] In the illustrated embodiment, a device for producing double-wall corrugated board is described, i.e., two sheets of corrugated paper (so-called fluted paper) are inserted between two sheets of flat paper (so-called liners), and an intermediate sheet is inserted between the two fluted sheets. Furthermore, the device is configured to form two stacks of board on two adjacent upper stackers.

[0026] However, it should be understood that the features described below in connection with the predictive diagnostic system and method can also be used in systems having different numbers of single-facers, and thus suitable for producing corrugated board comprising different numbers of sheets. Furthermore, the stacking systems can be different, for example, they can be adapted to form only a single stack of sheets or more than two stacks of sheets. In other embodiments, the stacking system can include components for fan-folding the corrugated board without cutting it into individual sheets.

[0027] Similarly, the double-facer and longitudinal cutting and creasing stations described below by way of example only may be configured differently than described and illustrated herein.

[0028] With reference to the accompanying drawings, the apparatus comprises a first section 3 for producing a first single-faced corrugated board, a second section 5 for producing a second single-faced corrugated board, and a third section 7 for feeding the two single-faced corrugated boards together with a flat sheet to a double-facer 8 in section 9. Section 9 comprises the double-facer 8 and corresponding accessories. A composite corrugated board emerges from section 9, formed from the single-faced corrugated board and another flat sheet glued to it, this flat sheet forming the second liner of the composite corrugated board.

[0029] Downstream of the section 9 is provided a section 11 in which an apparatus for removing trimmings is arranged, and downstream of the section 11 is provided a cutting and creasing section 13 for longitudinally cutting and creasing the corrugated board coming from the section 9 comprising the double-facer 8 so as to cut the corrugated board into a plurality of longitudinal strips of corrugated board and to form creasing lines along the longitudinal extension of the individual longitudinal strips of the continuous corrugated board.

[0030] In the embodiment shown, purely by way of example, the apparatus 1 further comprises a section 15 for transversely cutting the corrugated cardboard strips coming from the cutting and creasing section 13 , a double conveyor 17 , and two areas 19A and 19B for stacking the cardboard sheets cut in the section 15 and fed by means of the double conveyor 17 .

[0031] In section 3, a single facer 21 is arranged. Single facers that can be used to produce single face corrugated board are known. Only the main elements of a single facer will be described below, and single facers are disclosed, for example, in US 8,714,223 or EP 1362691, the contents of which are included in this specification.

[0032] Briefly, the single-facer 21 may include a first corrugating roller 23, which cooperates with a second corrugating roller 25 and a pressure roller 27 or other pressure member to couple a flat sheet and corrugated board, as described below. A first flat sheet N1 is fed to the single-facer 21 from a first unwinder 29. The unwinder 29 may be configured in a known manner, and therefore will not be described in detail. The unwinder 29 may include two locations: a first unwind reel B1 from which the first flat sheet N1 is supplied, and a second waiting reel B1X, which is unwound upon completion of the first unwind reel B1.

[0033] The second flat paper sheet N2 is unwound from a second uncoiler 31, which may be substantially identical to the uncoiler 29, with a first reel B2 and a second waiting reel B2X arranged thereon from which the second paper sheet N2 is unwound, and the waiting reel B2X starting to unwind when the first unwinding reel B1 is completed.

[0034] The first flat paper sheet N1 is fed to the first corrugating roller 23 after having passed around the heating roller 33. The contact arc of the paper sheet N1 around the heating roller 33 can be modified to transfer more or less heat from the inside of the heating roller 33 (e.g., heated by steam circulating therein) to the flat paper sheet N1.

[0035] The first flat paper sheet N1 is corrugated by passing through the nip formed between the first corrugating roller 23 and the second corrugating roller 25. In this manner, the corrugated paper sheet N1 is obtained at the exit of the nip between the first corrugating roller 23 and the second corrugating roller 25. Suitable glue is applied to the ridges formed on the corrugated paper sheet by the gluing unit 35 so that the corrugated paper sheet N1 can be glued to the second flat paper sheet N2 fed together with the corrugated paper sheet N1 through the nip formed between the second corrugating roller 25 and the pressure roller 27.

[0036] The gluing unit 35 may include an applicator roller 36 that contacts the ridges of the corrugated paper sheet N1 driven around the second corrugating roller 25. The applicator roller 36 receives glue from a screen roller or a dispenser roller 38 that draws glue from a tank 40 or the like. The distance between the applicator roller 36 and the dispenser roller 38 may be adjusted to adjust the amount of glue applied to the corrugated paper sheet N1.

[0037] In some embodiments, the second flat paper sheet N2 can be fed around one or more rollers 37, 39 disposed between the second uncoiler 31 and the single-facer 21 to be heated. The arc of contact between the second flat paper sheet N2 and either or both of the rollers 37, 39 can be modified to vary the amount of heat transferred to the flat paper sheet N2 by the roller(s) 37, 39 before the second flat paper sheet N2 contacts the pressure roller 27. Furthermore, the pressure roller 27 can be internally heated to bond the first and second paper sheets N1 and N2 together under high pressure and high temperature conditions.

[0038] At the output of the single-facer 21, a single-face corrugated board NS formed by a first corrugated paper sheet N1 and a second flat paper sheet N2 is obtained. Figure 2 The grooves or ridges O formed on the first paper sheet N1 are glued to the surface of the second flat paper sheet N2 facing the corrugated paper sheet N1 by means of glue C applied to the grooves O by the gluing unit 35.

[0039] Downstream of the single-facer 21, a bridge 41 is arranged, which extends towards the section 5 and the subsequent sections 7 and 9 of the apparatus 1. On the bridge 41, blanks S of single-faced corrugated board NS can be formed, forming appropriate cumulative folds, making it possible to make the operating speed of the single-facer 21 at least partially independent of the operating speed of the downstream sections.

[0040] The single-faced corrugated cardboard NS is then fed along a first path unfolded above the bridge 41 to a heating roller 43 around which it can be guided in an adjustable arc so as to be properly heated before reaching the double-facer 8 of section 9 .

[0041] In the embodiment shown, the apparatus 1 comprises a second section 5 substantially identical to the section 3, and a single-facer similar to the single-facer 21, where a second single-faced corrugated board, still indicated by NS, is formed by means of another pair of sheets N4, N5 coming from a reel similar to the first uncoiler 29 and the second uncoiler 31. This second single-faced corrugated board NS is fed to a bridge 41 to form a blank S and to a double-facer 8 of the section 9, guided around a heated roller 45 substantially identical to the heated roller 43.

[0042] In other embodiments, the section 5 and the corresponding single-facer can be omitted. Vice versa, in other embodiments, more than two sections 3, 5 can be provided, each with a corresponding single-facer and unwinder for the sheets to form corresponding single-faced corrugated boards NS, which are then glued together by means of a double-facer 8 in section 9.

[0043] The flat sheet N3 is unwound from another unwinder 47 and preferably fed around heated rollers 49 to the double-facer section 9. Gluing units 51, 53 apply glue in a known manner to the spines of the respective corrugated sheets of the two single-faced corrugated sheets NS to glue them together and to the flat sheet N3, which will form the second liner of the composite corrugated sheet CC emerging from section 9, the first liner of which is formed by the second flat sheet N2.

[0044] The portion 9 comprising the double facer may be configured in a known manner and will not be described in detail herein. Exemplary embodiments of single facers are disclosed in US 7,291,243 and US 2012 / 0193026, the contents of which are incorporated herein and to which reference may be made for further details of embodiments of this portion of the apparatus.

[0045] In section 11, a transverse rotary shear 61 is arranged, which can perform transverse cuts to completely or partially sever the composite corrugated cardboard CC fed from section 9. Transverse rotary shear 61 can be configured, for example, as described in US Pat. No. 6,722,243, the contents of which are incorporated herein. As will be described in more detail below, transverse rotary shear 61 can be used, in particular, to remove portions of the corrugated cardboard CC that have gluing defects or other defects.

[0046] The composite corrugated cardboard CC fed through the cutting and creasing section 13 is divided into strips that can deviate along two paths defined by the two conveyors 17A, 17B of the dual conveyor 17. The cutting and creasing section 13 can be configured in a known manner, for example as disclosed in US Pat. No. 5,951,454, US Pat. No. 6,165,117, US Pat. No. 6,092,452, US Pat. No. 6,684,749, US Pat. No. 8,342,068 or other prior art documents cited in the above-mentioned patent documents, the contents of which are incorporated into this specification.

[0047] Two conveyors 17A, 17B convey corrugated cardboard obtained by transversely cutting a continuous strip of composite corrugated cardboard in section 15, so as to form stacks P1, P2 on stacking planes 63, 65. These are configured as known and disclosed, for example, in EP 1710183, US Pat. No. 5,829,951, or other patent documents cited therein, the contents of which are incorporated herein. Reference numeral 62 designates a station for transversely cutting the continuous strip of composite corrugated cardboard from the longitudinal cutting and creasing section 13. Station 62 includes transverse shears 62A, 62B, which cut each continuous strip from the cutting and creasing section 13 into individual sheets of a given length. The production line may include a transverse shear 62A, 62B for each conveyor 17A, 17B.

[0048] Each section or station of the apparatus 1 briefly described above comprises one or more functional units, each of which can be supplied with electricity, pressurized air, steam, glue, or other materials or fluids, or a combination thereof. For example, each single-facer includes one or more motors for controlling the rotation of the corrugating and pressure rollers, a steam supply system for heating the rollers, and a system for supplying glue to be applied to the ridges of the corrugated sheet. Furthermore, in addition to thrust or radial bearings, each section of the production line or apparatus 1 also includes motion-transmitting components such as belts, chains, shafts, and joints. Some stations include rotating components subject to wear, such as disc knives, linear or spiral blades, creasing tools, and the like.

[0049] Functional units are subject to wear and tear; therefore, over time, they require maintenance, repair, or replacement. According to one aspect described herein, in order to avoid or reduce failures that would result in lengthy downtimes and / or to better schedule replacement, maintenance, and repair interventions, a method for managing and controlling operating parameters of a plant 1 is provided, thereby allowing for predictive diagnostics of a production line or one or more functional units of the plant 1. Below, an embodiment of a predictive diagnostics method will be generally described, followed by specific examples of the method applied to a range of functional units of the plant 1 (as non-limiting examples only).

[0050] One or more functional units of the device 1 may include one or more sensors for detecting at least one operating parameter or multiple operating parameters of the functional unit. The sensors are used to acquire the values ​​of the operating parameters during the learning step. After the initial learning step, the sensors are used to acquire the current value of the (one or more) operating parameters in order to perform the control and predictive diagnosis steps of the functional unit by using the current value and the historical data related to the value of the same parameter previously obtained in the learning step. As will be better explained below, the historical data are continuously updated by providing that the learning step is not only performed within an initial time interval; instead, the learning step is continuously performed within a movable time window (hereinafter referred to as movable learning time window Δt2). In this way, the historical values ​​of the data used for predictive diagnosis are continuously updated.

[0051] In some embodiments, the initial learning step can be avoided and the values ​​of the same parameters associated with a device having similar characteristics and already in operation can be used as historical values. Essentially, the values ​​of the parameters associated with the same functional units of a similar device that was previously installed and put into operation are used as historical data for the second device (at least during the initial operating steps).

[0052] For example, the functional unit may include an electric motor and one or more sensors for detecting one or more electrical parameters (e.g. voltage, current, active or reactive power) and / or one or more parameters related to mechanical quantities (such as torque, vibration, etc.).

[0053] More generally, based on the functional unit, one or more of the following sensors may be provided: current sensor; voltage sensor; temperature sensor; vibration sensor; speed sensor; acceleration sensor; air flow sensor; steam flow sensor; glue consumption sensor; pressure sensor; sensor or system for measuring power consumption; torque sensor.

[0054] The predictive diagnostic method may include a learning step during which a collection of historical data relating to an operating parameter characterizing a given functional unit is generated. In the case of an electric motor, this operating parameter may be, for example, the current drawn. As mentioned above, the learning step may alternatively or in combination be represented by a step of acquiring historical data from another identical or similar device or part thereof that has already been put into operation.

[0055] In general, some operating parameters can be obtained by suitable sensors, probes, or transducers. For example, voltage and power parameters can be obtained by voltage and current sensors. Torque parameters can be detected by torque sensors or by processing the electrical signals of the actuator motor. Temperature can be detected by means of temperature sensors. Force and pressure can be detected by means of sensors or load cells, pressure switches, etc. In some cases, some operating parameters can be given by actuating the corresponding actuators. In some cases, the operating parameters may already be available as process parameters (such as pressure, speed, etc.).

[0056] Figure 2 (A), 2(B), 2(C) and 2(D) schematically illustrate embodiments of the method disclosed in WO2019048437, which may be a starting point for the improved method described herein.

[0057] More specifically, Figure 2 (A) shows a general graph of a general operating parameter (e.g., the current drawn by the motor) as a function of time. Figure 2 Δt sample The operating parameter is sampled at the indicated sampling interval. Time is indicated on the horizontal axis, while the operating parameter is indicated on the vertical axis. For example, the sampling interval may be one second.

[0058] In some embodiments, the operating parameters may be pre-processed. For example, they may be filtered, interpolated, or processed in other ways.

[0059] A movable learning time window indicated by Δt2 is identified along the time axis. The movable learning time window may last for example several days or weeks. By way of example only, the movable learning time window (hereinafter also referred to as "learning time window") may last for 60 days. The data acquired by means of the control system during the movable learning time window Δt2 are processed and the results of the processing are stored. According to some embodiments, a single calculation window for calculating a statistical function related to the operating parameter in question is identified within the movable learning time window Δt2. Figure 2 In the figure, the window for calculating the statistical function is represented by Δt1. The calculation window is basically a time window whose duration is less than the duration of the learning time window Δt2. In some embodiments, the duration of the calculation window Δt1 for calculating the statistical function is several minutes.

[0060] The processing of the historical data acquired within the movable learning time window Δt2 may provide for the calculation of a first statistical function and, if necessary, a second statistical function within each statistical function calculation window Δt1. In some embodiments, the statistical function may be a power spectral density, or a root mean square, or a simple maximum and minimum value calculated on a data set related to the controlled parameter and acquired within the time window in question. In a particularly advantageous embodiment, the statistical function may be the variance of the values ​​of the operating parameter in question (in Figure 2 ) or the mean (indicated by σ in Figure 2 Indicated by μ in . In an advantageous embodiment, both the variance and the mean of the data acquired in each single statistical function calculation window Δt1 are calculated. It is also possible to calculate more than two statistical functions.

[0061] In practice, along the movable learning time window Δt2, the movable calculation window Δt1 can be moved at regular intervals (e.g., 1 second), during which the statistical function(s) associated with the data contained in this window are calculated. In this way, the statistical function is calculated for all historical data acquired within the time defined by the movable learning time window Δt2.

[0062] For each position of the calculation window Δt1, the variance and mean of the values ​​of the operating parameters contained in the calculation window can be calculated. For example, for the general i-th position (Δt1) of the calculation window i , the variance σ can be calculated in this way i and mean μ i (where i=1...N). For a given movable learning time window Δt2, all values ​​(μ1, μ i ,...μ N ) and all values ​​calculated for the variance (σ1, σ i , ...σ N) select the maximum and minimum values ​​of variance and mean, as follows Figure 2 As indicated in:

[0063] The maximum value of the mean: MAX(μ),

[0064] Minimum value of the mean: min(μ)

[0065] Maximum value of variance: MAX(σ)

[0066] Minimum value of variance: min(σ).

[0067] The maximum and minimum values ​​of the two statistical functions are stored by the control unit of the device 1. Since the movable learning time window Δt2 is a window movable over time, as will be described below, the four maximum and minimum values ​​of the variance and mean change over time when the device operates.

[0068] In some embodiments, the values ​​MAX(μ), min(μ), MAX(σ), and min(σ) may be detected by using a greater number of samples of the operating parameters. To this end, the following operations may be performed.

[0069] Choose a suitable sampling interval Δt sample For example, the sampling interval may be several seconds. As a non-limiting example only, the sampling interval Δt may be set to sample =1 second. For the operating parameter values ​​obtained during the calculation window Δt1 that has just passed, the values ​​of two statistical functions can be calculated every second: the variance and the mean. In the subsequent second, the calculation window Δt1 moves for 1 second, and the variance and the mean are calculated again for the values ​​of the operating parameters in the calculation window obtained by shifting the calculation window Δt1 by 1 second. For example, this process can last for a whole day. For each day, the maximum and minimum values ​​of the variance and mean calculated as described above can be detected and stored. The maximum and minimum values ​​can also be calculated within different time ranges (for example, every hour or every ten hours) instead of within 24 hours. The 24-hour range is selected only for practical reasons. In fact, with a sampling interval of 1 second, 24*60*60=86,400 variance values ​​and the same number of means are collected in one day. Each value is calculated over the calculation window Δt1. Based on the 86,400 values ​​collected each day for each of the two statistical functions, the maximum and minimum values ​​σ are determined. MAX , σ min 、μ MAX 、μ minAt the end of the learning step, i.e., once the time window Δt2 (which typically lasts G days) has expired, the system will have G maximum values ​​of the variance, G maximum values ​​of the mean, G minimum values ​​of the variance, and G minimum values ​​of the mean. If Δt2 = 60 days, then 60 maximum and minimum values ​​will be available for each statistical function. Based on each of the G = 60 elements of these four sets, the values ​​MAX(μ), min(μ), MAX(σ), and min(σ) are detected.

[0070] In other embodiments, the maximum and minimum values ​​among all collected samples may be directly identified.

[0071] Once the initial learning step is completed, or once values ​​associated with similar devices already in operation have been acquired, values ​​of the operating parameters are acquired continuously for a time interval Δt3, the duration of which is preferably less than the width (i.e., duration) of the movable learning time window Δt2. For example, the time interval Δt3 may have a duration of several days, for example from 0 to 20 days, typically 15 days. It should be understood that these numerical data and the previous data are given only as non-limiting examples.

[0072] Once the time interval Δt3 following the movable learning time window Δt2 has ended (or its data are added to the data acquired by the data collection done for similar devices), a predictive diagnostics process is started for the functional unit, which process refers to the detected and processed operating parameters. This step provides for the calculation of the current time window Δt act The first and second statistical functions (in the example shown, the variance and the mean) of the values ​​of the operating parameters detected during the current time window are calculated. In some embodiments, the duration of the current time window may be the same as the duration of the window Δt1 used to calculate the statistical functions. As will be better explained below, this is particularly preferred because in the current time window Δt act The data acquired during this step in will be used as historical data for dynamic learning. Moreover, the current time window Δt act is movable, ie it shifts in time similarly to the movable learning time window Δt2. The current time window is preferably kept at a fixed time distance (interval Δt3) from the movable learning time window.

[0073] exist Figure 2 In the figure, in the current time window Δt act The variance and mean values ​​of the statistical function used to calculate the operating parameters are represented by σ act and μ act These values ​​will also be indicated as the current values ​​of two statistical functions (mean and variance).

[0074] In the current time window Δtact The value σ calculated in act and μ act is compared with the values ​​MAX(μ), min(μ); MAX(σ) and min(σ) defined above and calculated in the movable learning time window Δt2. If the functional unit to which the operating parameter in question relates operates correctly, then the statistical value σ act and μ act It should be included between the maximum and minimum values ​​calculated in the movable learning time window. If necessary, a corresponding tolerance gap above and below the corresponding maximum and minimum values ​​can be provided for each of the two statistical functions. Starting from the values ​​MAX(μ), min(μ); MAX(σ) and min(σ), the extended interval including the tolerance margin is defined as follows:

[0075] The interval of the first statistical function (variance): [min(σ)-Δ; MAX(σ)+Δ]

[0076] The interval of the second statistical function (mean): [min(μ)-Δ; MAX(μ)+Δ]

[0077] In a particularly advantageous embodiment, within each of the intervals defined above, a corresponding intermediate interval may be defined:

[0078] [min(σ)-Δ'; MAX(σ)+Δ']

[0079] [min(μ)-Δ';MAX(μ)+Δ']

[0080] Where Δ'<Δ.

[0081] exist Figure 2 In (B), a Cartesian plot is shown; the mean is indicated on the horizontal axis and the variance is indicated on the vertical axis.

[0082] In this figure are shown: a first inner square, defined by the maximum and minimum values ​​of the variance and the mean; a middle square, containing the inner square; and an outer square, containing the inner square and the middle square, defined by the intervals indicated above.

[0083] In the example shown, the same values ​​Δ and Δ′ are used for the gaps of the variance and mean, respectively. However, this is not strictly necessary. It will be appreciated that, for example, different margins may be provided to widen the intervals of the mean and variance.

[0084] Furthermore, as noted, while two statistical functions (variance and mean) are used in the described examples, it should be understood that in other embodiments, different statistical functions and / or a different number of statistical functions may be used.

[0085] In each current time window Δtact The current values ​​of the variance and mean σ calculated in act and μ act Basically limited Figure 2 (B) is the coordinate of a point in the diagram. This point is also defined as the current operating point of the functional unit. If this point is within the square defined by MAX(μ), min(μ), MAX(σ) and min(σ), then the current value σ of the operating parameter for which the variance and mean are calculated is act and μ act The functional units operate correctly.

[0086] This square defines the range of permissible values ​​for the first and second statistical functions. This range is two-dimensional because two statistical functions have been used.

[0087] As already mentioned, the method can also be implemented using a different number of statistical functions, for example three or more statistical functions, or even only one statistical function, but it is presently preferred to use two statistical functions.

[0088] In the case of three statistical functions, the range of permissible values ​​is defined by a three-dimensional volume. In the case of N statistical functions, the volume of permissible values ​​is defined by an N-dimensional space. In the case of only one statistical function, the range of permissible values ​​is reduced to a line, that is, to a one-dimensional space.

[0089] Reference again Figure 2 For a two-dimensional example, if the coordinate σ act and μ act The current operating point is in the square defined by:

[0090] [min(σ)-Δ'; MAX(σ)+Δ']

[0091] [min(μ)-Δ';MAX(μ)+Δ']

[0092] Then no alarm or warning signal will be provided, because Δ' can be considered as a tolerance value around the on-time data. If the point is between the middle square and the outer square defined by:

[0093] [min(σ)-Δ; MAX(σ)+Δ]

[0094] [min(μ)-Δ;MAX(μ)+Δ]

[0095] Then a warning signal is generated, and if the point is outside the largest square, an alarm signal is generated. In other embodiments, if the currently operated point is between the inner square and the middle square, a warning is generated, and if the point is between the middle square and the outer square, or outside the latter, an alarm signal is generated.

[0096] These anomalies of the difference of the statistical function from the square calculated during the learning step (or obtained from historical data stored during operation of a similar device or a similar functional unit of another device) indicate an incipient fault, and the corresponding alarm therefore represents useful information for predictive diagnostics.

[0097] Anomalous data are useful for highlighting the onset of fault conditions, but they should not be used in the data acquisition step (i.e., the system learning step) because this would introduce errors. Therefore, provision can be made for automatically removing (e.g., by an algorithm) or manually removing (e.g., by an operator) anomalous data from the data series useful for the learning step.

[0098] In some embodiments, only one alarm threshold may be used, rather than the two alarm thresholds (or pre-alert and alarm) described above.

[0099] In some embodiments, a time threshold may be provided to avoid false alarms, such as temporary fluctuations in operating parameters due to factors unrelated to an incipient fault condition. Figure 2 In the diagram (B), the value σ act and μ act An alarm or warning is generated only if the defined point remains outside the square defined between the values ​​MAX(μ), min(μ), MAX(σ) and min(σ) and the tolerance gap (if any) for longer than a preset time threshold. Conversely, that is, if the anomaly ends after a time shorter than the preset time threshold, no alarm is generated.

[0100] Abnormal operation that would cause such a change in the statistical functions used (e.g. the values ​​of the variance and the mean) to trigger a warning or alarm signal could be due to different external causes, such as incorrect adjustment or incorrect use of a functional unit, incipient failure due to degradation caused by wear, or any other reason.

[0101] Figure 2 The square shown in (B) can be used on a monitor to give a direct visual indication that the operator can quickly understand. To give a more intuitive representation, the coordinates can be changed to be as shown in FIG. Figure 2 The circular diagram indicated in (C) shows Figure 2 (B) The same situation.

[0102] The method for controlling the functional unit to which the detected operating parameter relates may provide for sampling at a sampling interval Δt of, for example, 1 second. sample In the current time window Δt act Calculate the current value σ act and μ act . Move the current time window every second and calculate the plane σ, μ again ( Figure 2 (B) or Figure 2 (C)) The coordinates σ of the actual operating point act and μ act .

[0103] As indicated above, in the current time window Δt act The current value σ of the statistical function σ and μ calculated in act and μ act Compare with MAX(μ), min(μ), MAX(σ), and min(σ) identified in the movable learning time window Δt2, where Δt2 is the same as the current time window Δt act are separated in time by an interval Δt3. In this way, a discontinuity can be created between the learning period and the current period. This can be useful to take into account the fact that some operating parameters of a given functional unit may drift slowly over time, for example due to aging of one or more components. If the value σ act and μ act and the current time window Δt in time act If the maximum and minimum values ​​of the statistical function calculated over the adjacent learning time window Δt2 are compared, then this drift cannot be detected. Vice versa, by introducing the time interval Δt3, the gradual drift of the detected operating parameter leads to a signal or alarm, because the current value σ is act and μ act One or the other or both will be outside the square identified by the maximum and minimum values ​​of the statistical function computed over the movable learning time window.

[0104] As mentioned above, the learning steps are continuous and dynamic; this means that once the first learning step in the learning time window Δt2 is finished, the data related to the controlled operating parameters will continue to be stored and the learning time window Δt2 will be stored along the time axis ( Figure 2 The horizontal axis in (A) moves so that the act Always at the same time distance Δt3.

[0105] Figure 2 (A) with Figure 2The comparison between (D) makes this aspect clear. At each time step, corresponding for example to the time width of the computation window Δt1, the movable learning time window Δt2 is shifted by the same amount as the computation window Δt sample The width of the step, following the current time window Δt act The values ​​of the statistical function calculated on the oldest calculation window Δt1 are excluded and rejected, while the values ​​of the statistical function calculated on the data contained in the subsequent calculation window Δt1 enter the movable learning time window Δt2. Essentially, as by comparing Figure 2 (A) with Figure 2 (D) It is clear that the movable learning time window Δt2 is movable over time and is in close proximity to the current time window Δt act Move forward while maintaining the time distance Δt3. At each forward step, the most recent statistics are historized and the latest statistics are calculated.

[0106] Each time the movable learning time window Δt2 moves forward by a step size Δt1, the values ​​MAX(μ), min(μ), MAX(σ), and min(σ) are detected thereon (μ1, μ i ,...μ N ) and (σ1, σ i , ...σ N ) changes, so the maximum and minimum values ​​of the calculated statistical function may change. Figure 2 (B) and Figure 2 The square represented in (C) can be moved gradually over time. Therefore, learning is dynamic and continuous.

[0107] The learning time window Δt2 can be moved relative to the current time window Δt act The time interval Δt3 is always maintained. Therefore, as time passes, even if the maximum and minimum values ​​of the statistical function are updated, and thus ( Figure 2 (B) and Figure 2 (C)) is determined by the value σ act and μ act The square in which the defined point lies may move, and any slow drift of the operating parameters can always be detected. The duration of the time interval Δt3 can be constant. This simplifies the processing. However, this is not strictly necessary.

[0108] Even though reference has been made above to the case of defining a two-dimensional graph using two statistical functions (mean and variance) where the values ​​σ act 、μ actIt is also possible to use only one statistical function, such as only the variance or only the mean. In this case, all the considerations made above apply, the only difference being that there is only one statistical function and the plot will be one-dimensional instead of two-dimensional.

[0109] In other embodiments, more than two statistical functions may be used with the same criteria as described above. In this case, from a graphical point of view, the current operating point of the functional unit (or more precisely, the value of the operating parameter associated with this functional unit) should remain within the cubic (or spherical) volume defined by the maximum and minimum values ​​determined over the movable learning time window of the three statistical functions.

[0110] Even though specific reference is made in this specification to the variance and general mean of the values ​​of the operating parameters discussed, it should be understood that other statistical functions may be used. Furthermore, the mean may be an arithmetic mean, a weighted mean, a geometric mean, a harmonic mean, a power mean, an arithmetic geometric mean, an integral mean, a time mean, or any other function of the mean of defined values.

[0111] Described above are embodiments of the predictive diagnostic method disclosed in more detail in WO2019048437, the contents of which are fully incorporated into this document.

[0112] Even though reference has been made above to a single learning interval, multiple learning intervals may be used, resulting in inconsistent ranges of permissible values, for example, resulting in partially overlapping ranges. In this case, the range of permissible values ​​may be defined as the envelope of the ranges of permissible values ​​obtained in different learning phases or cycles. Figure 3 (A) shows four rectangles Q1, Q2, Q3 and Q4, defined by the maximum and minimum values ​​of the mean and variance calculated over four different learning intervals. Each of them represents a corresponding range of allowable values ​​associated with a learning stage. Figure 3 (B) shows an envelope rectangle containing all rectangles Q1, Q2, Q3, Q4, which represents the range of allowable values ​​used in the above-mentioned control method.

[0113] As mentioned above, the present invention is based on the surprising discovery that the coordinates (σ act μ act) and its deviation outside the range of permissible values ​​defined by MAX(μ), min(μ); MAX(σ), and min(σ) not only indicate an incipient fault but also allow for predictive diagnosis of the fault. The position of the current operating point relative to the range of permissible values ​​(outside the latter) is not random but rather correlates with the type of fault or malfunction that is occurring or about to occur. In other words, the fact that the current operating point is to the right of the range of permissible values ​​rather than to the left is not accidental; in fact, a displacement in one direction or another outside the range of permissible values ​​indicates one type of fault rather than another.

[0114] Thus, the method disclosed herein provides the steps of detecting where the current operating point is outside a range of allowable values ​​and using this location to obtain statistical information about the type or nature of the nuisance or incipient fault and to suggest possible solutions or possible recovery interventions.

[0115] This concept will become clearer after the following example.

[0116] Figure 4 A portion of the line of Figure 1 is shown in more detail. More specifically, Figure 4 The area where single-faced corrugated cardboard sheets NS are extracted by the single-facer is shown on bridge 41. Reference numeral 101 designates a belt system driven by motor 103. Belt 101 should be periodically adjusted and / or replaced due to wear. Possible malfunctions of the functional unit represented by motor 103 could be excessive or insufficient belt tension or overheating of the motor. A controlled operating parameter could be the current drawn by the motor.

[0117] Figure 5 A time diagram showing the trend of the current absorbed by the motor 103. Time is shown on the y-axis and the absorbed current is shown on the x-axis. Figure 5 In the figure, the belt is over-tensioned, so the current drawn by the motor 103 is higher than the mean, but substantially constant. Therefore, the variance of the operating parameter is between the maximum and minimum values ​​calculated during the learning step or based on historical data, while the mean is greater than the maximum value defining the range of permissible values. Therefore, the current value of the variance and the current value of the mean (σ act and μ act ) defines that the point of the current operation is located to the right of the range of allowable values.

[0118] This situation is Figure 6 is shown in FIG, which shows a method based on a method similar to that used for Figure 2 The range of permissible values ​​for the variance and mean of the criterion. The current operating point is outside the permissible range and to the right (the mean is greater than the maximum permissible value).

[0119] Figure 7 Shown with Figure 5 , where the current drawn (on the x-axis) as a function of time (on the y-axis) is lower than normal and fluctuates over time. This situation may occur, for example, when belt 101 is not sufficiently tensioned. The current drawn is lower than normal and suddenly decreases when the belt slips on the drive pulley driven by motor 103.

[0120] Due to this anomaly, the mean of the detected operating parameter (current drawn) of the monitored functional unit (motor 103 of belt 101) is below the minimum permissible value. Due to fluctuations caused by the loose belt slipping on the drive pulley 104, the variance is greater than the maximum permissible value.

[0121] Therefore, the point of current operation is outside the range of permissible values ​​and above its left side, as shown in Figure 8 As shown in .

[0122] By comparison Figure 6 and Figure 8 , from which it can be noted that different malfunctions (ie anomalies) affecting the same operating parameter (current absorbed) of the same functional unit (motor 103) lead to different situations in the graph of the statistical function.

[0123] By collecting a sufficient number of abnormal events occurring in one or more devices, it is possible not only to detect abnormalities that may lead to failures and require intervention, but also to correlate these abnormalities with the position of the current operating point in the graph of the statistical function. Conversely, it is also possible to provide information about the type of event (malfunction, failure, or incipient failure) that resulted in the generation of an alarm signal due to the fact that the current operating point was outside the range of permissible values ​​of the statistical function, and thus to recommend solutions or intervention measures to eliminate the abnormality and restore the device.

[0124] In some cases, the gradual and constant increase in current drawn by motor 103 may be due to reasons other than excessive tensioning of the belt. For example, malfunction of the cooling fan of motor 103 may cause the increase in current drawn. Figure 9 Shown in. Figure 10 A graph showing a statistical function in which the position of the current operating point is outside the range of allowable values ​​(to the right thereof) as the mean value of the absorbed current increases.

[0125] therefore, Figure 5 、 Figure 6 The situation and Figure 9 and Figure 10 The situations are different in terms of the failure or malfunction that causes the alarm signal, but they are the same or nearly the same with respect to the abnormal position of the current operating point relative to the range of allowable values.

[0126] Figure 9 and Figure 10 Statistically speaking, Figure 5 and Figure 6 The situation is even more unlikely.

[0127] As can be seen from the learning phase, there are two possible reasons for abnormal operation, and the device can be restored by removing these reasons. The first reason is statistically more frequent than the second reason. Figure 5 、 Figure 6 or Figure 9 、 Figure 10 When one of the alarm conditions shown in the example occurs, possible solutions to the anomaly in question can be listed (based on the data collected during the learning phase). These solutions can be listed in proportion to their probability of being the correct solution. In this particular case, the solution of a loose belt is statistically classified as more successful; therefore, it is listed before the other solutions (replacing the cooling fan or correctly assembling the cooling fan).

[0128] Based on the data collected during the learning phase (and generally during the use of one or more identical or similar devices), a database can be created in which each abnormal situation (point of current operation outside the range of permissible values) is associated with a corresponding cause or multiple possible causes, and with a corresponding solution. Figure 5 、 Figure 6 In the example, the cause is excessive belt tension; Figure 7 and Figure 8 In the example, the cause is insufficient belt tension; Figure 9 and Figure 10 In the example above, the (less common) cause is a malfunction of the cooling fan. As in the first and third cases, the position of the point of current operation relative to the range caused by the anomaly is the same, and in both cases, the same cause (and solution) will be indicated.

[0129] Once a database has been created, associating the values ​​of current statistical functions with faults, this database can be used to obtain one or more pieces of information (at least statistically, even if not punctually) about what happened in the functional unit when the current operating parameters were outside the range of permissible values, and therefore about the type of intervention required to solve the problem.

[0130] Figure 11A block diagram is shown, in which reference 1 designates a corrugated cardboard production facility, reference 105 designates a common functional unit, such as the aforementioned motor 103, and reference 107 designates a central control unit. The central control unit receives data related to the operation of the functional unit 105 from its sensor(s) and can identify any abnormalities, such as values ​​of one or two statistical functions outside the permissible range of values. The central control unit 107 communicates with a server 109. The server 109 has access to a database 111 containing information on the causes of malfunctions associated with the various positions of the current operating point relative to the permissible range of values ​​for each monitored functional unit of the facility 1.

[0131] When queried by central control unit 107, server 109 provides a list of possible causes and solutions for the detected anomaly, or more precisely, a list of solutions for device recovery. In practice, anomalies can typically be detected by a tendency of the signal to move toward a suboptimal operating region. An anomaly can have several causes. One or more solutions can address the cause of the anomaly. The method may also provide a step for suggesting multiple solutions, listed in descending order based on likelihood of success.

[0132] In the above example of the belt 101 and the motor 103, Figure 5 、 Figure 6 and Figure 9 、 Figure 10 In the case of a fault, server 109 provides two possible causes of malfunction and two corresponding solutions: the belt is over-tensioned (solution: loosen the belt); the motor cooling fan is not operating properly (solution: check fan operation). The first cause is statistically more frequent. The server can provide central control unit 107 with the causes ranked by their statistical frequency to allow the operator to intervene in a reasonable manner, starting with the most statistically likely cause for inspection and intervention.

[0133] The described system can also allow the occurrence of a different cause than those listed based on the contents of database 111 to be communicated to server 109. For example, an operator can verify that none of the listed causes of a fault occurred, and can communicate to the server that, in this particular case, the abnormal coordinates of the currently operating point correspond to a different cause than those already in the database. This new cause (and corresponding solution) can be stored in the database. In this way, content can be added to database 111 for future use.

[0134] Furthermore, an operator who checks which of the causes hypothesized based on the data contained in the database 111 is the actual detected cause can transmit this information to the server 109 in order to modify the "ranking" of the fault causes. This affects the actual content of the database, i.e. the frequency of occurrence of a fault cause (and the associated solution) relative to another fault cause can be varied over time in order to take into account the actual number of occurrences of this cause.

[0135] The above operations may be fully or partially automated, or performed manually, with intervention from the operator's device side and / or the operator's server side.

[0136] The connection between the device 1 and the server 109 can be a remote connection via the Internet. In some embodiments, the system can be implemented using cloud technology. The server 109 and the database 111 can also be installed directly at the same location as the device. However, it is preferred to centralize the server and database (for example, at the device supplier or seller, or at any other entity responsible for supervision and service) so that the information received from the individual devices can also be centralized and added to the existing content of the database.

[0137] Figure 12 An overview diagram of an embodiment of the method described herein is shown. The left portion of the diagram (indicated by (A)) shows functions performed on the device, for example, by a control unit. The right portion, indicated by (B), shows functions performed by a server. The unit controlling the device or a portion thereof may be connected to the server via the Internet or by any other suitable channel.

[0138] Block 201 shows the detection of an anomaly, represented by the fact that the point P of the current operation representing the coordinates (σ, μ) of the operating parameters of the functional unit is outside the range of permissible values. The anomaly is reported to the server (block 202).

[0139] Based on the values ​​of the coordinates (σ, μ), i.e., the position of the current operating point relative to the range of permissible values, the server queries a database (block 203) and receives (block 204) a list of possible anomalies and associated solutions, which may correspond to the detected coordinates. Each anomaly (Anomaly_i) and its solution (Solution_i) may be characterized by a probability value (Prob_i(%)), i.e., the likelihood of occurrence. The list of possible anomalies and solutions (in the example shown, comprising N anomalies and N solutions) is sent to the plant control unit, which receives it (block 205) and may, for example, display it on a user interface to allow an operator to perform the necessary checks.

[0140] The plant operator may perform a check (block 206 ) starting with the anomalies that are more frequently associated with the detected abnormal operating conditions (P(σ, μ)).

[0141] Whether an anomaly is found among the listed anomalies (box 207) may be communicated to the server 109 via the central control unit 107. If so, the server 109 is informed (box 208) which of the listed anomalies (and solutions) is the actual anomaly found. Figure 12 The block diagram shows that the actual detected anomaly is the jth anomaly (Anomaly_j), and it has been resolved according to the jth solution (Solution_j). In this way, the server 109 can update (box 209) the value of the anomaly probability based on the historical data plus the data related to the anomaly just resolved.

[0142] If none of the listed anomalies occur in the functional unit, and the operator identifies a different, unlisted cause of malfunction, he / she (if necessary with the help of a maintenance technician provided by the equipment manufacturer) prepares and applies a corresponding solution. In block 210, the new anomaly indicated by Anomaly_N+1 is transmitted to the server 109, which inserts it among the possible anomalies associated with the condition P(σ, μ) and updates the database, see block 211. In doing so, not only is the possible (N+1)th anomaly added to those that can correspond to the operating coordinates P(σ, μ), but the probability of possible anomalies associated with a given position of the operating point outside the range of permissible values ​​is also recalculated.

[0143] It should be understood that, in general, it is not possible to identify an anomaly (or series of anomalies) for every specific point in the current operation that is outside the range of permissible values ​​for σ and μ. Instead, there will be smaller or larger areas containing points corresponding to the same type of cause, and the server 109 can identify and propose corresponding solutions to the equipment operator.

[0144] refer to Figure 12 The data exchange described can occur with a higher or lower degree of automation. In principle, all communications can occur via one or more human operators. In practice, communication can occur at least partially automatically or semi-automatically, with an operator opening a transmission channel between the control unit (or the control unit and its associated computer or server) and the server 109; data and information can be exchanged via this channel.

[0145] In an initial learning phase, solutions to various faults or abnormal operating conditions can be obtained from the maintenance cards executed over time, with the database 111 gradually and continuously increasing. In this way, continuous learning allows providing increasingly accurate solutions to possible anomalies in the various functional units of the device 1.

[0146] exist Figure 11 and Figure 12 In the figures of , a solution has been illustrated in which the devices are physically and administratively separated and independent of a server managing a database providing information on the causes of detected anomalies and solutions. However, this architecture is not the only one possible. In fact, the database may reside, for example, in a computer or server belonging to the device or, in general, owned by the owner of the device, instead of in a server of a third party. In this case, the above references can be executed directly from the control unit of the device 1 or from a server connected to the device 1 and managed by the same body that manages the device. Figure 12 In this case, the database can be updated with data from a single device, but also with data from other devices if necessary. Several devices can exchange data with each other for storage in a database located at a single device, for example on a server of the company that owns the device.

[0147] For example, in this type of configuration, each device can contribute to updating and increasing a database with anomalies / solutions in order to create a global machine network.

Claims

1. A method for monitoring the operation of a corrugated board production plant, comprising the following steps: detecting at least one operating parameter of a functional unit of the device and calculating a current value of at least a first statistical function of said operating parameter in a current time window, the current value of said first statistical function defining first coordinates of a point of current operation of the functional unit; verifying whether the current operating point is within a range of allowable values ​​of the first statistical function, wherein the values ​​contained in the range of allowable values ​​correspond to correct operation of the functional unit; If the currently operated point is outside the range of permissible values, determining the position of the currently operated point relative to the range of permissible values; wherein, for at least some of the coordinates of the currently operated point that are outside the range of permissible values, the database contains a plurality of possible causes of the deviation of the currently operated point from the range of permissible values, and, for each of the possible causes, the database contains an indication of whether one of the possible causes has a higher or lower probability of occurrence relative to the other possible causes; Based on the location of the current operating point relative to the range of allowable values, querying a database via a server and receiving a list of possible anomalies and associated solutions corresponding to the location of the current operating point; wherein each anomaly and its solution are characterized by a probability value indicating the likelihood of occurrence of the anomaly; communicating, by the control unit to the server, which anomaly in the list of possible anomalies caused the deviation, and updating, by the server, a value of the anomaly likelihood based on historical data plus data related to the anomaly that was just resolved; If a different anomaly not included in the list of possible anomalies causes the deviation, the different anomaly is communicated to the server, added to the possible anomalies associated with the corresponding coordinates of the currently operated point in the database, and the likelihood of the possible anomaly associated with the coordinates is calculated again.

2. The method as claimed in claim 1 further includes the step of calculating a current value of a second statistical function of the operating parameter in the current time window, the current value of the first statistical function and the current value of the second statistical function respectively defining a first coordinate and a second coordinate of a point of current operation of the functional unit; and wherein the range of allowable values ​​is limited to a two-dimensional range of allowable values ​​of the first statistical function and the second statistical function.

Images