Method for life prediction and monitoring

Abstract

A method for life prediction and monitoring of a device and a corresponding system are provided. The method includes calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model having end of life (EOL) boundary conditions, the boundary conditions including a first boundary condition and a second boundary condition, the first boundary condition is a flux-free boundary condition and the second boundary condition is an absorption or partial absorption boundary condition; a condition-dependent observable measurement of the measuring device; obtaining first data representing a measured value of the observable measurement; calculating the possibility of aging variables according to the first data; updating the calculated probability density function of the aging variable based on the likelihood; and generating a signal indicative of a health prediction of the device based on the probability density function, the aging model, and the EOL boundary condition.

Description

Technical Field

[0001] The present disclosure relates to methods and systems for device monitoring, in particular for monitoring devices with respect to generation, transmission, distribution and / or use of electrical energy. Background Art

[0002] Device failure can have serious consequences. In many cases, the damage caused by unexpected device failure can far exceed the cost of the device. In some cases, device failure is prevented by monitoring the device. Monitoring includes detecting and indicating deviations in device properties that may indicate a deterioration in the state of the device. Where the device is monitored while it is in operation, this is also referred to as online monitoring. Existing monitoring uses physical sensors to observe device properties. Alternatively, in other cases, statistical predictions are used to prevent device failure. For example, a device may undergo maintenance periodically, such as after a minimum predetermined amount of time has passed or after the device has performed a predetermined number of cycles. Summary of the invention

[0003] An object of the present disclosure is to provide a more accurate method for preventing failures.

[0004] The object is solved by a method for monitoring a device. The method comprises: providing or obtaining an aging model, wherein the aging model comprises an end-of-life (EOL) boundary condition and at least one equation having an aging variable of the device; calculating a probability density function of (multiple) aging variables over time based on solving the at least one equation according to the aging model with the EOL boundary condition; measuring an observable quantity related to the operation of the device; obtaining first data representing the measured value of the observable quantity; calculating the likelihood of the aging variable based on the first data; updating the calculated probability density function of the aging variable based on the likelihood; and generating a signal indicating a health prediction and a remaining useful life (RUL) of the device based on the probability density function, the aging model, the EOL boundary condition and the reliability calculated therefrom.

[0005] The aforementioned object is also solved by another method for monitoring a device. The method includes: calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model with an end-of-life EOL boundary condition; measuring an observable quantity related to an operating condition of the device; obtaining first data representing the measured value of the observable quantity; calculating the likelihood of the aging variable based on the first data; updating the calculated probability density function of the aging variable based on the likelihood; and generating a signal indicating a health prediction of the device based on the probability density function, the aging model and the EOL boundary condition. The aging model may include the EOL boundary condition and the at least one equation with the aging variable of the device. The method may further include providing or obtaining the aging model.

[0006] The aforementioned object is also solved by another method for monitoring a device. The method includes: calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model with an end-of-life EOL boundary condition, wherein the boundary condition includes a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorption or partial absorption boundary condition; measuring an observable quantity related to an operating condition of the device; obtaining first data representing a measured value of the observable quantity; calculating a likelihood of the aging variable based on the first data; updating the calculated probability density function of the aging variable based on the likelihood; and generating a signal indicating a health prediction of the device based on the probability density function, the aging model and the EOL boundary condition. The aging model may include the EOL boundary condition and the at least one equation with the aging variable of the device. The method may further include: providing or obtaining the aging model.

[0007] By means of the aforementioned method, a more accurate and reliable signal indicating the health prediction (forecast) of the device is provided, so that device failure can be prevented more reliably and efficiently.

[0008] Aging variables reflect properties of the device that can be considered constant or nearly constant over the time scale of normal device operation. A vector including one or more aging variables can be represented as Note that vector arrows simply indicate the fact that the corresponding quantity can have more than one component, whereas different vectors (e.g. and the following ) may have different numbers of components. Hereinafter, when referring to the aging variable , including the case where there is only one aging variable μ.

[0009] The aging model defines the one or more aging variables The aging of a device is modeled by its dynamic behavior over time. Aging (also called degradation) refers to the change in the health of a device due to aging variables. Degradation that occurs, often irreversibly, as a result of significant temporal variation on a time scale that is usually long compared to the time scale of device operation. In the reliability context, this refers primarily to wear-out failures. Degradation can occur slowly and continuously. Degradation can occur rapidly but in small increments distributed over time, such as when the device is operating. Thus, aging variables are is often considered to be continuous or quasi-continuous on long-term aging time scales. As described above, the aging variables Can be considered a constant of the device operation. Aging occurs due to deterministic and stochastic stresses. For example, aging can occur during operation, which is called cyclic aging. Aging can also occur independently of operation due to external factors (such as environmental phenomena), which is called calendar aging.

[0010] EOL boundaries are aging variables A surface in variable space that defines the aging variables associated with acceptable or healthy settings (acceptable region) The values ​​of and aging variables associated with devices that have unacceptable conditions (which we call unhealthy) In other words, when the device has a high probability of being outside the acceptance region (say, P EOL ), the device is considered unhealthy. In one example, the EOL boundary can be defined by a destructive failure of the device. In another example, the EOL boundary can be defined by a fault. In yet another example, the EOL boundary can be defined by an additional safety margin to the failure boundary. In general, the EOL boundary conditions are defined with respect to one or more failure and / or unsatisfactory behavior scenarios that should be prevented using the method.

[0011] When the aging variable When the value of crosses the surface defined by the EOL boundary condition, the aging variable Leaving the acceptance region, which results in a degradation in reliability R, which is defined as the probability of being in this region. Aging variables reaching EOL boundary conditions can be absorbed, for example, when the corresponding device is removed (or fails). The EOL boundary condition can be an absorbing boundary condition associated with a probability density of disappearance. Similarly, the boundary condition can be a partially absorbing boundary condition, i.e., when the aging variable The aging variable reaches the EOL boundary condition when the value of May be absorbed, for example, when the corresponding device is removed (or fails).

[0012] In some examples, the receptive region or its portion of the boundary may also include a source of probability density. For example, when the aging variable When the EOL boundary is reached, maintenance or replacement of the device can be triggered. In this case, the aging variable The dynamics can be continued by restarting at another location in the acceptance area or on its surface.

[0013] Further boundary conditions may be defined that do not reflect EOL boundary conditions. For example, a no-flux boundary condition may be defined, i.e., a boundary condition in which there is no probabilistic decay. When the boundary conditions include a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition, aging may be accurately modeled. Thus, predictions are more accurate, and maintenance may be more efficient and more effective.

[0014] An aging model uses one or more equations to define one or more aging variables. Dynamic behavior over time. In one example, the dynamics of device aging is based on the aging variable In another example, the dynamic process of device aging is modeled according to its equivalent probability density function (PDF) for the associated The phenomenon is modeled by partial differential equations (PDEs), such as the Fokker-Planck equation (FPE).

[0015] PDF of aging variables over time Defined aging variables The probability distribution at time t.

[0016] The method comprises solving the at least one equation according to an aging model with EOL boundary conditions, for example numerically. As a result, the PDF of the aging variable over time is obtained: Therefore, although the aging variables are not directly calculated for any future time , but for any future time t the aging variable is obtained In general, the PDF is less localized at more distant time points, reflecting the fact that the aging variables are localized for more distant time points. The uncertainty of the value of increases gradually.

[0017] The method further comprises: measuring an observable quantity related to the operating condition of the device at time t1. The observable quantity is a variable ψ, or a plurality of variables These variables need to be available for measurement (e.g., using sensors). This includes the case where there is only one observable ψ.

[0018] Observables Need to be related to at least some aging variables A certain relationship is exhibited, such as a direct functional relationship or a random correlation. First data representing a measured value of an observable at time t1 is obtained.

[0019] Observables and aging variables The relationship between the aging variables At the measuring time t m Possibility

[0020] This likelihood is used to update the aging variable based on the likelihood The update of the calculated PDF is performed using Bayesian inference update. Thus, the update of the calculated PDF is performed using the Bayesian inference update. The information provided by the measured values ​​is incorporated into the aging model. For example, when the measured value indicates that the value of the aging variable is less than the predicted aging variable The calculated PDF is updated to reflect the unexpected dynamics, and the updated PDF For the measurement time t m As another example, when the measured value indicates that the value of the aging variable is less than the predicted aging variable, The calculated PDF is updated to reflect the unexpected dynamics, and the updated PDF For the measurement time t m Later time points are used to predict smaller values ​​of the aging variable.

[0021] Finally, the method includes: generating a signal indicative of a health prediction of the device based on the probability density function, the aging model, and the EOL boundary conditions. The signal indicative of the health prediction may include any of the following: a life curve, a reliability curve, a failure probability density curve, a remaining life value, a RUL value, a reliability value, an unreliability value, a failure probability value, and / or similar values ​​or curves related to predicted device health and / or predicted device reliability.

[0022] Various embodiments may preferably implement the following features:

[0023] Preferably, the method further comprises: providing or obtaining a device model, wherein the device model comprises equations having aging variables as device parameters; wherein the EOL boundary condition is based on the device model; and wherein the calculation of the likelihood is based on the device model.

[0024] The device model is based on the state variables The dynamics of the device are defined on a fast time scale of operation, thereby modeling the operational behavior of the device. In contrast, aging models are based on aging variables. Define the dynamics of the device on the slow time scale of aging, as described above. Treat the aging variables as continuous or quasi-continuous for the aging model Device parameters that are considered constant in the device model

[0025] The device model can be used to target device parameters Thus, the device parameters that exhibit unacceptable behavior in the simulation are Used to define EOL boundary conditions. Unacceptable behavior may, for example, include unstable behavior of state variables, state variables exceeding safety thresholds, and / or state variables associated with the end of life of the device. The EOL boundary conditions may be obtained from the device digitally or analytically.

[0026] The calculation of the probability can be based on the device model. The device model defines the state variables and device parameters This relationship can be expressed as the device parameters The state variable for the condition Then, in the first data In the case of is calculated as As defined by the device model.

[0027] By providing or obtaining a device model, the EOL boundary conditions better reflect actual failures, and the probabilities better reflect the actual information contained in the measurements. As a result, predictions become more accurate. In this way, unexpected device failures are prevented, and wasteful preventive maintenance is reduced.

[0028] Preferably, the method further comprises: providing or obtaining a monitoring model, wherein the monitoring model comprises equations with observables and aging variables; wherein the monitoring model is based on the device model; and wherein calculating the likelihood is based on the monitoring model.

[0029] Since the observable needs to be available for measurement (e.g. using a sensor), so the state variable or aging variables Some or all of may not be suitable as observables. As described above, observables At least some aging variables A certain relationship is shown. This relationship can be defined by the monitoring model and can be based on the device model.

[0030] The monitoring model consists of observable and aging variables This equation (which may include multiple sub-equations) defines the observable and aging variables For example, when the relationship is a random correlation, the equation can include a definition of conditional probability.

[0031] For example, in the observable The distribution corresponds to a very precise value μ1 = a and contains no information about other aging variables. It can be a normalized Gaussian function of μ1 that peaks sharply at a and is independent of other conditioning variables. In other words, in this case, the monitoring model consists of a Gaussian function that is strongly localized.

[0032] By providing or obtaining a monitoring model, the actual information contained in the measured values ​​can be extracted more thoroughly and accurately. As a result, predictions become more accurate. In this way, unexpected equipment failures are prevented and wasteful preventive maintenance is reduced.

[0033] Preferably, the generation of the signal includes computing a representation of the unreliability over time based on a probability density function, an aging model, and EOL boundary conditions. Unreliability is also known as failure probability and has time as its base space. It is based on a probability density function The probability density function is generated by as its basic space.

[0034] Reliability of the device At each time t the output device is in the acceptance region Ω a The probability that the device is in the range of t and therefore has not failed before this time. Therefore, the unreliability 1-R(t) corresponds to the probability that the device does not fail until this time. The reliability and this failure probability for a specific point in time are calculated by marginalizing the PDF (defined by solving the aging model) over this part of the variable space, and the acceptance region is partially defined by the EOL boundary.

[0035] In this case of absorbing EOL boundary conditions, the aging variable is absorbed at this boundary, i.e. In this sense, the probability of device failure is concentrated at the EOL boundary.

[0036] By calculating a representation of the unreliability over time, the predictions obtained from the aging model become understandable and actionable. Unreliability is also known as the probability of failure. The probability of failure over time allows for further analysis and allows for flexible responses to be adjusted based on both technical requirements and the remaining operating time of the device. In this way, managing the optimal life of the device without risking failure is facilitated.

[0037] Preferably, the generation of the signal comprises calculating a remaining useful life RUL value.

[0038] The RUL value reflects the amount of time before the device is no longer considered useful according to prediction. The RUL is obtained by calculating the probability of failure over time. The predetermined threshold P EOL Defines the maximum acceptable unreliability of the device, 1-R(t), or the probability of device failure. The failure probability exceeds (succeeds) the threshold (i.e., 1-R ≥ P EOL ) at time t EOL is the predicted point in time at which the device can no longer be considered useful. Then, RUL can be obtained as EOL )=1-P EOL Given t EOL -t, the amount of time until the EOL.

[0039] Further signals indicating health may include life PDF, hazard function and mean time to failure (MTTF). Life PDF indicates the development of reliability over time and describes the a The net probability current flows. The lifetime PDF is obtained by calculating:

[0040] Where R is reliability.

[0041] The hazard function is the failure rate at time t conditional on survival until time t or later. It is obtained by calculating:

[0042] Where R is reliability.

[0043] MTTF is the expected (ie, predicted) average amount of time until a device fails. It is obtained by calculating:

[0044] Where R is reliability.

[0045] Note that MTTF is not necessarily equal to t EOL ; In fact, reliability at MTTF is often unacceptable, and t EOL Earlier. By calculating RUL values, predictions obtained from aging models of multiple devices become comparable and manageable. For example, overhaul and replacement bottlenecks that could lead to device failure can be identified in advance, thus preventing them from happening. In this way, managing a fleet of devices without risking failure is facilitated.

[0046] Preferably, the aging model includes an equation having at least two aging variables of the device; wherein calculating the probability density function includes calculating a joint probability density function of the at least two aging variables; wherein calculating the likelihood includes calculating the likelihood of the at least two aging variables and the first data; and wherein updating the probability density function includes updating the joint probability density function.

[0047] As described above, the present disclosure allows the use of a single aging variable μ or a vector of at least two aging variables Therefore, when the aging model is based on at least two aging variables When defining the device dynamics, the calculation of the probability density function includes calculating the at least two aging variables Similarly, in this case, the calculation of the likelihood includes calculating the joint probability density function of the at least two aging variables and the likelihood of the first data, wherein the observable corresponding to the first data needs to have a certain relationship with at least one of the two referenced aging variables. Similarly, in this case, the updating of the probability density function includes updating the joint probability density function.

[0048] When the aging model includes equations with at least two aging variables of the device, the predictions obtained from the aging model can have a higher dimensionality and thus be more reliable. The interdependencies between different failure scenarios can be identified more reliably. In this way, unexpected device failures are prevented.

[0049] Preferably, the aging model comprises an equation having aging variables for at least two devices; wherein calculating the probability density function comprises calculating a joint probability density function of the aging variables for each of these devices; wherein measuring observables comprises measuring one or more observables of these devices, and wherein the first data represents a measured value of the observable; wherein calculating the likelihood comprises calculating the likelihood of the aging variables and the first data; and wherein updating the probability density function comprises updating the joint probability density function.

[0050] As described above, the method is also suitable for monitoring a team of at least two devices. For example, in a system comprising multiple devices, it may be even more important to monitor these devices and predict the failure of a single device, which may affect the entire system.

[0051] In this case, measuring the observables includes measuring one or more observables of these devices. In other words, at least one observable is monitored for each of the monitored devices using the current hybrid monitoring method. The observables may be measured together for all devices at one time (e.g., at regular time intervals). The observables may be measured at different times (e.g., in the case when different devices or different sensors provide measurements at different intervals or based on events). The first data represents the measurement value of the observable, that is, the first data may represent any measurement value of any observable available at a given time. In the case of measuring observables at different times, the first data may represent the measurement value of only part of or even only a single observable at a given time.

[0052] The hybrid monitoring method according to the present disclosure can be combined with known monitoring methods. For example, a portion of the devices can be monitored as described above, while a portion of the devices can be modeled using an aging model without measurements and without updating the probability density function. In this way, existing devices that do not have any online monitoring capabilities can be included in the monitoring.

[0053] If the aging model comprises equations with aging variables for at least two devices, the method can be applied to a fleet of devices, where preventive maintenance is particularly important. Furthermore, the predictions obtained from the aging model can be of higher dimensionality and thus more reliable. Interdependencies between different devices and upcoming maintenance bottlenecks can be identified more reliably. In this way, unexpected device failures are efficiently prevented.

[0054] Preferably, the device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device and an energy storage device. For example, using a circuit breaker as a reference, at least one of its observable quantities must be considered, such as the opening / closing speed of the movable contact, the travel of the movable contact, the total travel of the movable contact, the overtravel of the movable contact, the rebound of the movable contact, the opening / closing time, the opening / closing peak coil current of the circuit breaker, or the current, voltage and further physical, electrical or chemical quantities of the transformer, the power electronic device or the energy storage device.

[0055] The power electronic device may be, for example, a converter, an inverter or a cycloconverter. The energy storage device may be, for example, a battery, a capacitor, a fuel cell or a supercapacitor.

[0056] Aging variables that allow prediction of device failure may include, for example, drive spring stiffness, friction constant or coil resistance.

[0057] By applying the method to devices selected from the group consisting of: circuit breakers, transformers, power electronics, and energy storage devices, failure prevention can be targeted at some of the most critical infrastructure devices. Further, accurate monitoring is enabled for devices that are difficult to observe and exhibit failure scenarios that are particularly difficult to prevent.

[0058] By applying the method to devices selected from the group consisting of: circuit breakers, transformers, power electronics, and energy storage devices, failure prevention can be targeted at some of the most critical infrastructure devices. Further, accurate monitoring is enabled for devices that are difficult to observe and exhibit failure scenarios that are particularly difficult to prevent.

[0059] By measuring one of the opening / closing speed of the movable contact of the circuit breaker, the travel of the movable contact, the total travel of the movable contact, the overtravel of the movable contact, the rebound of the movable contact, the opening / closing time of the circuit breaker, and the opening / closing peak coil current of the circuit breaker, the observable quantity is easily measurable and at the same time reliable and informative. In this way, the method can be easily applied to many devices and effectively prevent device failure.

[0060] Preferably, the method further comprises triggering maintenance, overhaul, replacement or load shedding of the device in response to the generated signal indicating a critical health prognosis within a predetermined amount of time.

[0061] For example, when a device is operated in a fleet of devices, if the reliability of the device is low, the load can be reduced for the device, and if the reliability of the device is high, the load can be increased for the device. In this way, the failure of the device can be delayed until maintenance or replacement of the device is possible.

[0062] In another example, replacement of a device may be automatically triggered. The replacement device may even already be installed and only needs to be activated once the generated signal indicates a critical health prognosis (eg, high probability of failure).

[0063] Critical health predictions may include, for example, a RUL that is too short, an unreliability that is too high, or similar health indicators that exceed and / or violate critical thresholds or criteria.

[0064] The predictive capabilities of the method are automatically and continuously utilized by triggering maintenance, overhaul, replacement or load shedding, thus preventing human errors that could lead to unnecessary equipment failures.

[0065] Preferably, the method further comprises indicating a warning in response to the generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction for future maintenance scheduling.

[0066] For example, when the predicted reliability value of a device falls below a predetermined reliability value (e.g., below 0.9 or 0.95) within a predetermined time (e.g., a week), a warning may be indicated. In response to the warning, the device may be scheduled for regular weekly maintenance, thus preventing device failure without requiring emergency action. At the same time, the warning allows the life of the wearable part or device to be used to the best extent possible, since the safety margin is not based on speculation, but on the best available knowledge about the technical state of the device.

[0067] By giving warnings in response to signals generated indicating critical health predictions within a predetermined amount of time, an organized and balanced maintenance schedule is facilitated. In this way, the health status of the devices in the fleet can be balanced, and device failures can be prevented efficiently.

[0068] The object is further solved by a system or apparatus for monitoring a device. The system comprises: at least one sensor configured to measure an observable quantity related to an operating condition of the device; a memory configured to store: an aging model, wherein the aging model comprises an end-of-life (EOL) boundary condition and at least one equation having an aging variable of the device; a controller or another device with appropriate computing power configured to: calculate a probability density function of the aging variable over time based on solving the at least one equation according to the aging model with the EOL boundary condition; obtain first data representing a measured value of the observable quantity; calculate a likelihood of the aging variable based on the first data; update the calculated probability density function of the aging variable based on the likelihood; and generate a signal indicating a health prediction of the device based on the probability density function, the aging model and the EOL boundary condition.

[0069] The object is further solved by a system or device for monitoring a device. The system includes: at least one sensor configured to measure an observable quantity related to an operating condition of the device; and a controller or another device with appropriate computing capabilities, configured to: calculate a probability density function of an aging variable over time based on solving at least one equation according to an aging model with EOL boundary conditions; obtain first data representing a measured value of the observable quantity; calculate the likelihood of the aging variable based on the first data; update the calculated probability density function of the aging variable based on the likelihood; and generate a signal indicating a health prediction of the device based on the probability density function, the aging model and the EOL boundary conditions. The aging model may include the EOL boundary conditions and the at least one equation with the aging variables of the device. The system or device may further include a memory configured to store the aging model.

[0070] The object is further solved by a system or device for monitoring a device. The system includes: at least one sensor configured to measure an observable quantity related to an operating condition of the device; and a controller or another device with appropriate computing power configured to: calculate a probability density function of an aging variable over time based on solving at least one equation according to an aging model with EOL boundary conditions, wherein the boundary conditions include a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition; obtain first data representing a measured value of the observable quantity; calculate the likelihood of the aging variable based on the first data; update the calculated probability density function of the aging variable based on the likelihood; and generate a signal indicating a health prediction of the device based on the probability density function, the aging model and the EOL boundary conditions. The aging model may include the EOL boundary conditions and the at least one equation with the aging variable of the device. The system or device may further include a memory configured to store the aging model.

[0071] The controller may comprise at least one integrated circuit.Alternatively or additionally, the controller may comprise a memory.

[0072] Various embodiments may preferably implement features corresponding to those described above with reference to the method, and these embodiments have corresponding technical effects and benefits.

[0073] The following items relate to preferred embodiments:

[0074] 1. A method for monitoring a device, comprising:

[0075] Providing or obtaining an aging model, wherein the aging model includes an end-of-life (EOL) boundary condition and at least one equation having an aging variable of the device;

[0076] calculating a probability density function of the aging variable over time based on solving the at least one equation according to the aging model with the EOL boundary condition;

[0077] Measuring observables related to the operating conditions of the device;

[0078] obtaining first data representing a measurement value of the observable;

[0079] Calculating the likelihood of the aging variable based on the first data;

[0080] updating the calculated probability density function of the aging variable based on the likelihood; and

[0081] A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

[0082] 2. A method for monitoring a device, the method comprising:

[0083] Calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model with end-of-life (EOL) boundary conditions;

[0084] Measuring observables related to the operating conditions of the device;

[0085] obtaining first data representing a measurement value of the observable;

[0086] Calculating the likelihood of the aging variable based on the first data;

[0087] updating the calculated probability density function of the aging variable based on the likelihood; and

[0088] A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

[0089] 3. The method according to clause 2, wherein the aging model comprises the EOL boundary condition and the at least one equation having aging variables of the device.

[0090] 4. The method according to any one of items 1 to 3, further comprising:

[0091] providing or obtaining a device model, wherein the device model includes equations having the aging variables as device parameters;

[0092] wherein the EOL boundary condition is based on the device model; and

[0093] Therein, the calculation of the likelihood is based on the device model.

[0094] 5. The method according to any one of items 1 to 4, further comprising:

[0095] providing or obtaining a device model, wherein the device model includes an equation having the observable and the aging variable;

[0096] wherein the monitoring model is based on the device model; and

[0097] Wherein, calculating the possibility is based on the monitoring model.

[0098] 6. A method according to any one of clauses 1 to 5, wherein generating the signal comprises: calculating a representation of the unreliability over time based on the probability density function, the aging model and the EOL boundary condition.

[0099] 7. A method according to any one of clauses 1 to 6, wherein generating the signal comprises calculating a remaining useful life RUL value.

[0100] 8. The method according to any one of items 1 to 7,

[0101] wherein the aging model includes equations having at least two aging variables of the device;

[0102] Wherein, calculating the probability density function includes calculating a joint probability density function of the at least two aging variables;

[0103] wherein calculating the likelihood comprises calculating the likelihood of the at least two aging variables and the first data; and

[0104] Wherein, updating the probability density function includes updating the joint probability density function.

[0105] 9. The method according to any one of items 1 to 8,

[0106] wherein the aging model includes equations having aging variables for at least two devices;

[0107] wherein calculating the probability density function comprises calculating a joint probability density function of the aging variables of each of the devices;

[0108] wherein measuring the observable comprises measuring one or more observables of the device, and wherein the first data represents a measured value of the observable;

[0109] Wherein, calculating the likelihood includes calculating the likelihood of the aging variable and the first data; and

[0110] Wherein, updating the probability density function includes updating the joint probability density function.

[0111] 10. The method according to any one of items 1 to 9,

[0112] wherein the device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device, and an energy storage device; and

[0113] The observable quantity is one of: an opening / closing speed of a movable contact of the circuit breaker, a stroke of the movable contact, a total stroke of the movable contact, an overtravel of the movable contact, a rebound of the movable contact, an opening / closing time of the circuit breaker, and an opening / closing peak coil current of the circuit breaker.

[0114] 11. The method according to any one of items 1 to 10, further comprising:

[0115] Maintenance, overhaul, replacement, or load shedding of the device is triggered in response to a generated signal indicating a critical health prognosis within a predetermined amount of time.

[0116] 12. The method according to any one of items 1 to 11, further comprising:

[0117] An alert is indicated in response to a generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction for future maintenance scheduling.

[0118] 13. A system or apparatus for monitoring a device, comprising:

[0119] at least one sensor configured to measure an observable quantity related to an operating condition of the device;

[0120] A memory configured to store:

[0121] an aging model, wherein the aging model includes end-of-life (EOL) boundary conditions and at least one equation having aging variables of the device;

[0122] A controller configured to:

[0123] calculating a probability density function of the aging variable over time based on solving the at least one equation according to the aging model with the EOL boundary condition;

[0124] obtaining first data representing a measurement value of the observable;

[0125] Calculating the likelihood of the aging variable based on the first data;

[0126] updating the calculated probability density function of the aging variable based on the likelihood; and

[0127] A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

[0128] 14. A system or apparatus according to item 13,

[0129] wherein the memory is further configured to store a device model, wherein the device model includes equations having aging variables as device parameters, and

[0130] wherein the EOL boundary condition is based on the device model; and / or the memory is configured to store a monitoring model, wherein the monitoring model includes an equation having the observable and the aging variable, and wherein the monitoring model is based on the device model; and

[0131] Therein, the controller is further configured to calculate the possibility based on the device model and / or based on the monitoring model.

[0132] 15. A system or device according to item 13 or 14, wherein the controller is further configured to: generate the signal using a representation of the probability of failure over time calculated based on the probability density function, the aging model and the EOL boundary condition and / or using a calculated remaining useful life RUL value.

[0133] 16. A system or apparatus according to any one of items 13 to 15,

[0134] Therein, the memory, the controller and the at least one sensor are configured to perform and / or implement the method according to any one of items 1 to 10.

[0135] 17. A system or apparatus according to any one of items 13 to 16,

[0136] wherein the memory is further configured to store the aging model, the aging model comprising an equation having at least two aging variables of the device;

[0137] Wherein, the controller is further configured to:

[0138] calculating the probability density using a joint probability density function that calculates the at least two aging variables;

[0139] calculating the likelihood using the at least two aging variables and the first data; and

[0140] The probability density function is updated using updating the joint probability density function.

[0141] 18. A system or apparatus according to any one of items 13 to 17,

[0142] wherein the memory is further configured to store the aging model, the aging model comprising an equation having aging variables of at least two devices;

[0143] Wherein, the controller is further configured to:

[0144] calculating the probability density using a joint probability density function that calculates the aging variable for each of the at least two devices;

[0145] Calculating the likelihood includes calculating the likelihood of the aging variable and the first data; and

[0146] updating the probability density function using updating the joint probability density function; and

[0147] Wherein, the sensor is further configured to measure one or more observable quantities of the at least two devices, wherein the first data represents a measured value of the observable quantity.

[0148] 19. A system or apparatus according to any one of items 13 to 18,

[0149] wherein the device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device, and an energy storage device; and

[0150] The sensor is further configured to measure the observable quantity, which includes one of the following: the opening / closing speed of the movable contact of the circuit breaker, the stroke of the movable contact, the total stroke of the movable contact, the overtravel of the movable contact, the rebound of the movable contact, the opening / closing time of the circuit breaker, and the opening / closing peak coil current of the circuit breaker, or the current, voltage and further physical, electrical or chemical quantity of the transformer, power electronic device or energy storage device.

[0151] 20. A system or apparatus according to any one of clauses 13 to 18, further comprising:

[0152] Wherein the controller is further configured to trigger maintenance, overhaul, replacement or load shedding of the device in response to the generated signal indicating a critical health prediction within a predetermined amount of time.

[0153] 21. A system or apparatus according to any one of clauses 13 to 20, further comprising:

[0154] Wherein the controller is further configured to: indicate a warning in response to the generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction for future maintenance scheduling.

[0155] 22. A system or apparatus for monitoring a device, the system or apparatus comprising:

[0156] at least one sensor configured to measure an observable quantity related to an operating condition of the device;

[0157] A controller configured to:

[0158] Calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model with end-of-life (EOL) boundary conditions;

[0159] obtaining first data representing a measurement value of the observable;

[0160] Calculating the likelihood of the aging variable based on the first data;

[0161] updating the calculated probability density function of the aging variable based on the likelihood; and

[0162] A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

[0163] 23. The system or apparatus according to clause 22, further comprising:

[0164] A memory is configured to store the aging model.

[0165] 24. The system or apparatus of clause 22 or 24, wherein the aging model comprises the EOL boundary condition and the at least one equation having aging variables of the device.

[0166] 25. A system or apparatus according to any one of items 22 to 24,

[0167] wherein a / the memory is further configured to store a device model, wherein the device model comprises equations having aging variables as device parameters, and

[0168] wherein the EOL boundary condition is based on the device model; and / or a / the memory is configured to store a monitoring model, wherein the monitoring model comprises an equation having the observable and the aging variable, and wherein the monitoring model is based on the device model; and

[0169] Therein, the controller is further configured to calculate the possibility based on the device model and / or based on the monitoring model.

[0170] 26. A system or device according to any one of items 22 to 25, wherein the controller is further configured to generate the signal using a representation of the probability of failure over time calculated based on the probability density function, the aging model and the EOL boundary condition and / or using a calculated remaining useful life, RUL, value.

[0171] 27. A system or apparatus according to any one of items 22 to 26,

[0172] Therein, the memory, the controller and the at least one sensor are configured to perform and / or implement the method according to any one of items 1 to 10.

[0173] 28. A system or apparatus according to any one of items 22 to 27,

[0174] wherein the memory is further configured to store the aging model, the aging model comprising an equation having at least two aging variables of the device;

[0175] Wherein, the controller is further configured to:

[0176] calculating the probability density using a joint probability density function that calculates the at least two aging variables;

[0177] calculating the likelihood using the at least two aging variables and the first data; and

[0178] The probability density function is updated using updating the joint probability density function.

[0179] 29. A system or apparatus according to any one of items 22 to 28,

[0180] wherein the memory is further configured to store the aging model, the aging model comprising an equation having aging variables of at least two devices;

[0181] Wherein, the controller is further configured to:

[0182] calculating the probability density using a joint probability density function that calculates the aging variable for each of the at least two devices;

[0183] Calculating the likelihood includes calculating the likelihood of the aging variable and the first data; and

[0184] updating the probability density function using updating the joint probability density function; and

[0185] Wherein, the sensor is further configured to measure one or more observable quantities of the at least two devices, wherein the first data represents a measured value of the observable quantity.

[0186] 30. A system or apparatus according to any one of items 22 to 29,

[0187] wherein the device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device, and an energy storage device; and

[0188] The sensor is further configured to measure the observable quantity, which includes one of the following: the opening / closing speed of the movable contact of the circuit breaker, the stroke of the movable contact, the total stroke of the movable contact, the overtravel of the movable contact, the rebound of the movable contact, the opening / closing time of the circuit breaker, and the opening / closing peak coil current of the circuit breaker, or the current, voltage and further physical, electrical or chemical quantity of the transformer, power electronic device or energy storage device.

[0189] 32. A system or apparatus according to any one of clauses 22 to 31, further comprising:

[0190] Wherein the controller is further configured to trigger maintenance, overhaul, replacement or load shedding of the device in response to the generated signal indicating a critical health prediction within a predetermined amount of time.

[0191] 33. A system or apparatus according to any one of clauses 22 to 32, further comprising:

[0192] Wherein the controller is further configured to: indicate a warning in response to the generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction for future maintenance scheduling.

[0193] According to a preferred embodiment of any one of items 1 to 33, the boundary condition includes a first boundary condition and a second boundary condition. The first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition. When the boundary conditions are distinguished into a first no-flux boundary condition and a second absorbing or partially absorbing boundary condition in this way, aging can be accurately modeled. Therefore, prediction can be more accurate, and maintenance can be more efficient and more effective.

[0194] The exemplary embodiments disclosed herein are directed to providing some features that will become readily apparent by reference to the following description when understood in conjunction with the accompanying drawings. According to various embodiments, exemplary systems, methods, devices, and computer program products are disclosed herein. However, it should be understood that these embodiments are presented by way of example only and not limitation, and it will be apparent to those of ordinary skill in the art who read this disclosure that various modifications may be made to the disclosed embodiments while still within the scope of this disclosure.

[0195] Therefore, the present disclosure is not limited to the exemplary embodiments and applications described and illustrated herein. Additionally, the specific order and / or hierarchy of steps in the methods disclosed herein are merely exemplary methods. Based on design preferences, the specific order or hierarchy of steps of the disclosed methods or processes can be rearranged while still within the scope of the present disclosure. Therefore, it will be understood by those of ordinary skill in the art that the methods and techniques disclosed herein present various steps or actions in a sample order, and unless otherwise expressly stated, the present disclosure is not limited to the specific order or hierarchy presented. BRIEF DESCRIPTION OF THE DRAWINGS

[0196] The above and other aspects and embodiments thereof are described in more detail in the drawings, the description and the claims.

[0197] Figure 1 A diagram of a method according to an embodiment of the present disclosure is shown.

[0198] Figure 2 A diagram of an aging model according to a further embodiment of the present disclosure is shown.

[0199] Figure 3 A diagram illustrating an exemplary prediction output according to a further embodiment of the present disclosure is shown.

[0200] FIG. 4 shows a graphical representation of an exemplary prediction output according to a further embodiment of the present disclosure.

[0201] Figure 5 A comparison of an exemplary prediction according to a known method and an exemplary prediction output according to a further embodiment of the present disclosure is shown.

[0202] Figure 6 A diagram of a system according to a further embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0203] Figure 1 A diagram of a method according to an embodiment of the present disclosure is shown.

[0204] The state of health (SOH) of the device is based on the aging variables Starting from the start of life (BOL) of the device (corresponding to the BOL point in the aging variable space), the prediction of the end of life (EOL) of the device consists of the following tasks:

[0205] In a first step 101, the device is modeled: model parameters including physics-based insights are defined algebraic equations or ordinary or partial differential equations and determine by experimental or other methods Value at BOL.

[0206] According to an embodiment, the first step 101 comprises: providing or obtaining a device model, wherein the device model comprises equations with aging variables as device parameters; wherein the EOL boundary conditions are based on the device model; and wherein the calculation of the likelihood is based on the device model.

[0207] The device model is based on the state variables The dynamics of the device are defined on a fast time scale of operation, thereby modeling the operational behavior of the device. In contrast, aging models are based on aging variables. Define the dynamics of the device on the slow time scale of aging, as described above. Treat the aging variables as continuous or quasi-continuous for the aging model Device parameters that are considered constant over relatively short operating time scales in device models

[0208] The device model can be used to target device parameters Thus, device parameter values ​​that lead to unacceptable behavior in the simulation is used to define EOL boundaries. Unacceptable behavior may, for example, include unstable behavior of state variables, state variables exceeding safety thresholds, and / or state variable behavior associated with another type of end of life for the device. The EOL boundaries may be obtained digitally or analytically from a device model.

[0209] If the monitoring observable is linked to the device model state (e.g., via monitoring operations), the calculation of the likelihood can be based on the device model. The device model then defines the state variable and device parameters This relationship can be expressed as the device parameters The state variable for the condition Then, in the first data In the case of is calculated as As defined by the device model.

[0210] By providing or obtaining a device model in step 101, the EOL boundary conditions better reflect the actual failure risk, and the probability better reflects the actual information contained in the measured values. Therefore, the prediction becomes more localized and more accurate. In this way, unexpected device failures are prevented and wasteful preventive maintenance is reduced.

[0211] In a second step 102, failure modes or unhealthy criteria are identified: The EOL hypersurface in the space of , where in the aging variable space of (multiple) failure modes k, the healthy state is For example, from the solution of the plant model equation (101) associated with the fault.

[0212] In a third step 103, the aging of the device is modeled: a stochastic differential equation (SDE) or a Fokker-Planck equation (FPE) for the aging model is defined based on physical insights, and parameter values ​​are determined experimentally or by other methods. Some of the parameters are more specific to a single device, while others may be related to statistical moments of quantities across a group of devices. The quantity may eventually turn out to be better viewed as a random auxiliary variable This expands the aging model variable space. The identification of an aging model is a modeling task that may require empirical data from the field and / or specialized experiments.

[0213] According to one embodiment, the aging model includes an equation having at least two aging variables of the device; wherein calculating the probability density function includes calculating a joint probability density function of the at least two aging variables; wherein calculating the likelihood includes calculating the likelihood of the at least two aging variables and the first data; and wherein updating the probability density function includes updating the joint probability density function.

[0214] As described above, the present disclosure allows the use of a single aging variable μ or a vector of at least two aging variables Therefore, when the aging model is based on at least two aging variables When defining the device dynamics, the calculation of the probability density function includes calculating the at least two aging variables Similarly, in this case, the calculation of the likelihood includes calculating the joint probability density function of the at least two aging variables and the likelihood of the first data, wherein the observable corresponding to the first data needs to have a certain relationship with at least one of the at least one aging variable. Similarly, in this case, the updating of the probability density function includes updating the joint probability density function.

[0215] When the aging model includes equations with at least two aging variables of the device, the predictions obtained from the aging model can be of higher dimensionality and thus more reliable. Interdependencies between different failure scenarios can be identified more reliably. In this way, unexpected device failures are prevented.

[0216] According to one embodiment, the aging model includes equations having aging variables of at least two devices; wherein calculating the probability density function includes calculating a joint probability density function of the aging variables for each of these devices; wherein measuring observables includes measuring one or more observables of these devices, and wherein first data represents a measured value of the observable; wherein calculating the likelihood includes calculating the likelihood of the aging variables and the first data; and wherein updating the probability density function includes updating the joint probability density function.

[0217] The method as described above is also suitable for monitoring a team of at least two devices. For example, in a system comprising multiple devices, it may be even more important to monitor these devices and predict the failure of a single device, which may affect the entire system.

[0218] In this case, measuring the observables includes measuring one or more observables of these devices. In other words, at least one observable is monitored for each of the monitored devices using the current hybrid monitoring method. The observables may be measured together for all devices at one time (e.g., at regular time intervals). The observables may be measured at different times (e.g., in the case when different devices or different sensors provide measurements at different intervals or based on events). The first data represents the measurement value of the observable, that is, the first data may represent any measurement value of any observable available at a given time. In the case of measuring observables at different times, the first data may represent the measurement value of only part of or even only a single observable at a given time.

[0219] The hybrid monitoring method according to the present disclosure can be combined with known monitoring methods. For example, a portion of the devices can be monitored as described above, while a portion of the devices can be modeled using an aging model without measurements and without updating the probability density function. In this way, existing devices that do not have any online monitoring capabilities can be included in the method.

[0220] If the aging model comprises equations with aging variables for at least two devices, the method can be applied to a fleet of devices, where preventive maintenance is particularly important. Furthermore, the predictions obtained from the aging model can be of higher dimensionality and thus more reliable. Interdependencies between different devices and upcoming maintenance bottlenecks can be identified more reliably. In this way, unexpected device failures are efficiently prevented.

[0221] In a fourth step 104, monitoring is modeled: observables are determined and variables / parameters and The relationship between , for example, based on experiments or appropriate physical models. Preferably, direct measurement values, such as using normal physical sensors or using soft sensing, such as using soft sensors / virtual sensors.

[0222] According to an embodiment, the fourth step 104 comprises: providing or obtaining a monitoring model, wherein the monitoring model comprises equations with observables and aging variables; wherein the monitoring model is based on a device model; and wherein calculating the likelihood is based on the monitoring model.

[0223] Observables should be available for measurement (e.g., using sensors). They do not necessarily need to include or aging variables As described above, the observable At least some aging variables A certain relationship is shown. This relationship can be defined by the monitoring model and can be based on the device model.

[0224] The monitoring model consists of observable and aging variables This equation (which may include multiple sub-equations) defines the observable and aging variables For example, when the relationship is a random correlation, the equation can include a definition of conditional probability.

[0225] For example, in the observable The distribution corresponds to a very precise value of, for example, μ1 = a and contains no information about other aging variables. It can be a normalized Gaussian function of μ1 that peaks sharply at a and is independent of the other conditioning variables. In other words, in this case, the monitoring model comprises a Gaussian function (or other approximation of a Dirac delta function) that is strongly localized.

[0226] By providing or obtaining a monitoring model, the actual information contained in the measured values ​​can be extracted more thoroughly and accurately. As a result, predictions become more localized and more accurate. In this way, unexpected equipment failures are prevented and wasteful preventive maintenance is reduced.

[0227] In the fifth step 105, the prediction is initialized: the initial values ​​of all variables are determined / defined, and the parameters at BOL must be determined and Because it will target FPE is used in the program, so the initial distribution should be given An exemplary Gaussian distribution with an appropriate standard deviation will be used below to illustrate this initial distribution. Other distributions are also applicable without further complication.

[0228] In the sixth step 106, the reliability and life probability are calculated: for the case with initial conditions of The SDE or FPE of the aging model is solved until EOL is reached. Then, R(t) is calculated as described above, that is, Based on the reliability R(t) and the predetermined threshold, the lifetime t is determined as described above. EOL , and thus by calculating RUL(t)=t EOL -t to give the RUL.

[0229] In a seventh step 107, a diagnostic measurement is performed: at time t m, additional information is obtained through monitoring. Monitoring first provides binary information about death or survival, that is, whether the device has failed since the last information was obtained. In the latter case, the probability distribution can be derived For example, if the information corresponds to a very precise value (say, μ1=a) and contains no information about other device model parameters, then the distribution p m (μ1) can be a normalized Gaussian function of μ1 that peaks sharply at a (i.e., is strongly localized) and is independent of the other conditioning variables.

[0230] In an eighth step 108, the prediction (106) is updated: In the first prediction sub-step, the solution of the SDE or FPE of the aging model predicted that m The prior probability In the second correction sub-step, the prediction must be corrected with the aid of the diagnostic measurements from the seventh step (107). The updated probability can be calculated using a Bayesian inference procedure as follows (which is also called the posterior probability):

[0231]

[0232] The normalization constant is C and the probability is And the prior distribution is p ＜ The normalization is justified because, according to the assumption, one knows for sure from the diagnosis whether the device is alive or not. In exceptional cases where this binary information is not available, the constant C must be determined from the knowledge given about the probability of being in the fault state. After the correction step, it can be determined at t m With new initial conditions Start over again to complete the RUL calculation steps.

[0233] In the possibility p m and the prior probability p ＜ For the Gaussian case, they are conjugate and p ＞ is also Gaussian. The means and standard deviations of these three distributions then satisfy the relationship for the particular Kalman filter update. However, in general, as a solution to the FPE It is not Gaussian but depends on the underlying aging model; therefore, this approach is able to cover more general real-world cases.

[0234] In the ninth step 109, using the corrected state, from time t m The process starts over again with the sixth step 106 .

[0235] When the device reaches EOL, that is, when the true reliability is R(t EOL )=1-P EOL , the method stops.

[0236] At any point in the method, a signal indicating the health prediction of the device may be generated. The health prediction of the device is based on the probability density function, the aging model, and the EOL boundary conditions. For example, a signal indicating the health prediction of the device may be generated after the sixth step 106 or after the eighth step 108.

[0237] According to one embodiment, the generation of the signal includes computing a representation of unreliability over future time based on a probability density function, an aging model, and EOL boundary conditions.

[0238] The unreliability over time (also called the probability of failure over time) is here a function of the probability that the output device will fail at any given future time point. The probability of failure at a particular time point can also be referred to as the unreliability of the device at that time point. For a given time point, by a The marginalized PDF (defined by solving the aging model) is used to calculate the failure probability.

[0239] In the case of absorbing EOL boundary conditions, the aging variables Leaving the acceptance region through the surface defined by the EOL boundary condition, i.e. In this way, device failures are concentrated on the surface defined by the EOL boundary condition. The above marginalized PDF (defined by solving the aging model) is used to calculate the failure probability, that is, The failure probability is the complement of the reliability R, R = 1-P EOI .

[0240] By calculating a representation of the unreliability over time, the predictions obtained from the aging model become understandable and actionable. The prediction of the probability of failure over time allows further analysis and allows flexible responses to be adjusted based on both technical requirements and the remaining operating time of the device. In this way, it facilitates managing the optimal life of the device without risking failure.

[0241] According to one embodiment, the generation of the signal comprises calculating a remaining useful life RUL value.

[0242] The RUL value reflects the amount of time remaining until the device is no longer considered useful, as predicted based on operational requirements. The RUL is obtained by calculating the probability of failure over time. A predetermined threshold for the probability of failure defines the acceptable unreliability of the device. EOL (i.e., 1-R(t EOL )≥P EOL ) at time t EOLis the predicted point in time at which the device can no longer be considered useful. Then, the RUL can be obtained as t EOL -t, that is, until t EOL The amount of time until.

[0243] Further signals indicating health may include life PDF, hazard function and mean time to failure (MTTF). Life PDF indicates the development of reliability over time and describes the a The net probability current flows. It is obtained by calculating the following formula:

[0244] Where R is reliability.

[0245] The hazard function is the failure rate at time t conditional on survival until time t or later. It is obtained by calculating:

[0246] Where R is reliability.

[0247] MTTF is the expected (i.e., predicted) mean amount of time for a device to fail. It is obtained by calculating:

[0248] Where R is reliability. Note that MTTF is not necessarily equal to t EOL ; In fact, reliability at MTTF is often unacceptable, and t EOL Earlier.

[0249] By calculating the RUL value, the predictions obtained from the aging models of multiple devices become comparable and manageable. For example, overhaul and replacement bottlenecks that may lead to device failure can be identified in advance, thus preventing device failure. In this way, it is facilitated to manage a fleet of devices without risking failure.

[0250] According to one embodiment, the method further comprises triggering maintenance, overhaul, replacement, or load shedding of the device in response to the generated signal indicating a critical health prognosis within a predetermined amount of time.

[0251] For example, when a device is operated in a fleet of devices, if the reliability of the device is low, the load can be reduced for the device, and if the reliability of the device is high, the load can be increased for the device. In this way, the failure of the device can be delayed until maintenance or replacement of the device is possible.

[0252] In another example, replacement of a device may be automatically triggered. The replacement device may even already be installed and only needs to be activated once the generated signal indicates a critical health prognosis (eg, high probability of failure).

[0253] The predictive capabilities of the method are automatically and continuously utilized by triggering maintenance, overhaul, replacement or load shedding, thus preventing human errors that could lead to unnecessary equipment failures.

[0254] According to one embodiment, the method further comprises indicating a warning in response to the generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction for future maintenance scheduling.

[0255] For example, a warning may be indicated when a device is expected to fall below a predetermined reliability (e.g., below 0.9) within a predetermined time (e.g., within a week or a month). In response to the warning, the device may be scheduled for regular weekly or monthly maintenance, thus preventing device failure without requiring emergency action. At the same time, the warning allows the life of the wearable part or device to be used to the best extent possible, since the safety margin is not based on speculation, but on the best available knowledge about the technical state of the device.

[0256] By indicating a warning in response to a generated signal indicating a critical health prediction within a predetermined amount of time, an organized and balanced maintenance schedule is facilitated. In this way, the health status of devices in a fleet can be balanced and device failures can be prevented efficiently.

[0257] According to one embodiment, the device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device, and an energy storage device; and the at least one observable is one of: an opening / closing speed of a movable contact of the circuit breaker, a travel of the movable contact, a total travel of the movable contact, an overtravel of the movable contact, a rebound of the movable contact, an opening / closing time of the circuit breaker, and an opening / closing peak coil current of the circuit breaker. Additionally or alternatively, the observable may include a current value, a voltage value, and / or a state of charge value.

[0258] The power electronic device may be, for example, one of a converter, an inverter or a cycloconverter. The energy storage device may be, for example, a battery, a capacitor, a fuel cell or a supercapacitor.

[0259] Aging variables of the circuit breaker that allow prediction of device failure may include, for example, the drive spring stiffness, the friction constant or the trip coil resistance.

[0260] By applying the method to devices selected from the group consisting of: circuit breakers, transformers, power electronics, and energy storage devices, failure prevention can be targeted at some of the most critical infrastructure devices. Further, accurate monitoring is enabled for devices that are difficult to observe and exhibit failure scenarios that are particularly difficult to prevent.

[0261] By applying the method to devices selected from the group consisting of: circuit breakers, transformers, power electronics, and energy storage devices, failure prevention can be targeted at some of the most critical infrastructure devices. Further, accurate monitoring is enabled for devices that are difficult to observe and exhibit failure scenarios that are particularly difficult to prevent.

[0262] By measuring one of the opening / closing speed of the movable contact of the circuit breaker, the travel of the movable contact, the total travel of the movable contact, the overtravel of the movable contact, the rebound of the movable contact, the opening / closing time of the circuit breaker, and the opening / closing peak coil current of the circuit breaker, the observable quantity is easily measurable and at the same time reliable and informative. In this way, the method can be easily applied to many devices and effectively prevent device failure.

[0263] Figure 2 A diagram showing an aging model for a device according to a further embodiment of the present disclosure is shown. The aging model includes aging variables not depicted in the diagram These aging variables spans a variable space 200, which may also be referred to as a condition space or This variable space is illustrated in a two-dimensional plane because the example shows the vectors of two aging variables

[0264] The variable space includes the region (acceptance region) associated with the survival condition (associated with the device working properly) Ω a 201, the region is defined by the boundary surface having the boundary conditions of the aging model equations (i.e., SDE or FPE) 202 defines the boundary surface 202. The boundary surface 202 includes: a first boundary (illustrated as a straight thick line), which corresponds to a no-flux boundary condition; and a second boundary (illustrated as a light curve), which corresponds to an absorbing EOL boundary condition. More generally, the boundary Some parts of act as absorbers, i.e. they are EOL boundaries, while other parts of the boundary can be "closed" and described by a zero probability flux. Figure 2 Sources in the interior of the receiving area or in the boundary conditions corresponding to repair and overhaul are not shown in FIG, but these are also possible. The boundary conditions can also be partially absorbing boundary conditions, which are Figure 2 This is not shown, but is possible. The aging variable reaches the EOL boundary condition when the value of Can be absorbed, for example, when the corresponding device is removed (or fails). Figure 2In the embodiment illustrated in FIG. 1 , the boundary conditions include a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition. Therefore, aging can be accurately modeled. In this way, predictions can be more accurate, and maintenance can be more efficient and more effective.

[0265] The sketch illustrates the aging process in terms of the slow change of the aging variable over a long time scale. In the acceptance region, the aging process starts at the beginning of life (BOL) point 203, which corresponds to the beginning of the device's life and reliability R=1. The diagram shows the transition from the point BOL toward the boundary surface The trajectory is illustrated with the aid of arrows indicating the predicted highest probability paths and paths of probability density functions (PDFs) that become less localized, i.e., broader, over time. For any time point t1, the first PDF is selected from the solutions of the SDE or FPE 204. For another time point t2, another PDF close to the boundary surface is selected from the path. 205. When located in the region Ω a The probability R in the EOL ) or less, the device reaches the end of its life.

[0266] Figure 3 A diagram illustrating an exemplary prediction output according to a further embodiment of the present disclosure is shown.

[0267] The diagram shows an exemplary prediction output of an aging process for a single aging variable μ, which is plotted on a vertical scale against time, which is plotted on a horizontal scale. The prediction of the aging variable is illustrated by means of a trace 301a, which represents the time from The most likely path obtained. The aging variable PDF 302 for the time point t0=0 is specifically mentioned on the far left of the illustration. It can be seen that the trajectory 301a corresponds to the peak of the PDF 302. For the trajectory 301a, the RUL is calculated, which is illustrated when the trajectory reaches the EOL boundary given by the red line 304. At the time point t1, the measured value 303 is illustrated by the vertical line. At this time point, the prediction is corrected by updating the PDF, as described in this article. This is illustrated by the change of the starting point of the trajectory 301b from this time point. A new prediction is made again until the EOL boundary 304, which is represented by the horizontal line limiting the dimension of the aging variable. The life PDF 305 is shown at the top of the illustration. It can be seen that the point where the trajectory 301 crosses the EOL boundary 304 corresponds to the peak of the life PDF, which illustrates the EOL probability over time. The following illustrates a graph 306 of the failure probability (unreliability) over time, which corresponds to the cumulative form of the life PDF. Thus, the point at which trajectory 301 crosses EOL boundary 304 corresponds to a sharp increase in failure probability 1-R(t) (ie, a sharp decrease in reliability R(t)).

[0268] FIG. 4 shows a graphical representation of an exemplary prediction output according to a further embodiment of the present disclosure.

[0269] In sub-figure a), three predicted life curves are plotted for different aging rates for a two-dimensional aging variable space (μ1, μ2), where the EOL boundaries are not shown here. The figure shows the life in years on the vertical axis and the failure probability in percentage on the horizontal axis. Each curve shows the life as a function of the failure probability for a specific aging rate of the second variable μ2, while the aging rate of μ1 is constant. For example, when the aging rate of the second variable μ2 is f2=0V / y, a failure probability of 70% corresponds to a life of 40 years. In other words, if the device is no longer considered useful at a failure probability of more than 70%, the device can be expected to be useful in 40 years. As can be seen in sub-figure a), when very low failure probabilities are not considered acceptable, the life decreases particularly quickly, and when very high failure probabilities are considered acceptable, the life increases particularly quickly.

[0270] In sub-figure b), three reliability curves are plotted for the same different aging rates as in sub-figure a). The figure shows the reliability as a value between 0 and 1 on the vertical axis and the lifetime in years on the horizontal axis. Each curve shows the reliability as a function of time for a specific aging rate of the second aging variable. For example, when the aging rate is f2=0 V / y, the expected reliability after a lifetime of 40 years is 0.3, which corresponds to a failure rate of 70%. In other words, after 40 years, the reliability of the device is expected to drop below 0.3. As can be seen in sub-figure b), the predicted reliability drops particularly fast within a short critical time frame, depending on the aging rate. For example, for an aging rate of f2=-0.8 V / y, the reliability drops from 0.9 to 0.2 within 5 years after the 22nd year. Similar sharp reliability drops can be observed after the 25th year and after the 35th year for an aging rate of f2 = -0.4 V / y and an aging rate of f2 = -0 V / y, respectively.

[0271] Thus, the predicted reliability curve allows identification of times after which reliability deteriorates rapidly and after which it would therefore be particularly effective to consider maintenance or replacement.

[0272] In sub-figure c), three failure probability density curves are plotted for the same different aging rates as in sub-figure a). The figure shows the failure probability density on the vertical axis and the time in years on the horizontal axis. Each curve shows the failure probability density as a Gaussian function of time for a specific aging rate. For example, for an aging rate of f2 = -0.8V / y, the failure probability density has a peak at time t = 24y. In other words, the most likely time for failure is around the 24th year. This corresponds to a reliability drop around that time as observed in sub-figure b). The peak in the failure probability density curve corresponds to a time frame of sharp decline (as an observable observed in sub-figure b)).

[0273] In general, the failure probability density can be obtained as the derivative of the failure probability with respect to time. In other words, the failure probability density can be obtained as 1 minus the derivative of the reliability with respect to time. Conversely, the reliability can be obtained as 1 minus the integral of the failure probability density with respect to time.

[0274] The three curves according to sub-figure c) show different degrees of localization. The more localized the curve is, the more accurately the failure time is predicted. Since the solution of the FPE results in a broadening of the probability density, updating the calculated probability density function of the aging variable based on the likelihood according to an embodiment allows the failure probability density curve to be more localized.

[0275] Each of the life curve, reliability curve, and failure probability density curve as described above is a signal indicative of health prediction. Thus, the signal may include any of the curves, any value according to any of these curves, and / or a remaining life value (RUL value), a reliability value, an unreliability value, a failure probability value, or a similar value or curve related to predicted device health and / or predicted device reliability.

[0276] Figure 5 A comparison of an exemplary prediction according to a known method and an exemplary prediction output according to a further embodiment of the present disclosure is shown.

[0277] exist Figure 5 In the graph shown in , RUL is plotted on the vertical axis against time on the horizontal axis.

[0278] The dotted line 501 shows a deterministic remaining useful life (RUL) prediction. The RUL corresponds to the difference between the predicted useful life and the time elapsed. Figure 5 In the example shown in , the predicted deterministic useful life is 37 years. Therefore, for example at time t=10y, the RUL is RUI=37y-10y=27y. The deterministic RUL prediction is simply a linear continuation of the initial prediction of the useful life.

[0279] The solid line curve 502 shows the updated remaining useful life (RUL) prediction according to an embodiment. The RUL can be calculated as the difference between the updated RUL (i.e., the updated prediction of the remaining useful life) and the time that has passed since the update. For example, at time t5=5y, the updated useful life is 25.5 years. Therefore, for example, at time t6=6y=t5+1y (one year after the update), the RUL can be calculated as RUL=25.5y-1y=24.5y However, at time t6=6y, the updated useful life is 23.7 years. Therefore, Figure 5 The graph in shows a discontinuity at time t6=6y. The updated RUL prediction does not correspond to a linear continuation of the previous prediction of the service life.

[0280] As in Figure 5 As can be seen in FIG. 5 , the difference between the updated RUL curve 502 and the deterministic RUL curve 501 is that RUL 502 is shorter than 501. This difference is due to the fact that in this case, the deterministic RUL is close to the MTTF and the minimum acceptable reliability 1-P EOL is close to 50%, while RUL 501 corresponds to the minimum acceptable reliability 1-P EOL=95%. The second difference between the updated RUL curve 502 and the deterministic RUL curve 501 is the fluctuation of time. If the deterministic RUL method is to be used to prevent the failure of the device, a safety margin needs to be subtracted from the deterministic RUL. In this case, the safety margin corresponds to the maximum value of the difference between the updated RUL curve 502 and the deterministic RUL curve 501. If the updated RUL method is to be used to prevent the failure of the device, a much smaller safety margin is required. In this case, the safety margin corresponds to the maximum difference between the RUL according to one update and the RUL for the same time instance calculated according to subsequent updates. This safety margin is typically smaller than the safety margin according to the deterministic RUL method. Therefore, using the updated RUL method, the device can be put into use for a longer time. In this way, system efficiency and system performance can be improved.

[0281] Figure 6 A diagram of a system according to a further embodiment of the present disclosure is shown.

[0282] The system 600 comprises at least one sensor 601. The sensor is configured to measure an observable quantity related to an operating condition of the device.

[0283] The system 600 further includes a memory 610. The memory 610 is configured to store an aging model 611, wherein the aging model 611 includes an end-of-life EOL boundary condition and at least one equation with aging variables of the device.

[0284] Further, the memory is configured to store a device model 612. The device model 612 includes equations with aging variables as device parameters. In this case, the EOL boundary condition is based on the device model 612. The calculation of the likelihood is based on the device model 612. According to other embodiments, the memory may not store the device model 612.

[0285] Further, the memory is configured to store a monitoring model 613. The monitoring model 613 includes equations with observables and aging variables. The monitoring model 613 is based on the device model 612. The calculation of the likelihood is based on the monitoring model 613. According to other embodiments, the memory may not store the monitoring model 613.

[0286] Memory 610 may store any computer programs and / or models according to any of the methods disclosed herein.

[0287] The system 600 further includes a controller 620. The controller is configured to calculate a probability density function of an aging variable over time based on solving the at least one equation according to the aging model 611 with EOL boundary conditions. The controller is further configured to obtain first data representing a measurement value of an observable quantity from the sensor 601. The controller is further configured to calculate a likelihood of the aging variable based on the first data. The controller is further configured to update the calculated probability density function of the aging variable based on the likelihood. The controller is further configured to generate a signal indicating a health prediction of the device based on the probability density function, the aging model 611, and the EOL boundary conditions.

[0288] The controller 620 may be configured to perform any of the methods disclosed herein. The controller 620 may access data stored in the memory 610 and / or write data to the memory 610. The controller 620 may also obtain data from the sensor 601 and / or trigger a measurement of the sensor 601. The controller 620 may include at least one integrated circuit. As an alternative or in addition, the controller 620 may include the memory 610.

[0289] Although various embodiments of the present disclosure have been described above, it should be understood that they are presented only by way of example and not by way of limitation. Similarly, various figures may depict example architectures or configurations, which are provided to enable those of ordinary skill in the art to understand the exemplary features and functions of the present disclosure. However, such personnel will understand that the present disclosure is not limited to the illustrated example architectures or configurations, but may be implemented using various alternative architectures and configurations. Additionally, as will be understood by those of ordinary skill in the art, one or more features of an embodiment may be combined with one or more features of another embodiment described herein. Therefore, the breadth and scope of the present disclosure should not be limited by any of the exemplary embodiments described above.

[0290] It should also be understood that any reference to an element herein using names such as "first," "second," etc. does not generally limit the number or order of those elements. Rather, these names may be used herein as a convenient means of distinguishing between two or more elements or instances of an element. Thus, a reference to a first element and a second element does not mean that only two elements may be employed, or that the first element must be located before the second element in some manner.

[0291] Additionally, those of ordinary skill in the art will appreciate that information and signals may be represented using any of a variety of different techniques and technologies. For example, data, instructions, commands, information, signals, bits, and symbols such as those referenced in the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0292] The skilled person will further understand that any of the various illustrative logical blocks, units, processors, devices, circuits, methods, and functions described in conjunction with the aspects disclosed herein may be implemented by electronic hardware (e.g., digital implementations, analog implementations, or a combination of both), firmware, various forms of programs or design codes combined with instructions (which for convenience may be referred to herein as "software" or "software units"), or any combination of these technologies.

[0293] In order to clearly illustrate this interchangeability of hardware, firmware and software, various illustrative components, blocks, units, circuits and steps have been generally described above with respect to their functions. Whether such functions are implemented as hardware, firmware or software or a combination of these technologies depends on specific applications and the design constraints imposed on the entire system. Technicians can implement the described functions in various ways for each specific application, but such implementation decisions will not cause deviations from the scope of the present disclosure. According to various embodiments, processors, devices, components, circuits, structures, machines, units, etc. can be configured to perform one or more of the functions described herein. As used herein with respect to a specified operation or function, the term "configured to" or "configured to be used for" refers to a processor, device, component, circuit, structure, machine, unit, etc. that is physically constructed, programmed and / or arranged to perform the specified operation or function.

[0294] In addition, the skilled person will understand that the various illustrative methods, logic blocks, units, devices, components and circuits described herein may be implemented in or performed by an integrated circuit (IC), which may include: a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device or any combination thereof. The logic blocks, units and circuits may further include antennas and / or transceivers to communicate with various components within a network or within a device. The general-purpose processor may be a microprocessor, but in an alternative, the processor may be any conventional processor, controller or state machine. The processor may also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a DSP core, or any other suitable configuration to perform the functions described herein. If implemented in software, the functions may be stored on a computer-readable medium as one or more instructions or codes. Therefore, the steps of the methods or algorithms disclosed herein may be implemented as software stored on a computer-readable medium.

[0295] In this document, the term "unit" as used herein refers to any combination of software, firmware, hardware, and other elements for performing the associated functions described herein. Additionally, for the purpose of discussion, various units are described as discrete units; however, as will be apparent to one of ordinary skill in the art, two or more units may be combined to form a single unit that performs associated functions according to embodiments of the present disclosure.

[0296] Additionally, memory or other storage devices and communication components may be used in embodiments of the present disclosure. It will be appreciated that, for purposes of clarity, the above description has described embodiments of the present disclosure with reference to different functional units and processors. However, it will be apparent that any suitable distribution of functions between different functional units, processing logic elements, or domains may be used without devaluing the present disclosure. For example, functions illustrated as being performed by separate processing logic elements or controllers may be performed by the same processing logic element or controller. Therefore, references to specific functional units are only references to appropriate means for providing the described functions, rather than indicating a strict logical or physical structure or organization.

[0297] Various modifications to the embodiments described in this disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the claims. Therefore, the present disclosure is not intended to be limited to the embodiments shown herein, but is to be given the widest scope consistent with the novel features and principles disclosed herein, as recited in the appended claims.

Claims

1. A method for monitoring a device, comprising: Calculating a probability density function of an aging variable over time based on solving at least one equation according to an aging model having end-of-life (EOL) boundary conditions, wherein the boundary conditions include a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition; measuring an observable quantity related to an operating condition of the device; obtaining first data representing a measurement value of the observable; calculating the likelihood of the aging variable based on the first data; updating the calculated probability density function of the aging variable based on the likelihood; and A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

2. The method according to claim 1, further comprising: providing a device model, wherein the device model includes equations having the aging variables as device parameters; wherein the EOL boundary condition is based on the device model; and Therein, the calculation of the likelihood is based on the device model.

3. The method according to claim 1 or claim 2, further comprising: providing a monitoring model, wherein the monitoring model includes an equation having the observable and the aging variable; wherein the monitoring model is based on the device model; and Wherein, calculating the likelihood is based on the monitoring model.

4. The method according to any one of claims 1 to 3, wherein: Generating the signal includes computing a representation of unreliability over time based on the probability density function, the aging model, and the EOL boundary condition.

5. The method according to any one of claims 1 to 4, wherein: Generating the signal includes calculating a remaining useful life RUL value.

6. The method according to any one of claims 1 to 5, in, The aging model includes an equation having at least two aging variables of the device; Wherein, calculating the probability density function comprises calculating a joint probability density function of the at least two aging variables; wherein calculating the likelihood comprises calculating the likelihood of the at least two aging variables and the first data; and Wherein, updating the probability density function includes updating the joint probability density function.

7. The method according to any one of claims 1 to 6, in, The aging model includes an equation having aging variables for at least two devices; wherein calculating the probability density function comprises calculating a joint probability density function of the aging variables of each of the devices; wherein measuring the observable comprises measuring one or more observables of the device, and wherein the first data represents a measured value of the observable; wherein calculating the likelihood comprises calculating the likelihood of the aging variable and the first data; and Wherein, updating the probability density function includes updating the joint probability density function.

8. The method according to any one of claims 1 to 7, in, The device is selected from the group consisting of: a circuit breaker, a transformer, a power electronic device, and an energy storage device; and The observable quantity is one of the following: an opening / closing speed of a movable contact of the circuit breaker, a stroke of the movable contact, a total stroke of the movable contact, an overtravel of the movable contact, a rebound of the movable contact, an opening / closing time of the circuit breaker, and an opening / closing peak coil current of the circuit breaker.

9. The method according to any one of claims 1 to 8, further comprising: Maintenance, overhaul, replacement, or load shedding of the device is triggered in response to a generated signal indicating a critical health prognosis within a predetermined amount of time.

10. The method according to any one of claims 1 to 9, further comprising: An alert is indicated in response to the generated signal indicating a critical health prediction within a predetermined amount of time, the critical health prediction being used for future maintenance scheduling.

11. A system for monitoring a device, comprising: at least one sensor configured to measure an observable related to an operating condition of the device; A memory configured to store: an aging model, wherein the aging model comprises an end-of-life (EOL) boundary condition and at least one equation having aging variables of the device, wherein the boundary condition comprises a first boundary condition and a second boundary condition, wherein the first boundary condition is a no-flux boundary condition and the second boundary condition is an absorbing or partially absorbing boundary condition; A controller, the controller being configured to: calculating a probability density function of the aging variable over time based on solving the at least one equation according to the aging model with the EOL boundary conditions; obtaining first data representing a measurement value of the observable; calculating the likelihood of the aging variable based on the first data; updating the calculated probability density function of the aging variable based on the likelihood; and A signal indicative of a health prediction of the device is generated based on the probability density function, the aging model, and the EOL boundary condition.

12. The system according to claim 11, in, The memory is further configured to store a device model, wherein the device model includes equations having aging variables as device parameters, and wherein the EOL boundary condition is based on the device model; and / or the memory is configured to store a monitoring model, wherein the monitoring model comprises an equation having the observable and the aging variable, and wherein the monitoring model is based on the device model; and Therein, the controller is further configured to calculate the possibility based on the device model and / or based on the monitoring model.

13. The system according to claim 11 or 12, wherein: The controller is further configured to generate the signal using a representation of a probability of failure over time calculated based on the probability density function, the aging model and the EOL boundary condition and / or using a calculated remaining useful life RUL value.

14. A system according to any one of claims 11 to 13, in, The memory, the controller and the at least one sensor are configured to perform and / or implement the method according to any one of claims 1 to 10.

Images