Method and device for operating a device battery in a way that optimizes its service life after detecting a current or impending anomaly.

Abstract

The invention relates to a method for operating a device battery in a technical device (4), in particular after detecting an anomaly (S3) of the device battery (41), comprising the following steps: - providing (S1) temporal operating variable profiles (F) of several operating variables of a specific device battery (41); - predicting the temporal operating variable profiles (F) in order to obtain artificial operating variable profiles (F) depending on the provided temporal operating variable profiles (F) and a plurality of load parameters of a load pattern using a usage pattern model (10) which is designed to provide artificial profiles of at least one operating variable depending on the plurality of load parameters, wherein the plurality of load parameters correspond to stress factors for the aging of the device battery (41); - determining the remaining service life (RUL) of the device battery (41) depending on the artificial operating parameter curves (F) using an ageing state model (9); - performing (S4) an optimization process for adjusting the plurality of load parameters in order to determine optimized load parameters by which the remaining lifetime (RUL) can be increased when these are taken into account in the future operation of the technical device (4); - Operating (S5) the device battery (41) depending on at least one of the optimized load parameters.

Description

Technical field

[0001] The invention relates to methods for operating device batteries for technical devices for which an anomaly or faulty behavior has been detected, in particular to adapting operating methods after an anomaly has been detected. The invention further relates to the selection of suitable measures for extending the service life of a battery and methods for increasing the service life of a device battery affected by an anomaly. Technical background

[0002] The power supply for off-grid electrical devices and machines, such as electrically powered vehicles, is usually provided by device batteries or vehicle batteries. These provide electrical energy to operate the devices.

[0003] Portable batteries degrade over their service life and depending on their load and use. This so-called aging leads to a continuously decreasing maximum power or storage capacity. The state of aging corresponds to a measure of the aging of energy storage devices. According to convention, a new portable battery can have an aging state (in terms of its capacity, SOH-C) of 100%, which decreases noticeably over the course of its service life. The measure of the aging of a portable battery (change in the state of aging over time) depends on the individual load on the portable battery, i.e. in the case of vehicle batteries, the driver's usage behavior, external ambient conditions and the vehicle battery type.

[0004] For safety reasons, the proper functioning of the device battery in battery-operated technical devices must be regularly monitored for faults, especially at high energy densities. If a battery cell, a unit of several battery cells, or the entire device battery fails, depending on the fault, the technical device may become limited or completely inoperable. In the case of malfunctions that lead to a significant increase in temperature, the safety of the technical device and the user may also be compromised.

[0005] Device batteries in off-grid electrical devices, especially mobile devices, are therefore continuously monitored to detect sudden battery failures (sudden death), thermal fault events, rapid cyclic aging, or other fault events (collectively referred to herein as anomalies) in a timely manner. Depending on the type of anomaly detected, different measures can be taken, ranging from a simple warning to the user to the immediate shutdown of the technical device.

[0006] Such anomaly detection involves the detection of current or future anomalies based on measured temporal patterns of operating variables of the device battery. To ensure largely independent anomaly detection, anomaly detection models are implemented in the technical devices that evaluate the operating variable patterns and appropriately signal anomalies in the event of deviations from conventional operating patterns.

[0007] The document DE 10 2019 208 372 A1 discloses a computer-implemented method for detecting an anomaly in a technical system, comprising the following steps: detecting an operating variable vector which indicates an operating state of the technical system and comprises a number of operating state variables, wherein the operating state variables comprise at least one environmental state variable which indicates an environmental condition in which the technical system is operated, and system state variables which indicate internal system states of the technical system;Providing an environmental state model and an anomaly detection model, wherein the environmental state model indicates, depending on at least one of the environmental state variables, a verifiability of the operating variable vector with regard to the presence of an anomaly using the anomaly detection model, and wherein the anomaly detection model indicates the presence of an expected anomaly depending on the operating variable vector, signaling a presence of an anomaly or a non-anomaly depending on an evaluation of the at least one environmental state variable of the operating variable vector using the environmental state model and depending on an evaluation of the operating variable vector depending on the anomaly detection model;

[0008] Furthermore, the paper K. Park, Y. Choi, WJ Choi, H. -Y. Ryu and H. Kim, "LSTM-Based Battery Remaining Useful Life Prediction With Multi-Channel Charging Profiles," in IEEE Access, vol. 8, pp. 20786-20798, 2020, doi: 10.1109 / ACCESS.2020.2968939 discloses the prediction of a remaining useful life of a battery using LSTM models for different charging profiles. Disclosure of the invention

[0009] According to the invention, a method for operating a device battery after detection of a current or predicted anomaly and for determining optimized measures for extending the service life of a device battery, in particular following detection of a current or dedicated anomaly, according to claim 1 and a corresponding device according to the independent claim are provided.

[0010] Further embodiments are specified in the dependent claims.

[0011] According to a first aspect, a computer-implemented method for operating a device battery in a technical device, in particular after detection of an anomaly in the device battery, is provided, comprising the following steps: - Providing temporal operating variable profiles of several operating variables of a specific device battery; - Predicting the temporal operating variable profiles (F) in order to obtain artificial operating variable profiles (F) depending on the provided temporal operating variable profiles (F) and a plurality of load parameters of a load pattern using a usage pattern model which is designed to provide artificial profiles of at least one operating variable depending on the plurality of load parameters, wherein the plurality of load parameters correspond to stress factors for the aging of the device battery; - Determination of the remaining service life of the device battery depending on the artificial operating parameters; - performing an optimization procedure to adjust the plurality of load parameters in order to determine optimized load parameters that can increase the remaining lifetime when taken into account in the future operation of the technical device; - Operating the device battery depending on at least one of the optimized load parameters.

[0012] Furthermore, the temporal operating parameter profiles can be used to check whether there is an anomaly in the device battery that leads to a remaining service life of the device battery that is shorter than the guaranteed service life of the device battery, whereby the optimization procedure is only carried out if an anomaly is detected.

[0013] To monitor device batteries from a large number of devices, operating parameter data can be continuously recorded and transmitted in blocks as operating parameter curves to a central unit external to the device. To evaluate the operating parameter data, particularly in physical or electrochemical battery models based on differential equations, the operating parameter data are sampled as curves with a comparatively high temporal resolution (sampling rates), for example, between 1 and 100 Hz, and a battery status is determined from this using a time integration method.

[0014] To evaluate the operating parameter data, in particular to determine internal battery states, which also determine the aging state, an electrochemical battery model (P2D battery model) can be used. This model is based on a system of differential equations with a plurality of nonlinear differential equations. The operating parameter data enables modeling of the current battery state using a time integration method. Such electrochemical battery models are known, for example, from the publications US 2020 / 150,185A, WO2022017984A1, EP3919925A1, DE102020206915B3, and US20210373082A1.

[0015] Providing the operating parameter curves in the central unit enables the use and adaptation of the electrochemical battery model for a wide variety of device batteries with similar battery cells or cells with similar cell chemistry. Calculating the battery states using the differential equation system is computationally intensive, so outsourcing this to the central unit can also reduce the computational load on the device's internal computing devices.

[0016] In addition, an aging state can also be determined based on the operating parameter curves. The aging state (SOH: State of Health) is the key parameter for specifying the remaining battery capacity or remaining battery charge for portable batteries. The aging state represents a measure of the aging or performance of the portable battery. In the case of a portable battery, a battery module, or a battery cell, the aging state can be specified as the capacity retention rate (SOH-C). The capacity retention rate SOH-C, i.e. the capacity-related aging state, is specified as the ratio of the measured instantaneous capacity to an initial capacity of the fully charged battery and decreases with increasing aging. Alternatively, the aging state can be specified as an increase in the internal resistance (SOH-R) relative to an internal resistance at the beginning of the service life of the portable battery.The relative change in internal resistance SOH-R increases with increasing battery age.

[0017] Furthermore, as mentioned above, device batteries are continuously monitored to detect sudden battery failures (sudden death), thermal fault events, rapid cyclic aging, or other fault events (collectively referred to herein as anomalies) in a timely manner. Depending on the type of anomaly detected, different measures can be taken, ranging from a simple warning to a user to the immediate shutdown of the technical device. For this purpose, anomaly detection models are used that detect current or future anomalies based on an evaluation of the temporal progression of the operating variables.

[0018] Anomaly detection for portable batteries is typically performed by evaluating operating parameter curves. For portable batteries, these operating parameter curves typically include the battery voltage or cell voltage, the battery current, the battery temperature, and the state of charge. These operating parameter curves can be analyzed, particularly after subsequent feature extraction, using rule-based or data-based anomaly detection models to identify portable batteries with an anomaly or faults.

[0019] Anomaly detection methods can be used to detect current or future anomalies. Such methods can be rule-based or based on data-driven anomaly detection models and can predict various failure types. In particular, anomalies are all deviations from the normal operation of a device battery that lead to a reduction in service life or imminent failure. This allows for the detection of above-average degradation and critical failures of device batteries, such as sudden death or thermal runaway, or other failure types.

[0020] However, with rule-based anomaly detection, faults in device batteries are currently only detected when applied fault thresholds for operating variables, such as cell voltage, a module temperature, a current value or a state of charge value and an aging state value, are exceeded or undershot.

[0021] Anomaly detection models can also be designed as data-based classification models, for example. However, the creation and training of data-based anomaly detection models must also usually be performed externally to the device due to the high computational effort and the merging of data from a large number of device battery applications. Data-based anomaly detection models are known, for example, from DE 10 2022 202 111 A1 and DE 10 2022 203 343 A1 in the form of autoencoder-based approaches. Alternative methods are known, for example, from DE 10 2022 202 112 A1.

[0022] Anomaly detection can also be provided based on data, e.g., using an autoencoder or a variational autoencoder. Such an autoencoder or variational autoencoder can, for example, be trained using operating characteristic points from a large number of similar device batteries that operate error-free or have operated error-free and without failures in a previous period (e.g., up to 3 or 6 months ago). In particular, it can be trained unsupervised and updated regularly or continuously. The application of the anomaly detection model can determine an anomaly in a device battery based on an evaluation of a deviation measure of a distribution in the latent state space of the autoencoder or the variational autoencoder or a reconstruction error.

[0023] The autoencoder-based approaches used for anomaly detection are capable of classifying a current battery state as normal or anomalous and, in addition, reliably predicting when a failure event is likely to occur in the future. Since safety-critical events, such as a thermal runaway event or sudden death, are preceded by changes in battery states, such critical events can be predicted. The type of likely failure can also be identified based on the criteria of rule-based anomaly detection or through differentiated anomaly classification in data-based anomaly detection.

[0024] By predicting anomalies early on, it is possible to attempt to extend the service life or remaining service life (remaining service life) of the device battery or to ensure compliance with the guaranteed service life by intervening in the operation of the device battery. The service life of a device battery is determined by a point in time at which a predetermined degradation (aging state) is reached. The guaranteed service life specifies the minimum period until which degradation does not exceed a predetermined level. For example, it is common practice for vehicle batteries to guarantee a service life of eight years, with a decrease in the capacity-related aging state of 20% (to 80%). In particular, in the case of anomalies that lead to faster aging of the device battery and thus do not achieve the guaranteed service life, an extension of the predicted service life can be achieved through appropriate operation.

[0025] The remaining service life can be determined using an aging model for the time-series-based modeling of an aging state of the device battery, whereby a temporal aging state curve is determined depending on the provided temporal operating variable curves and the artificial operating variable curves, whereby the remaining service life is determined by the aging state curve depending on a point in time at which a predetermined limit aging state is reached.

[0026] The objective of the above method is to determine one or more of the load parameters that, as stress factors, influence or have the greatest influence on battery degradation, and to modify these load parameters / stress factors during further operation in order to enable the best possible extension of the device's battery life. Such load parameters can include, for example, the battery temperature (maximum or average), a state-of-charge window defined by the minimum and maximum state-of-charge or a corresponding voltage window for an electrical terminal voltage, and a charging curve that indicates a maximum charging current depending on a state-of-charge.

[0027] To determine the service life achievable with a specific device battery, an aging state model based on time series analysis of operating variables can be used. This aging state model uses recorded operating variable curves of the device battery, which can include, for example, battery current, battery voltage, battery temperature, and state of charge, to model the aging state of the device battery. Using predefined load parameters, which can be derived from the historical operating variable curves, artificial operating variable curves can be generated. Based on the thus determined artificial operating variable curves, a prediction of the aging state can be made by applying the aging state model.

[0028] The predicted service life determined in this way results from the time at which an aging threshold specified for the guaranteed service life is reached through the predicted course of the aging state.

[0029] An excessively high (average) deterioration gradient results in the device battery falling below a warranty limit, which is defined by a total service life and a deterioration limit. If the deterioration limit is reached or undercut due to the deterioration of, for example, 80% or less after a shorter period than the specified (guaranteed) service life, the warranty requirement for the device battery is not met. Consequently, a very high future warranty risk may exist, even though the device battery currently still has a deterioration level above the deterioration limit.

[0030] For example, the current aging state can be determined using a physical aging model, which represents a form of the electrochemical battery model and can be parameterized accordingly. The electrochemical battery model is based on a system of differential equations with a plurality of nonlinear differential equations. The time series of the operating variable data enable the modeling of the current battery state using a time integration method. Such electrochemical battery models are known, for example, from the publications US 2016 / 023,566, US 2016 / 023,567, and US 2020 / 150,185.

[0031] The aging model can be characterized by its model parameters and model the internal battery state of the device battery. The aging model is based on electrochemical model equations parameterized by the model parameters, which characterize the electrochemical states of a nonlinear differential equation system and can be continuously evaluated according to a time integration method. An indication of an aging state can be derived from the parameterized model parameters in a conventional manner, e.g., by linear combination.

[0032] To improve the accuracy of the aging state model, it can be provided in the form of a hybrid aging state model, i.e., a combination of the physical aging model and a data-based correction model. In a hybrid aging state model, a physical aging state can be determined using the physical or electrochemical aging model, and a correction value resulting from the data-based correction model can be applied to this model, particularly by addition or multiplication.

[0033] Furthermore, the correction model of the hybrid data-based aging state model can be designed with a probabilistic or artificial intelligence-based regression model, in particular a Gaussian process model, and can be trained to correct the aging state obtained by the physical aging model. Therefore, a data-based correction model of the aging state exists for correcting the capacitance-related aging state SOH-C and / or another for correcting the resistance-change-related aging state SOH-R. Possible alternatives to the Gaussian process are alternative supervised learning methods, such as those based on a random forest model, an AdaBoost model, a support vector machine, or a Bayesian neural network.

[0034] The correction model can use operating characteristics as input variables, which are determined from the temporal operating variable profiles using feature extraction or feature extraction methods. The features or characteristics are calculated using signal processing operations. The operating characteristics assigned to an operating variable profile define an operating characteristic point for the respective device battery for a specific point in time, which represents the load condition due to cyclic operation (cyclic aging) and the calendar aging of the device battery (elapsed time since commissioning or the start of its service life).

[0035] The operating characteristics may include cumulative load-based characteristics or aggregated characteristics and / or statistical values ​​determined over the entire lifetime to date.

[0036] In particular, operating characteristics can be determined from histogram data created from the curves of the operating variables. For example, histograms can be created relating to the battery current versus the battery temperature and the state of charge of the vehicle battery, a histogram of the battery temperature versus the state of charge of the vehicle battery, a histogram of the charging current versus a battery temperature, and a histogram of a discharging current versus the battery temperature. Furthermore, operating characteristics can be the accumulated total charge (Ah), an average capacity increase during a charging process (particularly for charging processes where the charge increase is above a threshold proportion [e.g. 20% ΔSOC] of the total battery capacity), the charging capacity, and an extreme value (e.g.a local maximum) of the smoothed differential capacity during a measured charging process with a sufficiently large change in the state of charge (smoothed curve of dQ / dU: change in charge divided by change in battery voltage) or the accumulated mileage since the device battery was put into operation can be taken into account. Further operating characteristics can correspond to a local extreme value of the spectral kurtosis, evaluated on a charging process for a current or voltage signal, one or more coefficients of a wavelet transform and / or one or more coefficients of the Fourier transform, each evaluated for a charging process for a current or voltage signal, or a transformed spectral value assigned to a defined frequency band.

[0037] Operating characteristics can thus be derived from histograms of operating variables. Using feature engineering or feature extraction methods, operating characteristics can be extracted, such as the mean, standard deviations of the histograms, as well as multidimensional statistical values ​​such as the mean, median, minimum, maximum, moments of the distribution, and the like.

[0038] It may be provided that the artificial operating parameter curves are created depending on the multiple load parameters. For this purpose, a data-based usage pattern model may be provided that converts specified load patterns into artificial curves of load variables L, which may include some of the operating variables and reflect the load on the device battery to which the device battery is exposed during the usage and operating mode specified by the usage pattern.

[0039] The load patterns can characterize the load parameters of the vehicle battery's previous use and load, or they can be defined differently. The load patterns can be specified and determined in such a way that, when applied with the usage pattern model, they result in the output of a temporal profile of a battery current and a battery temperature as load variables. Using a suitable, known battery dynamics model, the battery voltage and state of charge (SOC) curves can be determined from the load variable curves in order to generate the artificially generated operating variable curves together with the load variables.

[0040] The load patterns can be defined by load parameters, which are individually derived for previous operation from historical operating parameters and / or operating characteristics. The load patterns can also be adjusted by varying one or more load parameters in order to implement a desired measure to extend the remaining service life of the device's battery.

[0041] The load patterns can be evaluated by the usage pattern model, preferably with the help of data-based methods, and serve to indicate the typical usage behavior of an individual user with regard to the vehicle battery in question.

[0042] If implemented data-based, the usage pattern model can be designed, for example, as a transformer, which can be based on multi-head attention, or alternatively as a recurrent neural network, such as an LSTM or GRU, in particular as a Bayesian LSTM network, and trained based on load variables or operating variables that indicate a typical and representative type of vehicle battery use. The load variables or operating variables to be considered should be based on a period of the same type of use and the same operating mode of the vehicle battery. Furthermore, the usage pattern model can be designed based on physical domain knowledge, as described in US20230306803A1, or as a hybrid model that includes physical and data-driven models.

[0043] The load parameters that specify the load pattern N then correspond to the input variables of the usage pattern model and characterize the driving behavior of a driver or types of use of the vehicle. The usage pattern model can be trained in such a way that, when the load pattern and a calendar time are specified, an operating variable curve is predicted which, when evaluated accordingly, results in load parameters that correspond to the specified load pattern, or from which the load parameters can be derived when evaluated accordingly.

[0044] For evaluation, the usage pattern model can receive a calendar time and the load parameters of the load pattern of the previous use or changed load parameters of a changed load pattern on the input side and output a corresponding artificial course of the load variables (current, temperature, preferably as a time series) on the output side.

[0045] The usage pattern model can be designed as an overall model or have a sub-model for each load variable to be predicted. One sub-model of the usage pattern model is trained on the temporal temperature curve, and another sub-model of the usage pattern model is trained on the current curve based on the specified load pattern. Thus, by specifying a calendar time reference, such as the date and time, and desired load parameters, an artificial curve of the load variables and / or operating variables can be generated. The calendar time reference can also include the day of the week, the month, and information about public holidays, and can take seasonality into account, particularly through feature engineering.

[0046] By predicting artificial operating variables based on load parameters of a given load pattern, the aging state model can be used to extrapolate the course of a predicted aging state into the future. This enables the determination of the remaining service life for a specific load pattern. The predicted operating variables can be modeled based on specified load parameters or stress factors, which include, for example, an upper and lower value for the state of charge, a limitation of the battery temperature or the battery temperature range, an average or maximum ampere-hour throughput, and the selection of a specific charging curve. Other possible load parameters can include aggregated or abstracted variables such as: usage categories, annual mileage, fast-charging frequencies, driving style (e.g., average acceleration), temperature range (e.g.,ambient temperature), state of charge (SOC) and distributions of temperature, state of charge, etc.

[0047] By varying the load parameters or stress factors, determining the corresponding artificial operating parameters, and predicting the aging state progression until a aging state limit defined by warranty conditions, e.g., 80% SoH, is reached, the remaining service life can be determined for each of the varied load patterns. The optimal load pattern can be determined by maximizing the remaining service life in an optimization process.

[0048] Optimization is used to determine the load parameters that most strongly influence the aging of the device battery identified as abnormal. If the load parameter(s) identified in this way are taken into account in future operation, it may be possible to meet the manufacturer's warranty conditions despite a detected device battery anomaly.

[0049] The optimization method can be implemented as a simple gradient descent method. Alternatively, thanks to the probabilistic modeling of the aging state model, the optimization method can also be implemented in a conventional manner using a Bayesian optimization method to optimize the load parameters and to predictively optimize at least the state of charge window, the charging curve, and a temperature setpoint in the operating strategy, i.e., a limit value for a battery temperature. Thus, an appropriate adjustment of the load parameters or stress parameters can be made for each device battery for which an anomaly has been detected.

[0050] From the determined load parameters, those can then be selected that have the greatest influence on slowing down the aging of the device battery.

[0051] According to one embodiment, for the optimization, one of the at least one load parameter can be excluded depending on a usage behavior of the device battery, wherein the usage behavior is determined depending on the provided operating variable profile.

[0052] If necessary, load parameters can be restricted using constraints to meet usage specifications or user preferences. For example, if a device battery is always operated in fast-charging mode, this is a desired mode that should not be restricted, and thus, a load parameter should not be allowed to restrict the charging curve toward a more gentle charging process. In this case, however, a lifetime improvement can be achieved as a load parameter through thermal management or the state of charge shift. The permissible variations of the load parameters can be learned regularly from data for each of the device batteries, for example, through quantile analysis.

[0053] By using a probabilistic, data-based model to predict the course of the aging state, in particular using a Gaussian process, the course of the predicted aging state can be provided with a confidence interval. Using an optimization method, corresponding courses of the predicted aging state can then be simulated by varying the load parameters. The goal of the optimization is to achieve a specified guaranteed service life of, for example, eight years with a certain probability, such as 95%. The exact probability of achieving the specified service life can be determined using the probability density function resulting from the probabilistic modeling of the aging state model.

[0054] Optimization allows individual load parameters to be identified that have the greatest possible impact on extending the remaining service life. In particular, once identified, one or more of the identified load parameters can be further optimized while maintaining the values ​​of the remaining load parameters in order to adapt the operating strategy according to the further optimized identified load parameters.

[0055] The optimized load parameters can then be transmitted back from the central unit to the relevant technical device to be considered in an operating strategy for the further operation of the device battery. To this end, at least one parameter in an operating strategy can be adjusted so that the value of the optimized load parameter(s) can be achieved as a target state. The operating strategy can also consider only the most varied load parameters or one or more of the load parameters that have the greatest influence on extending the remaining service life. Brief description of the drawings

[0056] Embodiments are explained in more detail below with reference to the attached drawings. They show: Fig. 1 a schematic representation of a system for providing driver and vehicle-specific operating variables for detecting a predicted anomaly and for determining load parameters for the optimized operation of a vehicle battery in a central unit; Fig. 2 a schematic representation of a hybrid aging state model for probabilistic modeling of an aging state progression based on load patterns; and Fig. 3 is a flowchart illustrating a method for determining optimized operating parameters to extend the service life of a vehicle battery; Fig. 4 schematically shows a course of a predicted aging state in the case of an anomaly-affected vehicle battery and a vehicle battery with adapted load parameters. Description of embodiments

[0057] In the following, the method according to the invention is described using vehicle batteries as device batteries in a large number of motor vehicles as similar devices. For this purpose, one or more electrochemical battery models are evaluated and parameterized in the central unit based on operating parameter profiles.

[0058] The above example is representative of a variety of stationary or mobile devices with off-grid energy supply, such as vehicles (electric vehicles, pedelecs, etc.), systems, machine tools, household appliances, IOT devices and the like, which are connected to a device-external central unit (cloud) via a corresponding communication connection (e.g. LAN, Internet).

[0059] Fig. 1 shows a system 1 for collecting fleet data in a central unit 2 for creating, operating and evaluating an electrochemical battery model for modeling internal battery states of the vehicle battery and an aging state model for determining a current and predicted aging state of a vehicle battery in a motor vehicle.

[0060] Fig. 1 shows a vehicle fleet 3 with several motor vehicles 4. One of the motor vehicles 4 is in Fig. 1. The motor vehicles 4 each have a vehicle battery 41 with battery cells 45, an electric drive motor 42, and a control unit 43. The control unit 43 is connected to a communication device 44, which is suitable for transmitting data between the respective motor vehicle 4 and a central unit 2 (a so-called cloud).

[0061] The control unit 43 is particularly designed to record operating variables of the vehicle battery 41 recorded with the aid of a battery management system 46 with a high temporal resolution, such as between 1 and 50 Hz, such as 10 Hz, and to transmit these to the central unit 2 via the communication device 44.

[0062] The motor vehicles 4 send the operating variables F to the central unit 2, which at least indicate variables that influence or are influenced by the aging state of the vehicle battery 41 and are required for determining the internal battery states, an aging state, and parameterizing an electrochemical battery model. In the case of a vehicle battery, the operating variables F can indicate an instantaneous battery current, an instantaneous battery voltage, an instantaneous battery temperature, and an instantaneous state of charge (SOC), both at the pack, module, and / or cell level. The operating variables F are recorded as operating variable curves in a fast time frame of 0.1 Hz to 50 Hz and can be regularly transmitted to the central unit 2 in uncompressed and / or compressed form.For example, the time series can be transmitted to the central unit 2 in blocks at intervals of 10 minutes up to several hours using compression algorithms to minimize data traffic to the central unit 2.

[0063] The central unit 2 comprises a data processing unit 21 in which part of the method described below can be carried out, and a database 22 for storing data points, model parameters, states and the like.

[0064] The central unit 2 is configured to receive the operating variable profiles F. The central unit 2 can determine a current aging state for the respective vehicle battery 41 from the operating variable profiles in a manner known per se, e.g., using an aging state model described below. Furthermore, the central unit 2 can derive load patterns from the operating variable profiles and generate artificial operating variable profiles that would lead to a real load (load intensity) on the vehicle battery that corresponds to the historical load on the vehicle battery.

[0065] Fig. Figure 2 schematically shows, by way of example, the functional structure of an embodiment of a data-based aging state model 9 designed in a hybrid manner. The aging state model 9 comprises a physical aging model 5 and a data-based correction model 6.

[0066] Physical aging model 5 is a mathematical model based on nonlinear differential equations. Evaluating physical aging model 5 of aging state model 9 with operating parameter profiles, particularly since the beginning of the vehicle battery's service life, results in an internal state of the system of physical differential equations that corresponds to the physical internal state of the vehicle battery. Since the physical aging model is based on physical and electrochemical laws, the model parameters of the physical aging model are quantities that specify physical properties.

[0067] The time series of the operating variables x(t) of the vehicle battery 41 to be evaluated are therefore directly incorporated into the physical aging state model 5, which is preferably designed as an electrochemical model and models corresponding internal electrochemical states z(t), such as layer thicknesses (e.g. SEI thickness), changes in cyclable lithium due to anode / cathode side reactions, rapid consumption of electrolytes, slow consumption of electrolytes, loss of active material in the anode, loss of active material in the cathode, etc., using nonlinear differential equations in a multidimensional state vector.

[0068] The physical aging model 5 thus corresponds to an electrochemical model of the respective vehicle battery 3. This model determines internal physical battery states z(t) depending on the operating variable curves x(t) in order to obtain a physically based aging state SOHph = Phys[x(t)] of dimension of at least one, depending on the above-mentioned electrochemical states z(t), which are mapped linearly or non-linearly to a capacity maintenance rate (SOH-C) and / or an internal resistance increase rate (SOH-R) in order to provide these as aging states (SOH-C and SOH-R).

[0069] However, the model values ​​for the physical aging state SOHph provided by the electrochemical model are inaccurate in certain situations, and therefore, they are intended to be corrected with a correction variable k. The correction variable k is provided by the data-based correction model 6, which is trained using training data sets obtained by test bench 1.

[0070] Preferably, the data-based correction model corresponds to a Gaussian process model GP=GP[m(x(t)), z(x(t)), Phys[x(t)]))~N(μ, Σ), where µ is an m × 1-dimensional vector that specifies the predictive mean at an evaluation time t, and Σ is the m × m-dimensional predictive covariance matrix of the Gaussian process model. The formulas for the mean and variance of the Gaussian process are as follows: μ(x¯)=kTCN−1y ∑(x¯)=c−kTCN−1k where N corresponds to the number of labels, x is m x d dimensional and specifies a set of m new points in the input space, where x = [m(x(t)), z(x(t)), Phys[x(t)]]) and y each correspond to an N x 1 -dimensional vector, and where y each specifies a measured or labeled aging state. k corresponds to an N x m dimensional matrix of kernel evaluations specifying the correlations encoded in the kernel of the Gaussian process between the m new points and the N measured points. C corresponds to an N x N dimensional matrix of kernel evaluations between the N measured points and c to an m x m matrix with kernel evaluations between the m new points. T denotes the transpose. See also Bishop, “Pattern Recognition and Machine Learning”, 2006.

[0071] The correction model 6 receives operating characteristics m(x(t)) on the input side, which are derived from the predicted or given courses of the operating variables x(t 0 ...t n) for future evaluation times and can also include one or more of the internal electrochemical states of the differential equation system of the physical model 5. Furthermore, the correction model 6 can receive the physical aging state SOHph obtained from the physical aging model 5 on the input side. The operating characteristics m(t) of the current evaluation time t are generated in a feature extraction block 8 based on the operating variable profiles x(t0...t). Furthermore, the internal states z(x(t)) from the state vector of the electrochemical physical aging model 5 and advantageously the physical aging state SOHph are fed to the correction model 6. The feature vector m(x(t)) is robust because it is independent of the model quality of the physical aging model or its state.Therefore, the consideration of the feature vector m(x(t)) is a valuable addition to the internal states z(x(t)) of the physical or electrochemical aging model.

[0072] From the predicted or given company size curve x(t 0 ...t n ), operating characteristics m(x(t)) relating to an evaluation period can be generated in the central unit 2 for each vehicle fleet 3 or, in other embodiments, already in the respective motor vehicles. The evaluation period between two evaluation times can range from a few hours (e.g., 6 hours) to several weeks (e.g., one month) for determining the aging state. A typical value for the evaluation period is one week.

[0073] The operating characteristics m(x(t)) can, for example, include characteristics related to the evaluation period and / or accumulated characteristics and / or statistical quantities determined over the entire service life to date. In particular, the operating characteristics can include, for example: electrochemical states, such as SEI layer thickness, changes in cyclable lithium due to anode / cathode side reactions, rapid uptake of electrolyte solvents, slow uptake of electrolyte solvents, lithium deposition, loss of active anode material and loss of active cathode material, information on impedances orthe internal resistances, histogram features such as temperature versus state of charge, charging current versus temperature and discharging current versus temperature, in particular multi-dimensional histogram data relating to the battery temperature distribution versus state of charge, the charging current distribution versus temperature and / or the discharging current distribution versus temperature, the current throughput in ampere hours, the accumulated total charge (Ah), an average capacity increase during a charging process (in particular for charging processes where the charge increase is above a threshold proportion [e.g. 20% ΔSOC] of the total battery capacity), the charging capacity and an extreme value (e.g. maximum) of the differential capacity during a measured charging process with a sufficiently large swing in the state of charge (smoothed curve of dQ / dU: charge change divided by change in battery voltage) or the accumulated mileage.These variables are preferably converted in such a way that they best characterize the actual usage behavior and are normalized in the feature space. The operating characteristics m(t) can be used in whole or in part for the procedure described below.

[0074] To determine a corrected aging state SOH to be output, the outputs SOHph, k of the physical aging model 5 and the data-based correction model 6, which is preferably implemented as a Gaussian process model, are combined. In particular, these can be added or otherwise multiplied (not shown) in a summing block 7 to obtain the modeled aging state SOH to be output for a current evaluation period. In the addition case, the confidence of the Gaussian process can also be used as the confidence of the corrected aging value SOH of the hybrid model to be output. The confidence or confidence value of the Gaussian process model thus characterizes the modeling uncertainty of the mapping of operating characteristic points to an aging state.

[0075] The initial training of the hybrid aging state model 9 takes place in test bench 1. For this purpose, training data sets are created that assign the operating parameters of a vehicle battery operating in accordance with a load profile to an empirically or model-based aging state as a label. The goal is for the hybrid aging state model to be able to accurately predict the non-linear cyclical decrease in capacity. This can be achieved through a combination of electrochemically modeled causal chains and machine learning.

[0076] For example, a basis model can be provided as a label for training the hybrid or data-based state-of-age model to determine the state of aging. According to this model, a SOH-C measurement is performed by Coulomb counting or by forming a time integral of current during the charging process under defined environmental conditions, which is divided by the change in the charge state between the start and end of the relevant charging and / or discharging phase. A sufficiently reliable indication of the state of aging can be obtained, for example, if the vehicle battery is brought from a defined relaxed state under reproducible load and environmental conditions from a fully discharged charge state to a fully charged state during a charging process. The maximum charge thus detected can be related to an initial maximum charging capacity of the vehicle battery.Resistance-related aging states (SOH-R values) can also be calculated using voltage changes relative to a current change. These are usually related to a defined time interval, defined environmental conditions, and the energy flow direction of the system.

[0077] The determination of an aging state as a label can be carried out in a conventional manner by evaluating the operating parameter curves with an additional aging model under defined load and environmental conditions for label generation, such as constant temperature, constant current, and the like. Other models can be used to determine the aging state. The training of the data-based correction model can be carried out in the conventional manner based on the training data sets and the residuals of the modeled aging state.

[0078] One goal of the method described below is, after detecting an anomaly in a vehicle battery that leads to increased aging or degradation of the vehicle battery, to adjust load parameters that characterize or determine the further operation of the vehicle battery in order to reduce the further aging of the vehicle battery. This may potentially reduce the degradation of the vehicle battery 41 affected by an anomaly to such an extent that the specified warranty conditions (defined by an aging threshold and a service life) can be met despite the anomaly.

[0079] In order to determine the artificial operating parameter curves for predicting the aging state, load variables L can first be generated. For this purpose, a data-based usage pattern model 10 can be provided, for example. The usage pattern model 10 converts predetermined load patterns N into curves of load variables L that reflect the load on the device battery to which the device battery is exposed during the usage and operating mode specified by the usage pattern. The load patterns N characterize the load parameters of the previous use and load of the vehicle battery 41. The load patterns N can be predetermined and determined in such a way that, when used with the usage pattern model 10, they lead to the output of a temporal profile of a battery current I and a battery temperature T as load variables L.Using a suitable, known battery dynamics model 11, the battery voltage U and state of charge SOC curves can be determined from the load variables L. Together with the load variables, the artificially generated operating variables can be generated. The usage pattern model 10 and the battery dynamics model 11 can be operated depending on an aging condition.

[0080] The load patterns N can be defined by the load parameters, which are individually derived for the previous operation from the historical operating variables and / or the operating characteristics. The load patterns can also be adjusted by varying one or more load parameters in order to implement a desired measure to extend the remaining service life of the vehicle battery 41.

[0081] The load patterns can be evaluated by the usage pattern model, preferably with the help of data-based methods, and serve to indicate the typical usage behavior of an individual user with regard to the vehicle battery in question.

[0082] The usage pattern model 10 can be implemented as a recurrent neural network, such as an LSTM or GRU, in particular as a Bayesian LSTM network, and trained based on the profiles of load variables L or operating variables F that indicate a typical and representative type of use of the vehicle battery. The profiles of load variables L or operating variables F to be considered should be based on a period of the same type of use and the same operating mode of the vehicle battery 41.

[0083] The load parameters that specify the load pattern N then correspond to the input variables of the usage pattern model 10 and characterize the driving behavior of a driver or the types of use of the vehicle. The usage pattern model 10 is trained in such a way that, given the load pattern N and a calendar time specification, an operating parameter curve is predicted which, upon appropriate evaluation, yields load parameters that correspond to the specified load pattern N, or from which the load parameters can be derived upon appropriate evaluation.

[0084] For evaluation, the usage pattern model 10 can be supplied with a calendar time and the load parameters of the load pattern of the previous use or changed load parameters of a changed load pattern on the input side and output an artificial course of the load variables (current, temperature, preferably as a time series) on the output side.

[0085] The usage pattern model 10 can be designed as an overall model or have a sub-model for each load variable L to be predicted. One sub-model of the usage pattern model 10 is trained on the driver-specific temporal temperature profile, and another sub-model of the usage pattern model 10 is trained on the driver-specific current profile. Thus, by specifying a calendar time reference, such as the date and time, and desired load parameters, an artificial profile of the load variables and / or operating variables can be generated. The calendar time reference can also include the day of the week, the month, and information about public holidays, and can take seasonality into account, particularly through feature engineering.

[0086] The usage pattern model 10 can be trained regularly, taking new data into account, as long as no driver change has been detected, e.g., once a month. In this case, the training data set can typically include trends in operating variables or load variables from the last identified driver change up to the most recently recorded operating variables or load variables. Typically, in the middle of this section, a time period is extracted from the training data set that is not used for training, but rather for validating the usage pattern model. Typically, a Bayesian LSTM network is used for the driver-specific characterization of the current curve. Furthermore, a Bayesian LSTM network is used for the driver-specific characterization of the temperature curve. Alternatives to the LSTM approach are ARIMAX models and transformer models based on multi-head attention.

[0087] The usage pattern model 10 can thus be developed directly from raw data of the load variables L and / or the operating variables F. For the vehicle battery 41, typical patterns of the current profile, e.g., due to recurring commuter routes and typical idle and rest times and loads at temperature ranges, are thus recognized and made reproducible.

[0088] The load patterns can thus indicate types of load on the vehicle battery 41 based on the load parameters, in particular periodic loads.

[0089] The load patterns N are specified for each individual vehicle and characterize the type of use and operation or the use and operating behavior of the respective vehicle batteries 41 based on the load parameters.

[0090] The use of the usage pattern model 10 represents only one possibility for generating artificial operating variables from the load parameters of a previous load on the vehicle battery 41. In the simplest case, temporal segments of operating variables from the past can be pieced together to generate artificial operating variables for prediction. For example, an artificial operating variable profile with a comparable load on the vehicle battery 41 as in the past can be generated by copying, for each weekday, an operating variable profile from a previous corresponding (same) weekday that comes closest to the desired load parameters for this weekday with regard to the load parameters under consideration.Stress factors are taken into account, which may include, for example, an average or maximum ampere-hour throughput, an upper and lower limit for the state of charge (charging is limited by the upper state of charge and discharging by the lower state of charge), a maximum and / or average battery temperature, a maximum discharge current, and a selection of a specific charging curve with a maximum charging current.

[0091] The method for determining measures to extend the service life of an anomalous vehicle battery 41 is carried out in the data processing device of the central unit 2 and is subsequently described using the flow chart of the Fig. 3 described in more detail.

[0092] In step S1, operating variables are first continuously transmitted from each of the vehicles 4 to the central unit 2.

[0093] The operating parameter profiles can be evaluated in a step S2 to determine an anomaly in the vehicle battery. For this purpose, known methods can be used, which are disclosed, for example, in DE 10 2022 202 111 A1 and DE 10 2022 203 343 A1 in the form of autoencoder-based approaches. Alternative methods are known, for example, from DE 10 2022 202 112 A1. Further methods include PCA or variational autoencoder approaches as well as normalizing flows. Reference is made to the prior art of the documents US20220101666A1, DE102021211158A1, DE102022203343A1, DE102021205879A1, US20230280406A1, and DE102020214234A1.

[0094] In step S3, a check is made to determine whether an anomaly exists in the vehicle battery 41. If an anomaly is detected (alternative: yes), the method continues with step S4; otherwise, the method returns to step S1.

[0095] In step S4, ranges for load parameters or stress factors are defined that can form the basis for future operation of the vehicle battery 41 affected by the anomaly. The load parameters can be used to specify an upper and lower limit for a state of charge or an upper and lower limit for a battery voltage at idle, a derating temperature (maximum temperature) for thermal management, and several charging curves (each with a specified maximum charging current) for carrying out charging processes and a maximum discharging current. Since an anomaly has been detected, the goal is to mitigate one or more of these load parameters / stress factors, i.e., to change the value of the load parameter accordingly, so that the further operation of the vehicle battery is subject to restrictions imposed by the load parameters.

[0096] Using a suitable optimization method, artificial operating variables are predicted based on various load patterns that take into account the corresponding combinations of load parameters. For example, charging and discharging take place within the limits of the state of charge, so that the artificial operating variables are designed accordingly. Vehicle battery operation is also designed so that the battery temperature remains below the temperature value specified by the derating temperature. Furthermore, a charging curve can be selected with which a charging process is carried out. The charging curve usually specifies a maximum charging current depending on the respective state of charge. For the driving cycles, the maximum discharge current can be set according to the relevant load parameter.Other load parameters that influence the degradation of the vehicle battery can also be taken into account and optimized in the optimization process.

[0097] In step S4, the optimization process is executed, in which the load patterns are varied in a simulated manner and artificial operating parameter curves for future vehicle operation are determined. The remaining service life is optimized, which results from the predicted aging state curve. The optimization process iteratively solves the optimization problem for variations of the individual load parameters to maximize the service life, i.e., the time until the specified limit aging state is reached. The optimization process can be carried out based on a gradient descent method.

[0098] The goal of the optimization is to extend the service life, i.e., extend the remaining service life of the anomalous vehicle battery, by adjusting the load parameters. To this end, artificial operating parameter profiles are iteratively generated, and an aging state profile is predicted from them, e.g., using the method described above. The aging state profile is analyzed to determine the point in time at which the limit aging state is reached. This point in time signals the end of the guaranteed service life and serves as the key point in time for calculating the remaining service life.

[0099] In Fig.Figure 4 schematically shows a progression of a predicted aging state for a vehicle battery 41 with an anomaly (dash-dotted curve) and a vehicle battery 41 with adjusted load parameters (dotted curve) as the aging state SOH over the calendar service life t (in years). The extension of the service life when adjusting the load parameters is evident. When using the hybrid aging state model 9, the aging state progressions can be specified with confidence intervals K.

[0100] The optimization procedure varies the load parameters, e.g., according to a gradient descent method, in order to obtain the optimal load pattern with optimized load parameters for lifetime extension.

[0101] Bayesian optimization, for example, can be used as an optimization method. This involves creating a probabilistic model for determining the remaining service life depending on the load parameters for various combinations of the load parameters listed above. The evaluation of the model then yields optimized load parameters for the continued operation of the vehicle battery.

[0102] Optimization allows individual load parameters to be identified that have the greatest possible impact on extending the remaining service life. In particular, once identified, one or more of the identified load parameters can be further optimized while maintaining the values ​​of the remaining load parameters in order to adapt the operating strategy according to the further optimized identified load parameters.

[0103] The determined optimized load parameters can now be transmitted to the vehicle in question in step S5 in order to limit the operation of the vehicle battery 41 during further use of the vehicle battery 41 according to an adapted operating strategy.

[0104] During the optimization process, constraints can be taken into account. In particular, specifications for certain load parameters can be taken into account, such as the requirement that the vehicle battery 41 must always be charged using a rapid charging process. In particular, if, as described above, the aging state model 9 is designed probabilistically, a corresponding confidence interval can be taken into account when the limit aging state is reached. Therefore, the remaining service life can be specified by reaching the limit aging state with a specific predicted probability.

[0105] The above procedure can be carried out repeatedly at regular or predetermined times to check whether the limitation by changing the load parameters has resulted in a correspondingly reduced aging.

[0106] If necessary, the load parameters can be further adjusted by re-executing the optimization procedure, for example, to attempt to comply with the warranty limit if the current aging process deviates from a previously predicted aging process. This can also be done if it is determined that aging is progressing significantly slower than predicted and, as a result, the predicted remaining service life of the vehicle battery 41 is significantly above the warranty limit.

[0107] According to a further embodiment, the at least one load parameter used for optimization can also be determined based on historical usage of the vehicle battery. For example, if the driver travels long distances with multiple charging stops, they rely on the highest possible state of charge variation, so the load parameters that specify the lower and upper state of charge limits should be changed as little as possible or not at all. In other cases, where the user frequently charges their vehicle using a fast-charging process, a variation in the charging profile can be disregarded during optimization, as this is unlikely to result in sufficient user acceptance if the charging profile is restricted.

[0108] Those vehicle batteries 41 that apply modified load parameters or operating strategies according to the above method are subsequently monitored, with the actual aging state profile (evaluated using one of the calculation models described above) being compared with the previously simulated predicted aging state profile. If the actual aging state profile deviates from the predicted simulated aging state profile, the optimization process described above can be performed again. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] DE 10 2019 208 372 A1

[0007] US 2020 / 150,185 [0014, 0030] WO 2022017984A1

[0014] EP 3919925A1

[0014] DE 102020206915B3

[0014] US 20210373082A1

[0014] DE 10 2022 202 111 A1 [0021, 0093] DE 10 2022 203 343 A1 [0021, 0093] DE 10 2022 202 112 A1 [0021, 0093] US 2016 / 023,566

[0030] US 2016 / 023,567

[0030] US 20230306803A1

[0042] US 20220101666A1

[0093] DE 102021211158A1

[0093] DE 102021205879A1

[0093] US 20230280406A1

[0093] DE 102020214234A1

[0093] Zitierte Nicht-Patentliteratur

[0000] K. Park, Y. Choi, W. J. Choi, H. -Y. Ryu and H. Kim, „LSTM-Based Battery Remaining Useful Life Prediction With Multi-Channel Charging Profiles,“ in IEEE Access, vol. 8, pp. 20786-20798, 2020, doi: 10.1109 / ACCESS.2020.2968939

[0008] Bishop, „Pattern Recognition and Machine Learning“, 2006

[0070]

Claims

[1] Method for operating a device battery (41) in a technical device (4), in particular after detection of an anomaly (S3) of the device battery (41), comprising the following steps: - providing (S1) temporal operating variable profiles (F) of several operating variables of a specific device battery (41); - predicting the temporal operating variable profiles (F) in order to obtain artificial operating variable profiles (F) depending on the provided temporal operating variable profiles (F) and a plurality of load parameters of a load pattern using a usage pattern model (10) which is designed to provide artificial profiles of at least one operating variable depending on the plurality of load parameters, wherein the plurality of load parameters correspond to stress factors for the aging of the device battery (41); - determining the remaining service life (RUL) of the device battery (41) depending on the artificial operating parameter curves (F) using an ageing state model (9); - performing (S4) an optimization method for adapting or determining the plurality of load parameters of the load pattern in order to determine optimized load parameters by which the remaining lifetime (RUL) can be increased or maximized when these are taken into account in an operating strategy in the future operation of the technical device (4); - Operating (S5) the device battery (41) depending on at least one of the optimized load parameters. [2] Method according to claim 1, wherein the temporal operating variable profiles are used to check whether there is an anomaly in the device battery (41) which leads to a remaining service life of the device battery (41) which is shorter than the guaranteed service life of the device battery (41), wherein the optimization method is only carried out if an anomaly is detected. [3] Method according to claim 1 or 2, wherein the remaining service life is determined using an aging state model (9) for time-series-based modeling of an aging state of the device battery (41), wherein a temporal aging state profile is determined depending on the provided temporal operating variable profiles and the artificial operating variable profiles, wherein the remaining service life (RUL) is determined by the aging state profile depending on a time at which a predetermined limit aging state is reached. [4] Method according to one of claims 1 to 3, wherein the plurality of load parameters comprises at least one of the following limitations: an upper limit for the state of charge, a lower limit for the state of charge, a limitation of the maximum battery temperature, a limitation of the average battery temperature, a maximum discharge current, an average discharge current during a driving cycle, a selection of a specific charging curve with a maximum charging current. [5] Method according to one of claims 1 to 4, wherein the artificial operating variable profiles are created as a function of the load pattern (N) using a data-based usage pattern model (10), wherein the usage pattern model (10) is trained such that, when the plurality of load parameters of the load pattern (N) and in particular a calendar time specification are specified, one or more profiles of the operating variables are predicted which, when evaluated accordingly, result in the load parameters which correspond to the specified load pattern (N). [6] Method according to one of claims 1 to 5, wherein for the optimization one of the at least one load parameter is excluded depending on a usage behavior of the device battery (41), wherein the usage behavior is determined depending on the provided operating variable profile. [7] Method according to one of claims 1 to 5, wherein the aging state model (9) is data-based using a probabilistic model, wherein the optimization method is a Bayesian optimization method. [8] Method according to one of claims 1 to 7, wherein the aging state model (9) is data-based with the aid of a probabilistic model, wherein the determination of the remaining service life (RUL) is carried out based on a point in time at which the reaching of a predetermined aging state limit is predicted with a predetermined probability, in particular between 90 and 99.999%, which can result from the integration over a probability density which is taken from a probabilistic prediction model. [9] A method according to any one of claims 1 to 8, wherein the method is carried out regularly or continuously. [10] Method according to one of claims 1 to 9, wherein the aging state model (9) and the usage pattern model (10) are regularly retrained. [11] Apparatus for carrying out one of the methods according to one of claims 1 to 10. [12] Computer program product comprising instructions which, when the program is executed by at least one data processing device, cause the device to carry out the steps of the method according to one of claims 1 to 10. [13] Machine-readable storage medium comprising instructions which, when executed by at least one data processing device, cause the device to carry out the steps of the method according to one of claims 1 to 10.

Images