Computer-implemented method for controlling system based on detected trend

Abstract

A computer-implemented method for controlling a system based on a detected trend. The invention relates to a computer-implemented method for controlling a system on the basis of trends detected in a time series of data detected by a sensor, wherein the method comprises the following steps: detecting and providing a time series of data by means of the sensor (S10); applying at least one linear filtering function with a resolution L to each point in time t of the time series of data (S12), and obtaining, for each resolution L, a characteristic value for quantifying the temporal progression of the time series of data (S14); detecting a trend in the time series of data when one of the characteristic values of the time series of data reaches a threshold value (S16); and triggering a response of the system when a trend in the time series of data has been detected (S18).

Description

Technical Field

[0001] The present invention relates to the field of automatic trend identification in data and to control systems based on the identified trends. Background Art

[0002] The identification of trends in the workflow of a system can be crucial, for example to identify wear, abnormal behavior or maintenance needs of the system. The earlier a trend, especially one that is harmful to the system, is identified, the better it can be counteracted or measures can be taken in order to eliminate it. It is all the more important to automatically read out such trends from the data available at an early stage.

[0003] To this end, various methods exist to automatically identify trends in time series of data.

[0004] Estimation for linear trends is known from Bianchi et al., A comparison of methods for trend estimation, Applied Economics Letters 6.2 (1999): 103-109. The method proposed therein makes use of statistics in order to decompose the data time series into trend, seasonality and noise. In this mode of operation, there are two types of trend identification, namely model-based trend identification and non-parametric trend identification. In the case of non-parametric trend identification, for example, a convolution is applied between a triangular filter and a finite symmetric window of the data time series.

[0005] Wavelet analysis, which can be compared with Fourier analysis, is known from Percival et al., "Wavelet methods for time series analysis", Vol. 4., Cambridge university press, 2000. However, for this purpose, wavelets are used instead of sine and cosine functions, which represent local oscillations around 0. Signal and wavelet are then convolved with each other at different resolutions.

[0006] Prewitt, Judith MS., "Object enhancement and extraction.", Pictureprocessing and Psychopictorics 10.1, 1970: 15-19 discloses filters for finding edges in image processing. The so-called Prewitt operator consists of two 3×3 kernels, which are convolved with a grayscale image.

[0007] Other approaches use deep learning. However, deep learning has the following disadvantages: It requires relatively more resources for its application. The computational operations are complex, which not only increases the current consumption of the execution system, but also takes up computing power that could be used in other aspects or requires a generally larger system.

[0008] The object of the present invention is therefore to provide a resource-saving method for trend detection and for the control of a system associated therewith.

[0009] This object is achieved by the subject matter of the independent claims. Summary of the invention

[0010] According to a first aspect of the invention, the task is accomplished by a computer-implemented method for controlling a system based on trends detected in a time series of data detected by a sensor.

[0011] The method comprises the following steps:

[0012] - Detect through sensors and provide data time series;

[0013] - applying at least one linear filter function with a resolution L to each time point of the data time series and obtaining for each resolution L a characteristic value for quantifying the time evolution of the data time series;

[0014] - Detecting a trend in a data time series when one of the characteristic values ​​of the data time series reaches a threshold;

[0015] -Triggering a system response when a trend has been detected in a data time series.

[0016] The detected data time series is in particular a time series of one-dimensional data, such as a temperature profile, a level profile, a pressure profile or the like. In other words, the data time series includes the values ​​of the measurable variable detected within a specific time period.

[0017] The variable is detected by a sensor, in the working process of which a trend should be identified. Thus, the sensor can be a commercially common sensor for detecting a measured variable, such as a thermometer, a pressure sensor, a level gauge or even an electronic measuring device, such as an ammeter, a voltage measuring device or a resistance measuring device, but not limited thereto. The sensor can be part of the system or monitor a parameter external to the system.

[0018] A linear filter function is applied to the data time series for each time point. Applying the filter function to the data time series in particular comprises a mathematical operation on the data time series. A new function is generated by applying the filter which can be analyzed. Such an analysis enables the application of the filter to the data time series to be quantified. In particular, the average slope of the new function can provide a hint of the presence of a trend. Thus, a separate characteristic value is determined from the new function for each resolution, which characteristic value is suitable for quantifying the time development of the data time series.

[0019] The characteristic value can then preferably be compared with a threshold value. The threshold value can preferably be selected such that small trends in the data time series are not recognized as trends in the sense of the invention. This can prevent, for example, small trends that are within the tolerance range of the system from triggering the set action. In other words, the threshold value can be selected such that common measurement errors of sensors, play of possibly moving parts in the system or general tolerances of the system are taken into account.

[0020] If the characteristic value exceeds the threshold, this will undoubtedly lead to the system's response being triggered. Since the threshold already takes into account the error range in the detection of the data time series and the regular operation of the system, there is an actual trend in the data time series that requires a response from the system. For example, the temperature may rise or fall slowly beyond a certain fluctuation, so that the system can, for example, respond accordingly with more or less input power.

[0021] In particular, fixed threshold values ​​can be used if the error range is known based on knowledge of the physical system. If the values ​​are in an abnormal range outside the usual or expected operating range, the threshold values ​​can be determined statistically from past observations if necessary. Alternatively, the threshold values ​​can also be calibrated on a validation data set so that known anomalies are reliably detected with the lowest possible false positive rate.

[0022] The response of the system can depend on the system itself and / or the monitored data time series. Trends do not always have to have a negative effect on the system. Positive effects are also possible. If, for example, a new machine is put into operation and the data time series detects vibrations of the machine, the vibrations may decrease over time until the machine has been run-in and, for example, the lubricant in its components has been sufficiently distributed for normal operation.

[0023] Since the application of a linear filter function to a data time series requires only very low demands on the resources of the system, the method according to the invention is suitable in a system-dependent manner for identifying trends in the data time series which trigger a response of the system.

[0024] In one embodiment, the trend to be identified is an upward trend or a downward trend.

[0025] Depending on whether an upward trend or a downward trend is to be detected, the filter can be implemented with different symbols, thereby making the implementation of the corresponding filter particularly simple.

[0026] The detection of trends is related to direction. The direction of the expected trend in turn depends on the monitored system and the measured parameters. If, for example, the state of a battery cell is monitored, in principle a number of parameters can be suitable for quantifying this state. Temperature or the output power are such parameters. For example, the temperature at the time of power output can increase with aging. That is, in the operation of the battery, an increasing trend in temperature becomes apparent over time. Conversely, the power of the battery may decrease with developing aging. If the power is observed over a long period of time, a decreasing trend in the power of the battery can be observed. However, in this special case, the charging cycles of the battery should be taken into account, since a fully charged battery generally has a higher power output than a half-empty battery.

[0027] To accomplish this, a quantization of the data time series is compared to a threshold value such that fluctuations on smaller time scales are less taken into account.

[0028] In one embodiment, applying the filter function for the resolution L comprises the following steps:

[0029] - specify the resolution L;

[0030] - Create a linear function phi with L values L (t);

[0031] - determining a segment of the data time series for a region tL to t having L values;

[0032] - The linear function phi L (t) multiplied by the segment of the data time series; and

[0033] - for all L values ​​of the linear function phi L All products of (t) with the data time series are summed.

[0034] The data time series is decomposed into segments for multiplication with the filter function, each of which includes L segments. Therefore, each segment f(t) has L values ​​from f(t-L+0) to f(t-L+L)=f(t).

[0035] The resolution L determines how many values ​​of the data time series are checked for trends. Depending on the system and the monitored process, the resolution L can be selected differently. The sensors used and the data time series checked can also influence the resolution. If, for example, the measured values ​​are detected once a day and the data time series consists of values ​​over a period of weeks, months or even years, a different, especially lower, resolution may be suitable. If, on the other hand, vibrations in the motor of the system are monitored at the millisecond or even microsecond level, the checked section of the data time series may include a multiple of this number of values. Accordingly, the resolution should be higher so that more values ​​are taken into account when searching for trends.

[0036] The linear filter is designed in particular such that, depending on whether the trend to be detected is upward or downward, it increases or decreases in L steps from -1 to +1. L (t) is multiplied by the segments of the data time series f(t), and the products thus formed are summed:

[0037]

[0038] Where 1 is from 0 to Lo

[0039] The characteristic value can be detected in particular as a numerical value in order to also identify a downward trend. In this case, the characteristic value would be K=|k|. On the other hand, a negative product k can also indicate a negative trend. The characteristic value is thus evaluated not only according to the numerical value but also according to the sign. If K does not exceed or exceeds the threshold value S, a trend is detected.

[0040] In one embodiment, each product phi L (1)*f(1) are weighted differently in the sum. For this purpose, the characteristic value K can be calculated, for example, using a modified formula:

[0041] K=|k|=∑ l w(l)*phi L (l)*f(l).

[0042] The weight factor w(1) may be weighted more strongly, in particular for the edge regions of the considered data time series. For example, but not limited to, it would then be:

[0043]

[0044] Weighting certain areas of the data time series for detecting trends advantageously results in that linear trends can be detected more easily. In particular, if edge areas of the examined data time series are monitored, finding trends can be facilitated because these areas will be particularly far apart from each other when trends actually exist.

[0045] In one embodiment, the system is a test bench for motor vehicles and the system's response to the detection of a trend includes shutting down the test bench. The data time series can be, in particular, parameters to be checked of the vehicle tested on the test bench. For example, exhaust gas values ​​or the temperature of the exhaust gas and the temperature or power of a battery pack of an electric vehicle belong to this, without being limited thereto.

[0046] If a trend is detected, this can indicate a malfunction or maintenance requirement of the vehicle. The test bench itself is then no longer needed, since the question of maintenance requirements has been clarified. Switching off the system thus saves energy and resources, since another vehicle can be brought to the test bench if necessary, while the vehicle being tested is now placed in the workshop. In addition, the vehicle being tested can also be switched off, so that less exhaust gases are generated or the battery pack is unnecessarily loaded.

[0047] In another embodiment, the sensor is a sensor for measuring temperature. In this embodiment, the data time series may be the temperature of a subsystem of the motor vehicle, in particular a battery pack, a fuel cell or a drive unit.

[0048] Too high a temperature in a subsystem of a vehicle usually means impending damage to the vehicle. If the battery gets too hot, the individual battery cells may age more quickly. If the exhaust gas is too hot, this indicates a faulty combustion of the propulsion fuel.

[0049] In another embodiment, the system is a vehicle and the sensor detects a parameter of the vehicle or a subsystem thereof.

[0050] Advantageously, the proposed method is particularly suitable for execution by an onboard computer of a vehicle, since onboard computers are often equipped with very limited resources. Therefore, the resource-conscious detection of trends in a data time series according to the invention is particularly suitable for onboard applications in a vehicle.

[0051] In one embodiment, the system is a moving part in a machine. The sensor may then in particular be a sensor for detecting vibrations of a part of the machine, wherein the response of the system comprises shutting down the machine or marking the machine as requiring maintenance.

[0052] It can be said that every machine must inevitably be maintained. However, it is often difficult to estimate the true wear and tear, so that maintenance is performed that would still be unnecessary. Maintenance costs the operator of the machine money and time, so that maintenance is delayed if possible. Indicators that indicate the necessity of maintenance enable maintenance to be performed at the most efficient time point.

[0053] Machines with moving parts usually vibrate. The vibrations can act like fingerprints for the working process of the machine. If wear sets in, more friction can form at the moving parts, for example due to lack of lubricating fluid, worn lubrication plates or increased play between two parts. However, these effects lead to stronger vibrations in the machine.

[0054] If the amplitude of the vibration of a machine is detected as a data time series, the trend of the data time series can be checked using the proposed method. An upward trend in the data time series is then synonymous with stronger vibrations and therefore with higher wear. The moment from which the wear is no longer tolerable can be set above a threshold, which triggers a response of the system. In some cases, the machine can then be shut down for safety reasons. In other cases, it is sufficient to mark the machine. In particular, the machine can be marked as "maintenance required" in order to signal the operator that the machine should actually be repaired. This embodiment makes it possible to plan maintenance into the operation of the machine at the necessary time. As a result, costs for maintenance that is not yet necessary do not have to be incurred, which makes the operation of the machine more beneficial.

[0055] In another case, a machine can be marked as "ready to operate". If a new or freshly maintained machine is put into operation, it is possible that it must first be prepared for operation. For example, lubricant must first be distributed in the moving parts, or specific components must reach a rated temperature. For example, a sensor can detect the vibrations of a machine as a data time series. If the amplitude of the vibrations decreases over time, this can be identified as a trend as long as the characteristic value for the vibration time series is above a corresponding threshold value. For example, such a downward trend may trigger the marking of the machine as "heating up", "in preparation", etc. However, if the characteristic value reaches a level that is below the threshold value, the marking expires and the machine is indirectly or directly marked as "ready to operate". Indirect marking as "ready to operate" may in particular include omitting the marking "not ready to operate" or a similar marking.

[0056] In another embodiment, the system is a fuel cell. The sensor detects the pressure in the fuel cell, the current provided by the fuel cell, the temperature at or in the fuel cell, or generally a virtually determined operating state of the fuel cell. In this embodiment, the response of the system may include, in particular, reducing the power of the fuel cell or shutting down the fuel cell.

[0057] Fuel cells are based on the principle of electrochemical reaction between fuel and oxidant, thereby generating electrical energy. The fuel cell consists of a number of main components, including an anode, a cathode and an electrolyte. The anode serves as a place for oxidation of the fuel, while the cathode is responsible for the reduction of the oxidant. The electrolyte enables the transport of ions between the anode and the cathode. The electrolyte layer between the anode and the cathode enables the flow of ions, thereby generating an electric current.

[0058] In order to ensure the high efficiency and long-term power of the fuel cell, it is very important to choose the right catalyst. The catalyst should be highly active in order to accelerate the reactions at the anode and cathode. In addition, the catalyst should be stable and have good tolerance to the operating conditions of the fuel cell, including temperature, pressure and fuel composition.

[0059] Each of the parameters mentioned can be detected by a corresponding sensor as a data time series. If a trend is detected in one of these data time series, this can indicate an abnormal, possibly inefficient, operation of the fuel cell.

[0060] By switching off or reducing the power of the fuel cell, the fuel cell can be protected from damage if necessary or at least protected before it is damaged. In addition, the fuel cell can be labeled, in particular as "maintenance required".

[0061] In another aspect, the invention relates to a computer program with a program code for performing a method as previously described for controlling a system based on trends detected in a time series of data detected by a sensor, when the computer program is executed on a computer.

[0062] In another aspect, the invention relates to a computer-readable data carrier having a program code of a computer program for performing a method as previously described for controlling a system based on trends detected in a data time series detected by a sensor, when the computer program is executed on a computer.

[0063] In another aspect, the present invention relates to a subsystem for a system for executing a method for controlling the system based on trends detected in a time series of data detected by a sensor, wherein the subsystem has a device for executing the method for controlling the system based on the detected trends as previously described.

[0064] The subsystem can be implemented in particular as part of the system. For example, the subsystem can be designed as part of a fuel cell, which takes over the monitoring of the detected data time series.

[0065] In general, a computer-implemented method for controlling a system based on trends detected in a time series of data sensed by a sensor and an apparatus for performing the method are thus described.

[0066] The described embodiments and developments can be combined with one another as desired.

[0067] Further possible embodiments, refinements and implementations of the invention also include combinations of features of the invention described above or below with reference to the exemplary embodiments, which are not explicitly mentioned. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The accompanying drawings should facilitate a further understanding of the embodiments of the present invention. The drawings illustrate the embodiments and, together with the description, serve to explain the principles and concepts of the present invention.

[0069] Further embodiments and many of the advantages mentioned are apparent from the drawings. The elements shown in the drawings are not necessarily shown in true scale with respect to one another.

[0070] Figure 1 Schematically illustrating the process of a method according to one embodiment;

[0071] Figure 2 Schematically illustrating a partial process of a method according to one embodiment;

[0072] Figure 3a and 3b Two exemplary data time series with linear filters and their products are shown in tabular format;

[0073] Figure 4 showing a time series of data represented as a graph in which one or more trends have been identified; and

[0074] Figure 5 A subsystem for performing the method according to one embodiment is shown.

[0075] Unless otherwise indicated, in the figures of the drawings, identical reference signs denote identical or functionally identical elements, components or parts. DETAILED DESCRIPTION

[0076] Figure 1 The flow of the method according to one embodiment of the present invention is schematically shown. The method starts with providing a time series of data detected by a sensor, that is, starting from step S10.

[0077] In one embodiment, the data time series map characterizes the time course of a parameter of the system controlled within the scope of the method. The sensor can preferably be part of the system and forward the data time series to the subsystem of the system monitored in this way. This can make it possible to check the trend of the data time series during the ongoing operation of the system.

[0078] In one embodiment, it is possible that the sensor detects parameters of a system different from the system that triggers the response as a data time series. For example, the sensor can detect parameters of the vehicle, in particular the battery pack status, and forward these parameters to the test bench. If a trend is detected in the data time series of the vehicle parameters, a response is triggered at the test bench.

[0079] The data time series can be stored temporarily for carrying out the method. Preferably, the system has a dedicated working memory or an area in its working memory for this purpose. The size of the working memory or the working memory provided determines the number of data points of the data time series that can be used for trend identification.

[0080] In step S12, a linear filter function is applied to the detected data time series for at least one resolution L. By applying the filter function to the data time series, a new function appears, which is quantifiable. Therefore, from this new function, a characteristic value can be obtained in step S14, based on which an upward trend or a downward trend can be identified. The characteristic value can especially be the slope of a regression line through the new function.

[0081] In step S16, the characteristic value of the data time series is compared with a threshold value and a check is performed to see whether the characteristic value reaches the threshold value. Depending on the monitored system, the detected data time series and the expected trend, the threshold value may be selectively not exceeded or exceeded. If the aim is to protect the system from a too drastic increase in a parameter, such as the amplitude of a vibration detected by a motion sensor, a trend is detected when the characteristic value exceeds the threshold value. However, if, for example, the power of the system is measured and the system is monitored in such a way that the power should not fall below a certain value, a downward trend is detected when the characteristic variable does not exceed the threshold value.

[0082] If the characteristic value reaches a threshold value, a system response to the detection of a trend in the data time series is triggered in step S18. The triggered response depends on the system and the monitored parameters. The response can include marking the system with a specific operating state, in particular "maintenance required", or shutting down the system. In an embodiment, special processes can also be triggered when a trend is identified. For example, if the level shows an increasing trend, an additional drain valve for the intermediate storage of the liquid can be opened.

[0083] If no trend is detected when comparing the characteristic value with the threshold value, the method continues with step S10 and the process starts again.

[0084] In an embodiment, a second pass of the method can be performed with a second data time series, wherein a portion of the values ​​coincide with the values ​​of the first data time series from the first pass. Thus, for example, a parameter can be permanently monitored and its trend checked. For example, the method can be performed with a data time series comprising 100 values. In the case of the first pass, the trend of the first 100 values ​​is checked. In the second pass, 50 new values ​​are detected, so that the 100 values ​​of the data time series for the second pass consist of 50 values ​​of the first data time series of the first pass and 50 newly detected values. Thus, trends can be detected with double time resolution.

[0085] Figure 2 Schematically illustrating a method according to one embodiment Figure 1 The substep of step S12, namely the application of the filter function. In a first substep S12.1, the resolution for the data time series to be checked is specified. This specification can be made individually, that is, adapted to the conditions of the monitored system and / or external conditions. In particular, one or more resolutions can be predefined. Alternatively, the resolution can also be specified systematically and identically for each round.

[0086] Next, generate the linear function phi L (t)(S12.2), the linear function forms a linear filter. Function phi L (t) comprises as many values ​​as there are data time series or segments of data time series in such a way that trends are to be detected. The number of values ​​corresponds to the resolution L.

[0087] The data time series is in principle not limited to a specific number of values. Within the scope of the invention, the data time series consists of a defined number L of values ​​from an infinitely detectable data time series. This makes it possible to check the trend of the continuously detected data time series at regular intervals. For this purpose, the values ​​of the data time series can be continuously written to an intermediate memory, wherein the oldest values ​​of the data time series are overwritten with the latest values. The intermediate memory is read each time a trend in the data time series is to be detected. In step S12.3, the data time series detected in this way of length L is compared with the linear filter function phi. L (t) multiply.

[0088] exist Figure 2 In the final step S 12.4 shown in FIG. 1 , the linear filter function phi L (t) and the product of the data time series f(t). This results in the further method runs (see Figure 1 ) can be used as unique values ​​of the feature value.

[0089] If a linear function phi with L values L (t) appears to make the linear function increase linearly from -1 to +1, then when no trend can be identified in the data time series because the individual products should balance each other, the eigenvalue will be 0 or close to 0. However, if there is a trend, the eigenvalue will be significantly greater or less than 0, because later values ​​in the data time series will have a greater influence on the eigenvalue than earlier values.

[0090] Figure 3a and 3bThe two data time series f(t) are shown in a tabular format and the data time series are compared with the linear filter function phi L (t) and thus form a new function u(t). The values ​​of the new function u(t) can be aggregated to form the eigenvalues ​​of the data time series f(t).

[0091] exist Figure 3a A slight upward trend can be identified in the data time series f(t) shown in . This can be seen especially from the graphical representation of the data time series f(t). Figure 3b The data time series f(t) fluctuates back and forth between the values ​​1 and 2, and thus shows no trend, neither downward nor upward.

[0092] Since the data time series f(t) consists of five values ​​in this example, the linear filter function phi L (t) is divided into the same five steps. The filter function phi L (t) stretched linearly from -1 to +1 in 0.5 steps.

[0093] If the data time series f(t) is compared item by item with the filter function phi L (t) and the new function u(t) is obtained. Therefore, the function u(t) is calculated using the following

[0094] u(t)=f(t)*phi L (t).

[0095] The eigenvalue k of the data time series f(t) is then constructed as the sum of the values ​​of the function u(t).

[0096] k=∑ t u(t)=∑ t f(t)*phi L (t).

[0097] Therefore, from Figure 3a The characteristic value k of the data time series f(t) is 2.5 (=-1+-1.5++0+2+3). Figure 3b The characteristic value of the data time series f(t) is 0 (=-1+-1+0+1+1).

[0098] The eigenvalue is compared with a threshold value. The threshold value may be, for example, at 1. All eigenvalues ​​k above 1 or below -1 will then represent trends in the data time series f(t). Positive eigenvalues ​​will also characterize an upward trend, and negative eigenvalues ​​will characterize a downward trend.

[0099] However, it should be understood in principle that Figure 3a and 3bThe example shown in for a data time series fl(t) is a strong simplification which is intended only to illustrate the principle of the invention. In order to be able to make statistically relevant statements about trends, it is advantageous to use a data time series with as many values ​​as possible.

[0100] exist Figure 4 Another data time series f(t) over time t is shown in . The data time series itself is represented as a dotted line. The data time series may have been recorded when analyzing or monitoring the system and may contain one or more processes.

[0101] In this example, the data time series f(t) is checked for an upward trend. An upward trend is detected in segments A, B, C, D and E. It should also be noted that the upward trend in segment C is steeper than the upward trend in segment B. This can be identified, for example, by the fact that the feature value for segment C is higher than the feature value for segment B.

[0102] The data time series f(t) decreases between the segments A and B, C and D, and D and E. In principle, a decreasing trend can be detected here, but in this embodiment, it is not noticeable due to the selection of positive characteristic values ​​above a threshold value.

[0103] However, in addition to the rising trend in segments A to E, a long-term rising trend can also be detected over the entire data time series f(t) (summarized as segment F). The straight line reflecting this trend is also entered in the graph as a dotted line. It should be noted that the time scale and trend of the data time series f(t) must be checked. The data time series can be monitored on different time scales in order to detect micro or macro trends. To this end, the individual values ​​of the data time series can be stored in an intermediate memory over a longer period of time. If a micro trend is of interest, the trend of a portion of the data time series can be checked. If a macro trend is to be detected, the trend over a longer period of time is checked, and more values ​​are read out from the system accordingly.

[0104] Figure 5 One embodiment of a subsystem 10 for controlling a system based on trends detected in a time series of data acquired by a sensor 12 is shown schematically and exemplarily.

[0105] The subsystem 10 is designed to receive and process a data time series from a sensor 12. For this purpose, the subsystem 10 is communicatively connected to the sensor 12. The communication connection can be implemented wired or wirelessly. In particular, the sensor 12 or another subsystem that accommodates the sensor 12 sends the data time series as a continuous data stream or in the form of packets containing a plurality of values ​​to an intermediate memory 14 of the subsystem 10. The values ​​of the data time series are intermediately stored in the intermediate memory 14 until the values ​​are read out and further processed. The intermediate storage can be carried out until, for example, there is enough data to perform trend detection.

[0106] The computing unit 16 can read out the data time series completely or partially from the intermediate memory 14 in order to detect trends in the data time series or parts thereof according to the invention. The computing unit 16 can include, for example, a CPU, a GPU or any unit designed to perform computing operations.

[0107] The result of the check, i.e. the information about whether a trend was detected and, if so, which trend was detected and when it was detected, can be stored by the computing unit 16 in a further intermediate memory 18. In an embodiment, the intermediate memory 18 can be part of the intermediate memory 14. Both intermediate memories 14 and 18 can be implemented as virtual intermediate memories of the subsystem 10, located on a common hardware module or arranged separately from each other in different hardware modules.

[0108] If the computing unit 16 has detected a trend, the subsystem 10 sends a signal via the interface 20 to the system, which can then respond to the detected trend according to the signal. The interface 20 can be, for example, a USB or wireless interface, in particular a Bluetooth or WLAN interface. For example, the signal can contain specific action instructions for the system or activate action instructions stored in the system, so that the system itself decides which response is triggered using the information about the detected trend. The latter case is particularly advantageous because the subsystem can be combined with any system as an autonomous subsystem as long as a suitable interface for the interface 20 is present in the system.

Claims

1. A computer implemented method for controlling a system based on trends detected in a time series of data detected by a sensor, The method comprises the following steps: - Detecting and providing data time series through sensors (S10); - applying at least one linear filter function with a resolution L to each time point t of the data time series (S12), and obtaining for each resolution L a characteristic value for quantifying the time evolution of the data time series (S14); - when one of the characteristic values ​​of the data time series reaches a threshold value, a trend in the data time series is detected (S16); and - When a trend has been detected in the data time series, triggering a response of the system (S18). 2 . The computer-implemented method of claim 1 , wherein the trend is an upward trend or a downward trend.

3. The computer-implemented method according to any one of the preceding claims, wherein applying the filter function (S12) for a resolution L comprises the following steps: -specify the resolution L (S12.1); - Create a linear function phi with L values L (t)(S12.2); - determining a segment of the data time series for a region tL to t having L values; - The linear function phi L (t) multiplying by the segment of the data time series (S12.3); and - for all L values ​​of the linear function phi L (t) and all products of the data time series are summed up (S12.4). 4 . The computer-implemented method of claim 1 , wherein the system is a test bench for a motor vehicle and the response of the system comprises shutting down the test bench and / or the motor vehicle. 5 . 5 . The computer-implemented method of claim 4 , wherein the sensor is a sensor for measuring temperature, and wherein the time series of data determines a temperature of a subsystem of the motor vehicle.

6. The computer-implemented method of any one of the preceding claims, wherein the system is a moving part in a machine, wherein the sensor is a sensor for detecting vibration of a component, and The response of the system may include shutting down the machine or marking the machine, in particular as requiring maintenance.

7. The computer-implemented method of any one of the preceding claims, wherein the system is a fuel cell, wherein the sensor is a sensor for detecting pressure in the fuel cell, a current provided by the fuel cell, a temperature sensor at the fuel cell or a virtual sensor for detecting an operating state of the fuel cell, and The response of the system includes reducing the power of the fuel cell or shutting down the fuel cell.

8. A computer program having a program code for carrying out the method according to any one of the preceding claims for controlling a system based on trends detected in a time series of data acquired by a sensor, when the computer program is executed on a computer.

9. A computer-readable data carrier having a program code of a computer program for executing the method according to any one of claims 1 to 7 for controlling a system based on trends detected in a time series of data acquired by a sensor, when the computer program is executed on a computer.

10. A subsystem for a system for executing a method for controlling the system based on a trend detected in a data time series detected by a sensor 12, wherein the subsystem has means for executing the method for controlling the system based on the detected trend according to any one of claims 1 to 7.

Images