Method for synchronizing time sequence of sensor values relating to machining process

Abstract

The invention relates to a method for synchronizing time sequences of sensor values relating to a machining process, the method (1) having the following steps: during at least two implementation of the machining process, at least two time sequences (2) of sensor values are detected by means of the same sensor in each case; -for each of the at least two time sequences of sensor values, ascertaining a deviation of a different point of the respective time sequence from a respective point of the reference time sequence, and shifting the respective time sequence with a value ascertained on the basis of the ascertained deviation in order to synchronize the time sequences (3); and-providing a synchronized time sequence (4).

Description

Technical Field

[0001] The present invention relates to a method for synchronizing time series of sensor values regarding a processing process, by which the time series of sensor values regarding the processing process can be synchronized in a simple and intuitive way and manner. Background Art

[0002] Herein, the processing process or production process is understood as the standardized, i.e., specific requirements of the corresponding workflow for manufacturing or processing products, such as semiconductor modules. Herein, the processing process consists of a pre-given processing method and working and production means, such that salable products are manufactured especially by machining and processing.

[0003] In the scope of quality assurance during the processing process, the products manufactured by the processing process usually undergo inspection after the actual processing, wherein it is checked whether there are deviations from the standard or anomalies. Based on this inspection, it can then be decided, for example, whether the corresponding product should be directly further processed or applied, or scrapped or removed, for example in order to avoid safety risks when using the object subsequently.

[0004] Herein, "anomaly" is understood as the formed anomaly, irregularity or deviation from a pre-given standard, such as scratches formed on the surface of the machined component or undesired gaps or openings formed between the individual parts of the object.

[0005] Herein, for example, such anomalies can be identified based on the sensor values regarding the processing process, i.e., based on the sensor values detected by the processing process or the corresponding time series of sensor values, i.e., the corresponding series of continuously detected sensor values, during the implementation of the processing process or the processing of the product. However, it has proven disadvantageous here that such time series of sensor values often shift relative to each other, i.e., the characteristic time points during a repeatable processing process occur at different time points and are not the same in different time series of sensor values. Therefore, the individual time series of sensor values must be synchronized before their evaluation, especially before their use for anomaly detection.

[0006] A method for synchronizing a plurality of sensor systems for providing data for corresponding phases is known from the published document DE 102019216517 B3, wherein the sensor systems each periodically provide data, wherein for each of the plurality of sensor systems at least one evaluation component's phase requirements are generated, wherein at least one evaluation component is configured to receive the data of the plurality of sensor systems, check all the phase requirements for feasibility, and as long as the phase requirements have been checked as feasible, generate target phase requirements, and transmit the corresponding target phase requirements to the corresponding sensor systems for synchronizing the phases of the plurality of different sensor systems. Summary of the Invention

[0007] Therefore, the object underlying the present invention is to describe an improved method for synchronizing time series of sensor values regarding a machining process.

[0008] This object is achieved by a method for synchronizing time series of sensor values regarding a machining process according to the features of claim 1.

[0009] Furthermore, this object is also achieved by a system for synchronizing time series of sensor values regarding a machining process according to the features of claim 6.

[0010] According to an embodiment of the present invention, this object is achieved by a method for synchronizing time series of sensor values regarding a machining process, wherein the method has: detecting time series of sensor values by the same sensor during at least two executions of the machining process; for each of at least two time series of sensor values, respectively obtaining the deviation between different points of the corresponding time series and the corresponding points of a reference time series, and shifting the corresponding time series with values obtained based on the obtained deviations in order to synchronize the time series; and providing the synchronized time series.

[0011] A sensor, which is also referred to as a detector, (measurement parameter or measurement) receiver, or (measurement) detector, is a technical component that can qualitatively or quantitatively detect specific physical or chemical properties and / or substance states of its surrounding environment as a measurement parameter.

[0012] Here, the "reference time series" is understood as a time series of reference values or a comparison time series. The deviation between different points of the corresponding time series and the corresponding points of the reference time series and the shifting of the corresponding time series herein mean respectively obtaining the deviation or difference between characteristic time points within or above the corresponding time series during the machining process and the corresponding time points within or above the reference time series.

[0013] Methods of this kind have the advantage that they can synchronize time series of sensor values regarding a machining process in a simple and intuitive way and manner. In particular, the time series can be synchronized based on non-invasive methods, especially when each time series only progresses with certain values, that is to say, the sensor values hardly change and the characteristics of the corresponding time series are basically retained. In addition, there is no urgent need to define hyperparameters or configuration parameters.

[0014] Overall, therefore, an improved method for synchronizing time series of sensor values regarding a machining process is proposed.

[0015] Step: For each of at least two time series of sensor values, determine the deviation of the corresponding points of the respective time series from the corresponding points of a reference time series, and shift the respective time series with the value obtained based on the determined deviation in order to synchronize the time series. This step can use the dynamic time warping algorithm here.

[0016] Here, dynamic time normalization or dynamic time warping represents an algorithm that maps numerical sequences of different lengths to each other.

[0017] Therefore, relatively accurate results can be obtained automatically in a simple way and manner without the urgent need for resource-intensive adjustment.

[0018] In one embodiment, the method further has the step here: approximate the reference time series based on the training data of the dynamic time warping algorithm.

[0019] Here, approximating or approaching the reference time series based on at least two time series of sensor values means obtaining or evaluating the reference time series based on at least two time series of sensor values.

[0020] Therefore, the reference time series can be derived based on known values in a simple way and manner without the urgent need for costly and resource-intensive adjustment.

[0021] However, "approximate the reference time series based on the training data of the dynamic time warping algorithm" is only a feasible embodiment. More precisely, for example, the reference time series can also be determined by a process expert.

[0022] In addition, in the step of, for each of at least two time series of sensor values, respectively obtaining the deviation between different points of the corresponding time series and the corresponding points of the reference time series, and shifting the corresponding time series with a value obtained or derived based on the obtained deviation, sensor values outside a pre-given value range can be disregarded. Thereby, it can be ensured that the synchronization is not distorted due to outliers in the sensor values and, more precisely, the synchronization is based on relevant sensor values.

[0023] Herein, the pre-given value range can be pre-given by a process expert, for example.

[0024] In addition, in the step of, for each of at least two time series of sensor values, respectively obtaining the deviation between different points of the corresponding time series and the corresponding points of the reference time series, and shifting the corresponding time series according to the value obtained based on the obtained deviation, sensor values can be disregarded which, when shifting the corresponding time series, have an interval greater than an interval threshold relative to the reference curve. Thereby, the user and / or the process expert who pre-give the threshold can determine the value for shifting the time series that can be maximally shifted.

[0025] With a further embodiment of the present invention, a method for identifying anomalies in products manufactured by a manufacturing process is also proposed, wherein the method has: during at least two implementations of the manufacturing process, respectively detecting time series of sensor values by the same sensor; synchronizing at least two time series by the method for synchronizing time series of sensor values regarding the manufacturing process described above; and identifying anomalies in products manufactured by the manufacturing process based on the synchronized time series.

[0026] Therefore, a method for identifying anomalies in products manufactured by a manufacturing process is proposed, which is based on the following sensor time series that are synchronized by an improved method for synchronizing time series of sensor values regarding the manufacturing process. The method for synchronizing time series of sensor values regarding the manufacturing process particularly has the advantage that the time series of sensor values regarding the manufacturing process can be synchronized in a simple and intuitive way and manner. In particular, the time series can be synchronized based on a non-invasive method, especially each time series is only shifted by a determined value, that is, the sensor values are hardly changed and the characteristics of the corresponding time series are basically retained. In addition, it is not urgently necessary to define hyperparameters or configuration parameters.

[0027] Furthermore, another embodiment of the present invention also provides a system for synchronizing time series of sensor values regarding a machining process, wherein the system has: a sensor configured to respectively detect time series of sensor values during at least two executions of the machining process; a synchronization unit configured to, for each of at least two time series of sensor values, respectively determine the deviation between different points of the corresponding time series and the corresponding points of a reference time series, and shift the corresponding time series with values obtained based on the determined deviations in order to synchronize the time series; and a providing unit configured to provide the synchronized time series.

[0028] Accordingly, an improved system for synchronizing time series of sensor values regarding a machining process is proposed. Such a system has the advantage that the time series of sensor values regarding a machining process can be synchronized in a simple and intuitive manner. In particular, the time series can be synchronized based on a non-invasive method, especially each time series is only shifted by a determined value, that is, the sensor values are hardly changed and the characteristics of the corresponding time series are basically retained. In addition, there is no urgent need to define hyperparameters or configuration parameters.

[0029] Here, the synchronization unit can be configured to use the dynamic time warping algorithm to synchronize the time series. Thus, relatively accurate results can be obtained automatically in a simple manner and without the urgent need for resource-intensive adjustment.

[0030] In one embodiment, the system further has an approximation unit configured to approximate the reference time series based on the training data of the dynamic time warping algorithm. Thus, the reference time series can be derived based on known values in a simple manner without the urgent need for costly and resource-intensive adjustment.

[0031] However, "the system further has an approximation unit configured to approximate the reference time series based on the training data of the dynamic time warping algorithm" is only a feasible embodiment. More precisely, for example, the reference time series can also be determined by a process expert.

[0032] Furthermore, the synchronization unit can be configured to ignore sensor values outside a pre-given value range. Thereby, it can be ensured that the synchronization is not distorted due to outliers in the sensor values and more precisely the synchronization is based on relevant sensor values.

[0033] In addition, the synchronization unit is capable of constructing such that, when shifting the corresponding time series, sensor values are not considered that have an interval greater than the interval threshold with respect to the corresponding points on the reference curve. Thereby, it is possible for the user and / or process expert who pre-specify the threshold to determine the values for shifting the time series that can be maximally shifted.

[0034] In addition, with a further embodiment of the present invention, a system for identifying anomalies in products manufactured by a manufacturing process is proposed, wherein the system has: a sensor configured to detect a time series of sensor values respectively during at least two implementations of the manufacturing process; the system for synchronizing the time series of sensor values regarding the manufacturing process described above, which is configured to synchronize at least two detected time series; and an identification unit configured to identify anomalies in products manufactured by the manufacturing process based on the synchronized time series.

[0035] Thus, a system for identifying anomalies in products manufactured by a manufacturing process is proposed, which is based on the following sensor time series, and the sensor time series is synchronized by an improved system for synchronizing the time series of sensor values regarding the manufacturing process. The system for synchronizing the time series of sensor values regarding the manufacturing process particularly has the advantage that the time series of sensor values regarding the manufacturing process can be synchronized in a simple and intuitive manner. In particular, the time series can be synchronized based on a non-invasive method, and in particular, each time series is only shifted by a determined value, that is, the sensor values are hardly changed and the characteristics of the corresponding time series are basically retained. In addition, it is not urgently necessary to define hyperparameters or configuration parameters.

[0036] Generally speaking, it is determined that with the present invention, a method for synchronizing the time series of sensor values regarding the manufacturing process is proposed, and by using this method, the time series of sensor values regarding the manufacturing process can be synchronized in a simple and intuitive manner.

[0037] The described design solutions and improvements can be combined with each other arbitrarily.

[0038] Further feasible design solutions, improvements and implementations of the present invention also include combinations of features of the present invention described above or below regarding the embodiments that are not explicitly mentioned. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings are intended to introduce a further understanding of the embodiments of the present invention. The drawings show the embodiments and are used in combination with the description to explain the principles and solutions of the present invention.

[0040] Many advantages of other embodiments and those mentioned are obtained from the accompanying drawings. The elements shown in the drawings are not necessarily shown in true proportion relative to one another.

[0041] Wherein:

[0042] Figure 1 A flowchart of a method for synchronizing time series of sensor values regarding a machining process according to an embodiment of the present invention is shown; and

[0043] Figure 2 A schematic block diagram of a system for synchronizing time series of sensor values regarding a machining process according to an embodiment of the present invention is shown.

[0044] As long as there is no contrary indication, in the figures of the drawings, the same reference numerals denote the same or functionally identical elements, components or assemblies. Detailed Description

[0045] Figure 1 A flowchart of a method for synchronizing time series of sensor values regarding a machining process 1 according to an embodiment of the present invention is shown.

[0046] In the scope of quality assurance during a machining process, the products manufactured by the machining process usually undergo inspection after the actual machining, wherein it is checked whether there are deviations from the standard or anomalies. Based on this inspection, it can then be decided, for example, whether the corresponding product should be directly further processed or applied, or scrapped or removed, for example in order to avoid safety risks when using the object subsequently.

[0047] Herein, "anomaly" is understood as a formed anomaly, irregularity or deviation from a pre-given standard, for example a scratch formed on the surface of a machined component or an undesired gap or opening formed between the individual parts of an object.

[0048] Herein, such anomalies can be identified, for example, based on sensor values regarding the machining process, that is to say based on the sensor values detected by the machining process or the corresponding time series of sensor values, that is to say the corresponding sequences of continuously detected sensor values, during the implementation of the machining process or the machining of the product. However, it has proven disadvantageous here that such time series of sensor values often shift relative to one another, that is to say the characteristic time points during a repeatable machining process occur at different time points in different time series of sensor values and are not the same. Therefore, the individual time series of sensor values must be synchronized before they are evaluated, in particular before they are used for anomaly detection.

[0049] Herein, Figure 1A method 1 is shown, which has: Step 2: Detecting a time series of sensor values accordingly by the same sensor during at least two implementations of the processing; Step 3: For each of at least two time series of sensor values, correspondingly obtaining the deviation of different points of the corresponding time series from the corresponding points of a reference time series, and shifting the corresponding time series by the value obtained based on the obtained deviation in order to synchronize the time series; and Step 4: Providing the synchronized time series.

[0050] This kind of method 1 has the advantage that the time series of sensor values regarding the processing can be synchronized in a simple and intuitive way and manner. In particular, the time series can be synchronized based on a non-invasive method, especially each time series is only shifted by a determined value, that is, the sensor values are hardly changed and the characteristics of the corresponding time series are basically retained. In addition, it is not urgently necessary to define hyperparameters or configuration parameters.

[0051] Therefore, an improved method for synchronizing the time series of sensor values regarding the processing is generally proposed.

[0052] Figure 1 In particular, a method is shown in which sensor time series with different start time points are moved to overlap each other.

[0053] Step 3: For each of at least two time series of sensor values, correspondingly obtaining the deviation of different points of the corresponding time series from the corresponding points of a reference time series, and shifting the corresponding time series by the value obtained based on the obtained deviation in order to synchronize the time series. This step is based on the dynamic time warping algorithm according to Figure 1 the implementation mode. Here, for each point among different points of the time series, the corresponding point on the reference time series can be obtained respectively at one or more identical time points. Based on the intervals obtained in this way, the most frequently occurring interval can then be obtained, and the time series is shifted to the left or right by the corresponding value in the corresponding time sequence diagram. In particular, if the corresponding time series is shifted to the right, then subsequently, for example, the missing or required additional values of the time series can be filled accordingly by linear interpolation or a retrograde filling principle (ein auffüllendes Prinzip). In contrast, if the time series is shifted to the left, the corresponding number of original or initial sensor values is rejected or deleted.

[0054] In addition, for each process step or each working method, the individual time series of sensor values can be detected separately and synchronized accordingly. In addition, the user can specify or designate a determined range in which the time series should be synchronized, where the method 1 only executes for this range.

[0055] According to Figure 1 an embodiment of, the method further has step 5: approximating a reference time series.

[0056] Herein, the reference time series can in particular be approximated based on the training data of a corresponding dynamic time warping algorithm, for example in such a way that: a barycenter or an average value based on a specific distance measure is obtained based on the training data.

[0057] According to Figure 1 an embodiment of, step 3: for each of at least two time series of sensor values, correspondingly obtaining the deviation of different points of the corresponding time series from the corresponding points of the reference time series and shifting the corresponding time series according to the value obtained based on the obtained deviation, and this step does not consider sensor values outside a pre-given value range.

[0058] According to Figure 1 an embodiment of, step 3: for each of at least two time series of sensor values, correspondingly obtaining the deviation of different points of the corresponding time series from the corresponding points of the reference time series, and shifting the corresponding time series according to the value obtained based on the obtained deviation, and this step also does not consider the following sensor values, which have an interval greater than an interval threshold relative to the corresponding points on the reference curve when shifting the corresponding time series.

[0059] Subsequently, the corresponding synchronized time series can be used as the basis for further data analysis or as input data for a corresponding machine learning algorithm. In particular, the synchronized time series can form the basis for anomaly detection.

[0060] Figure 2 FIG. shows a schematic block diagram of a system for synchronizing time series of sensor values regarding a machining process 10 according to an embodiment of the present invention.

[0061] As Figure 2 shown, the system 10 herein has: a sensor 11 configured to respectively detect a time series of sensor values during at least two implementations of the machining process; a synchronization unit 12 configured to, for each of at least two time series of sensor values, respectively obtain the deviation of different points of the corresponding time series from the corresponding points of the reference time series, and shift the corresponding time series by the value obtained based on the obtained deviation so as to synchronize the time series; and a providing unit 13 configured to provide the synchronized time series.

[0062] Herein, the sensor can for example be a temperature sensor, a pressure sensor or a vibration sensor.

[0063] Furthermore, the synchronization unit can be implemented, for example, based on code stored in a memory and executable by a processor.

[0064] Furthermore, the providing unit is in particular a transmitter configured to transmit the corresponding data.

[0065] According to Figure 2 an embodiment, the synchronization unit is configured to use a dynamic time warping algorithm to synchronize at least two time series of sensor values.

[0066] As Figure 2 further shown, the system 10 also has an approximation unit 14 configured to approximate a reference time series based on training data of the dynamic time warping algorithm.

[0067] The approximation unit can in turn be implemented, for example, based on code stored in a memory and executable by a processor.

[0068] According to Figure 2 an embodiment, the synchronization unit 12 is also configured to disregard sensor values outside a pre-given value range.

[0069] According to Figure 2 an embodiment, the synchronization unit 12 is also configured to disregard sensor values that have a distance greater than a distance threshold relative to a corresponding point on a reference curve when shifting the corresponding time series.

[0070] Furthermore, the illustrated system 10 is configured to perform the method described above for synchronizing time series of sensor values regarding a machining process.

Claims

1. A method for synchronizing a time series of sensor values ​​relating to a machining process, wherein: The method (1) comprises the following steps: - during at least two executions of the machining process, at least two time series (2) of sensor values ​​are respectively detected by the same sensor; - for each of the at least two time series of sensor values, respectively determine the deviation of different points of the corresponding time series from the corresponding point of the reference time series, and shift the corresponding time series by a value determined based on the determined deviation, so as to synchronize the time series (3); and - Provides synchronized time series (4).

2. The method (1) according to claim 1, wherein: The steps include: for each of at least two time series of sensor values, respectively calculating the deviations between different points of the corresponding time series and corresponding points of the reference time series, and shifting the corresponding time series by a value calculated based on the calculated deviations so as to synchronize the time series (3), and this step uses a dynamic time warping algorithm.

3. The method (1) according to claim 2, wherein: The method (1) further comprises the following steps: - Approximating the reference time series based on training data of a dynamic time warping algorithm (5).

4. The method (1) according to any one of claims 1 to 3, wherein: The steps include: for each of at least two time series of sensor values, respectively calculating the deviations of different points of the corresponding time series with respect to the corresponding points of the reference time series, and shifting the corresponding time series according to a value calculated based on the calculated deviation (3), wherein the step does not take into account sensor values ​​outside a predetermined value range.

5. The method (1) according to any one of claims 1 to 4, wherein: The steps include: for each of at least two time series of sensor values, respectively calculating the deviations of different points of the corresponding time series with respect to the corresponding points of the reference time series, and shifting the corresponding time series according to the values ​​calculated based on the calculated deviations (3). This step does not take into account the following sensor values, which have an interval greater than an interval threshold relative to the corresponding point on the reference curve when shifting the corresponding time series.

6. A method for identifying anomalies in a product manufactured by a process, wherein: The method comprises the following steps: - during at least two executions of the machining process, respectively detecting a temporal sequence of sensor values ​​by means of the same sensor; - synchronizing at least two time series by means of a method for synchronizing time series of sensor values ​​relating to a machining process according to any one of claims 1 to 5; and -Identify anomalies in products manufactured through a machining process based on synchronized time series.

7. A system for synchronizing a time series of sensor values ​​relating to a machining process, wherein: The system (10) comprises: a sensor (11) which is designed to detect a temporal sequence of sensor values ​​during at least two executions of the machining process; A synchronization unit (12) configured to determine, for each of at least two time series of sensor values, deviations of different points of the corresponding time series from corresponding points of a reference time series, and to shift the corresponding time series by a value determined based on the determined deviations, in order to synchronize the time series; as well as A providing unit (13) is designed to provide a synchronized time series.

8. The system (10) according to claim 7, wherein: The synchronization unit (12) is designed to use a dynamic time warping algorithm to synchronize the time series.

9. The system (10) according to claim 7 or 8, wherein: The synchronization unit (12) is designed to exclude sensor values ​​which have a spacing greater than a spacing threshold value relative to a corresponding point on a reference curve when shifting the corresponding time series.

10. A system for identifying anomalies in a product manufactured by a process, wherein: The system has: a sensor, which is designed to detect a temporal sequence of sensor values ​​during at least two executions of the machining process; The system for synchronizing time series of sensor values ​​related to a machining process according to any one of claims 7 to 9, the system being designed to synchronize at least two detected time series; as well as A recognition unit is designed to recognize anomalies in a product produced by a manufacturing process based on the synchronized time sequence.

Images