Monitoring method for coating equipment and corresponding coating equipment

Abstract

The invention relates to a method for monitoring a coating device for applying a coating agent, in particular a coating device for coating a motor vehicle body part, the method comprising the following steps: - determining first raw sensor data (16) representing an operating variable of the coating device, - obtaining a first control signal for controlling the coating device, and - extracting features (15) from the first raw sensor data (16), the features (15) comprising a reduced amount of data compared to the first raw sensor data (16) and serving as a data basis for a machine learning algorithm. The invention also comprises a coating device monitored accordingly.

Description

Technical Field

[0001] The invention relates to a method for monitoring an application system for applying a coating agent, in particular for a coating system for coating motor vehicle body parts. The invention also includes a correspondingly designed coating system. Background Art

[0002] In modern painting plants for painting motor vehicle body parts, the operation is controlled by control signals, wherein a number of sensors detect operating variables of the painting plant and generate raw sensor data, so that the operating state of the painting plant can be monitored by evaluating the raw sensor data and the control signals. However, monitoring methods for monitoring the operation of painting plants have hitherto not been completely satisfactory.

[0003] DE 10 2007 062 132 A1 discloses a method for monitoring a coating system for applying coating agents, in particular a coating system for coating motor vehicle body parts. Here, raw sensor data are determined and control signals are recorded and used for function checking.

[0004] DE 10 2018 214 170 A1 discloses a method and a device for quality assessment of a technical system, in which raw sensor data (“image data”) and control signals (“observation condition data”) are used as the data basis for a machine learning algorithm.

[0005] The publication DE 10 2016 217 948 A1 describes a method for predicting the quality of an adhesive joint, wherein the adhesive joint is applied by an adhesive processing device, which in particular has an application device for applying the adhesive. The measurement data are analyzed by comparison with a data model. The statistical data model created by means of a multivariate analysis method can thus be optimized by machine learning. Summary of the invention

[0006] The invention is therefore based on the object of creating an improved monitoring method and a corresponding coating device.

[0007] This object is achieved by the monitoring method according to the invention and by a corresponding application device according to the invention.

[0008] The present invention is based on the technical knowledge that machine learning methods can be advantageously used for monitoring application equipment (e.g. painting equipment), such as those known from Fayyad U.; Patetsky-Shapiro, G.; Smyth, P.: "From Data Mining to Knowledge Discovery in Databases", AI Magazine, Vol. 17, No. 3 (1996). Another literature source for machine learning methods is Bremer, M.: "Principles of Data Mining", Springer-Verlag (2009).

[0009] According to known monitoring methods, the monitoring method according to the invention first provides for determining first raw sensor data reflecting an operating variable of the coating device. For example, the first raw sensor data may reflect, as an operating variable, the shaping air pressure of the shaping air for forming the spray jet of the rotary atomizer. However, the first raw sensor data may also reflect other operating variables of the coating device, as will be explained in more detail below.

[0010] Furthermore, the monitoring method according to the invention also provides for obtaining a first control signal for controlling the application device according to known monitoring methods. For example, the first control signal can define a target value for a shaping air flow for forming the spray of a rotary atomizer. However, the first control signal can also be used to set other operating variables of the application device, as will be explained in more detail below.

[0011] The control signals together with the raw sensor data thus form a signal set which is evaluated within the scope of the invention. For example, the signal set can take into account the set value of the shaping air flow as control signal and the actual signal of the shaping air pressure as raw sensor data. However, the signal set can also include more than one actual signal and more than one control signal. In general, the signal set can include 1 ... m different control signals and 1 ... k different actual signals, whereby these signals can also belong to different control loops.

[0012] The monitoring method according to the invention differs from the prior art in that so-called features are extracted from the first raw sensor data, which features comprise a reduced amount of data compared to the first raw sensor data and serve as a data basis for a machine learning algorithm.

[0013] The term "feature" as used in the context of the present invention firstly has the meaning of reducing the amount of data by feature extraction, ie the extracted features have a reduced amount of data compared to the original sensor data.

[0014] Furthermore, the concept of "feature" used in the context of the present invention has the meaning of a feature abstracted from the exact operating process without losing information relevant to the state detection. For example, it is no longer necessary to compare only identical sequences of the same workpiece variable (e.g. body variable). Instead, it is sufficient to compare similar "brushes" in completely different car bodies, for example.

[0015] Furthermore, the term "feature" used in the context of the present invention has the meaning that the feature is suitable as a data basis for a machine learning algorithm, wherein the known machine learning algorithms themselves are based on the extracted features. Therefore, these features form a bridge between the raw sensor data and the machine learning algorithm, which can usually only achieve a stable data analysis process with the machine learning algorithm that can master limited resources.

[0016] In a preferred embodiment of the present invention, first and second raw sensor data reflecting different operating variables of the application equipment are determined, and features are extracted from both the first raw sensor data and the second raw sensor data, which can then be used as a database for a machine learning algorithm. For example, the first raw sensor data can represent the shaping air flow as an operating variable, while the second raw sensor data can represent the speed of the drive turbine of the rotary atomizer. However, the first and second raw sensor data can also reflect other operating variables of the application equipment, as will be described in more detail below.

[0017] Preferably, extracting features from the first and second raw sensor data, respectively, is performed as follows.

[0018] In one step, an observation period window is first defined within which the first or second raw sensor data is to be analyzed. For example, the observation period window may be a car body cycle, i.e. the period between the introduction of a motor vehicle body to be painted into the paint booth and the introduction of the next motor vehicle body into the paint booth. This has the advantage that status and monitoring can be performed during normal production. However, there are various possibilities within the scope of the invention regarding the observation window, which will be described in detail below.

[0019] In a further step, the observation period window is then divided into individual segments, the so-called comparison periods. In this way, the comparison periods can be defined by changes (jumps) in the control signals. For example, a new comparison period begins each time a control signal relevant for the application is changed within a vehicle body cycle. For example, the set value of the shaping air flow of the rotary atomizer can be analyzed for this purpose. However, within the scope of the invention, various other control signals can also be analyzed to determine the comparison period, as will be described in detail.

[0020] In a further step, the comparison period itself is then subdivided into a plurality of successive subsections, the subdivision being predetermined by the parameters.

[0021] For example, the comparison period may be divided into 0...n initial adjustment phases, 0...m subsequent adjusted phases, and 0...l subsequent remaining phases as sub-segments.

[0022] The regulating phase is an initial time phase within the comparison period, in which the first and / or second raw sensor data react to jumps in the first and / or second control signal, so that a stationary state cannot yet be described.

[0023] The controlled phase is a time period after the controlled phase in which the first and / or the second raw sensor data at least partially react to a jump in the control signal.

[0024] In order to avoid misunderstandings, it should be pointed out that the method itself does not check whether the raw sensor data in the adjustment phase or in the adjusted phase have oscillations. Instead, the subdivision of the adjustment phase and the adjusted phase is fixed by parameters, namely by a predefined duration of the adjustment phase and a likewise predefined duration of the adjusted phase.

[0025] On the other hand, the remaining phase is the remaining period after the adjustment phase and the adjusted phase until the end of the comparison period. In this case, the comparison period may contain a single remaining phase or no remaining phase at all. For example, if the sum of the adjustment phase and the adjusted phase is exactly equal to the duration of the comparison period, then the comparison period does not contain any remaining phase.

[0026] It should be noted here that, in the case where the comparison period is long enough to include the adjusted phase and the remaining phase in addition to the adjustment phase, the adjusted phase and the remaining phase are optional. However, if the comparison period is short, the comparison period may be divided into only the adjustment phase and the remaining phase. If the comparison period is very short, it is even possible that the comparison period only includes an adjustment phase and no adjusted phase at all.

[0027] In a further step, at least one statistical parameter is then calculated for each subsection of the first and / or second raw sensor data and, if necessary, for further raw sensor data within each subsection, the calculated statistical parameter being a component of the extracted features.

[0028] It has been briefly mentioned above that the comparison period may be limited by jumps or changes in the control signal, ie one change in the control signal starts the comparison period, and the next change in the control signal ends the current comparison period and starts the next comparison period.

[0029] In one variant of the invention, the comparison period is limited by changes (jumps) of the same control signal. In this case, only one control signal, for example a set value for the shaping air flow of a rotary atomizer, is evaluated when determining the comparison period. On the other hand, in this variant of the invention, possible jumps of other control signals are not used to determine the comparison period.

[0030] On the other hand, in another inventive variant, a plurality of control signals are used to define the comparison period. Here, the comparison period is also limited by the change or jump of the control signal. However, different types of control signals are analyzed and processed. For example, a jump of one control signal can mark the beginning of the comparison period, while a jump of another control signal marks the end of the current comparison period and the beginning of the next comparison period.

[0031] It has been briefly mentioned above that the statistical features determined in each sub-segment of the comparison period are components of the extracted features. However, within the scope of the present invention, the extracted features may also contain further information. For example, these features may include the amount of setpoint change in the control signal and the time period since the last change, to name a few examples.

[0032] Furthermore, it has already been mentioned above that the comparison period is subdivided into subsections, i.e., for example, 0 ... n control phases, 0 ... m controlled phases and 0 ... 1 remaining phases. It is pointed out again at this point that the comparison period does not necessarily have to contain a plurality of control phases and a plurality of controlled phases. For example, the comparison period may also contain only one control phase. The number of subsections and the duration of the subsections can be determined here based on the time constants of the raw sensor data. For this purpose, the time constants of the first and / or second raw sensor data are determined, and then the number of subsections in the comparison period, the duration of a single control phase and / or the duration of a single controlled phase are determined based on the determined time constants. Only one time constant is selected for each use case, which may also contain a plurality of control and sensor data.

[0033] Alternatively, however, the duration of the individual control phases and / or the individual controlled phases can also be determined randomly.

[0034] As mentioned above, the observation period window can be, for example, a body cycle. This is a period of time that is relevant to the component to be coated, i.e., in particular to the component of the motor vehicle body to be painted. However, the observation period window can also alternatively be another period of time that is relevant to the component, for example, the period of time required to prepare the painting equipment for the subsequent coating of the component to be coated.

[0035] Furthermore, within the scope of the present invention, the observation period window may be a time period associated with the application equipment, such as a time period of a setup operation, a test operation, a manual operation, or a maintenance operation.

[0036] In addition, the observation period window can also be a time period related to time, such as one hour, one day, one week, one month, one quarter or one year. Another example of a time period related to the observation period window is the shift length of a work shift.

[0037] It has already been mentioned above that statistical parameters are determined in the individual subsections, which are then components of the extracted features.

[0038] In one variant of the invention, the statistical parameter is a univariate parameter taking into account only one type of raw sensor data. Examples of such univariate statistical parameters are the arithmetic or geometric mean, median, variance, maximum or minimum value within a sub-segment, wherein the univariate statistical parameter always relates to the same raw sensor data, i.e. to a single manipulated variable.

[0039] In another variant of the invention, the statistical parameter is a multivariate statistical parameter calculated from different raw sensor data. Examples of such multivariate statistical parameters are Pearson correlation coefficient and rank correlation coefficient. However, the multivariate statistical parameter may also take into account three or more different raw sensor data of the signal set.

[0040] Furthermore, it should be noted that the present invention is not limited to raw sensor data to be analyzed. On the contrary, within the scope of the present invention, a wide variety of raw sensor data generated during the operation of the application device can be analyzed. The following raw sensor data can be mentioned as examples:

[0041] - speed of the turbine of the rotary atomizer,

[0042] - the air pressure of the driving air for driving the turbine of the rotary atomizer,

[0043] - coating agent pressure at the coating pressure regulator,

[0044] -Filling current of electrostatic coating agent filling system,

[0045] -Filling voltage of electrostatic coating filling system,

[0046] - humidity in the coating booth,

[0047] - the air pressure of the shaping air for forming the coating agent spray,

[0048] - the flow rate of the shaping air used to form the coating agent spray,

[0049] - air temperature in the coating booth,

[0050] - the position of the coating impact point of the application device,

[0051] - the speed of movement of the coating impact point of the application device,

[0052] - valve positions of valves, in particular coating agent valves, flushing agent valves, pulse air valves or lubricant valves,

[0053] - drive variables of a drive, in particular a robot drive, in particular position, velocity, acceleration, current, voltage, power or temperature,

[0054] - the flow rate of the coating agent pump or metering unit,

[0055] - the position of the linear conveyor that transports the parts to be coated past the applicator,

[0056] - the flow rate, temperature and pressure of the material,

[0057] -Speed ​​of the swirl applicator.

[0058] It has already been mentioned above that various raw sensor data can be analyzed as part of the monitoring process according to the invention. In such monitoring of different raw sensor data, any combination of the above examples of raw sensor data is possible. For example, the rotation speed of the turbine of the rotary atomizer can be analyzed together with the flow rate of the coating agent, just to cite one example. Furthermore, within the scope of the invention, more than two different types of raw sensor data can also be analyzed, wherein any combination of the above examples of raw sensor data is possible.

[0059] For example, the raw sensor data may reflect an operating variable of one of the following components of the application equipment:

[0060] - Motor,

[0061] -Robot joints for coating robots,

[0062] - a drive controller for a robot drive of a coating robot,

[0063] - a turbine for driving a rotary atomizer,

[0064] - a shaping air controller for controlling the flow of shaping air through the spray of the shaping atomizer,

[0065] a pump, in particular a coating agent pump,

[0066] - a metering device for metering the coating agent,

[0067] - valves, in particular proportional valves,

[0068] -Paint pressure regulator,

[0069] - High voltage generator for electrostatic coating agent pick-up,

[0070] -switch,

[0071] -sensor,

[0072] - heater,

[0073] - control systems,

[0074] - electrical fuses,

[0075] -Battery,

[0076] - uninterrupted power supply,

[0077] -Uninterrupted signal transmission,

[0078] -transformer,

[0079] - fluid lines, in particular hoses or pipes, for conveying fluids, in particular coating agents, thick materials, adhesives, diluents, air or water,

[0080] -Swirl applicator.

[0081] Regarding the provision of raw sensor data, various possibilities exist within the scope of the invention. For example, the raw sensor data can be measured continuously and then analyzed. Alternatively, the raw sensor data can be read out from a database, provided that the raw sensor data is measured with an earlier time delay.

[0082] It was mentioned above that the features extracted from the raw sensor data are used as a database for the machine learning algorithm. These machine learning algorithms usually use so-called rules that are applied to the features and allow the operating state of the equipment to be inferred. To create this rule, historical raw sensor data measured and stored earlier can be used, where the so-called labels reflect the operating state of the application equipment during the measurement of the historical raw sensor data (for example, fault-free or faulty). Features are then extracted from the historical raw sensor data in the manner described above. Subsequently, rules are determined with the help of the features extracted from the historical raw sensor data and the features extracted from the known labels by the machine learning algorithm. The rules determined in this way can then be applied to the features extracted from the currently measured raw sensor data.

[0083] For example, the analysis of features extracted from the currently measured raw sensor data can be used to identify the operating state of the application device. For example, one of the following situations can be detected:

[0084] - wear or defect of the paint pressure controller,

[0085] - wear or defects in the mixer which mixes the various components of the coating agent with one another,

[0086] - wear or defects of the pump, in particular the coating agent pump,

[0087] - wear or defects of valves, in particular valves for controlling paint, thick substances, adhesives, thinners, air or water,

[0088] - wear or defect of the heater,

[0089] - wear or defect of the drive motor,

[0090] - electrical contact defects,

[0091] - detect bubbles in the paint or interruptions in the paint supply,

[0092] - the properties of the operating medium of the application device, in particular air, water, varnish, adhesive or thick substance,

[0093] - Analytical processing of the application of stress and motion sequences on machine parts or equipment,

[0094] - Check the spray from the bell of the rotary atomizer,

[0095] - detect contamination and / or moisture on the atomizer,

[0096] - Detect anomalies in the operating behavior of motors, pumps, pistons, shaping air, fans, turbines, high pressure, pressure flow regulators,

[0097] - abnormality in the sense of significant deviation of the overall detection signal curve from the normal curve,

[0098] -Predict maintenance intervals.

[0099] Furthermore, it should be mentioned that the term coating apparatus used in the context of the present invention is to be understood in general terms and includes in particular coating apparatus for coating motor vehicle body parts. However, the present invention can also be used, for example, in systems for bonding, sealing or insulating. Therefore, the term coating agent used in the context of the present invention is to be understood in general terms and includes in particular coatings and thick materials, such as adhesives or insulating materials.

[0100] Furthermore, it should be mentioned that the present invention does not only claim protection for the monitoring method according to the invention described above. Instead, the present invention also claims protection for a corresponding coating device, which comprises an applicator (e.g. a rotary atomizer), a manipulator (e.g. a multi-axis painting robot), a sensor for measuring raw sensor data, and a control device for controlling the coating device. The coating device according to the invention is characterized in that the control device performs the monitoring method according to the invention described above.

[0101] It should be mentioned here that the control in this sense can be distributed over multiple hardware components. It should also be mentioned that it does not have to be the application equipment control. The monitoring can run on completely different hardware that is completely unrelated to the application equipment control. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Further advantageous further embodiments of the invention are indicated in the dependent claims or are explained in more detail below together with the description of preferred embodiments of the invention with reference to the drawings. The drawings show:

[0103] Figure 1 is a schematic representation of the structure of a system for monitoring a coating device according to the present invention,

[0104] Figure 2 is a schematic diagram of the structure of the analysis software,

[0105] Figure 3 is a schematic illustration of the operating mode of the analysis software for feature extraction,

[0106] Figure 4 is a flow chart illustrating a monitoring method according to the present invention,

[0107] Figure 5 is a timing diagram illustrating feature extraction,

[0108] Figure 6 is a flow chart illustrating a monitoring method according to the present invention, and

[0109] Figure 7 is a time diagram illustrating feature extraction considering multiple control signals and multiple raw sensor data. DETAILED DESCRIPTION

[0110] Figure 1 The system structure of the system according to the invention is shown, which comprises a machine 1 , a controller 2 with embedded analysis software 3 , analysis hardware 4 with analysis software 5 , a visualization device 6 and hardware 7 belonging to an external system.

[0111] In the present exemplary embodiment, the machine 1 is a painting installation for painting motor vehicle body parts and all their components.

[0112] Here, the controller 2 outputs control signals to the machine 1 and receives raw sensor data from the machine 1 , and the raw sensor data can be analyzed by the analysis software 3 .

[0113] In addition, the raw sensor data can also be evaluated by the analysis software 5 of the analysis hardware 4 .

[0114] Figure 2A schematic representation of the modular structure of the analysis software 3 or 5 is shown.

[0115] The analysis software 3 can thus access the data storage 8 and the gateway 9 in order to communicate with other components such as the machine 1 or the hardware 7 .

[0116] Furthermore, the analysis software 3 includes a feature extraction module 10 , a rule generation module 11 and a rule application module 12 .

[0117] Figure 3 A schematic diagram illustrating the monitoring method according to the present invention is shown.

[0118] First, the feature extraction module 10 extracts features from historical raw data 13 previously measured in a known operating state of the machine 1, where the known operating state of the machine 1 is represented by a label 14. Then, the feature extraction module 10 extracts features 15 from the historical raw sensor data 13, as will be described in detail. The feature extraction of the module 10 has the task of reducing the data volume of the historical raw sensor data 13 so that the subsequent machine learning algorithm can be based on the features 15.

[0119] The module 11 then uses the extracted features 15 and the known labels 14 to create rules which are then used to later analyze the raw data 16 of the current measurement.

[0120] During normal operation, the module 10 then again extracts features from the new raw sensor data 16 currently being measured, and the module 12 applies the previously determined rules to the extracted features to determine the state 17 of the machine 1 .

[0121] In the following, reference will now be made to Figure 5 The timing diagram is used to describe the Figure 4 Flowchart of the process.

[0122] In a first step S1 , the application device is controlled, whereby raw signals are recorded, ie control signals for controlling the application device and actual signals (raw sensor data) which are measured by sensors and represent operating variables of the application device.

[0123] The original signals (control signal and actual signal) are then stored in step S2.

[0124] In step S3, the previously defined parameters for the respective application case are then retrieved. On the one hand, this includes the observation period window [T A , T E ], which can be, for example, a vehicle body cycle. On the other hand, the retrieved parameters also define the selection of the desired signal set, i.e. the control signal and the actual signal to be analyzed. Finally, the retrieved parameters also include parameters for dividing the comparison period [t1, t2] into subsegments, e.g. the regulation phase dDYN The number of n, the adjustment stage d DYN Duration and conditioning phase d STAT duration. After subtracting the adjustment phase d DYN After that, the static adjustment stage d STAT The amount is still appropriate for the comparison period.

[0125] In step S4, the previously stored data of the original signals (control signal and actual signal) are then retrieved, in particular for the observation period window [T A , T E ] and signal sets.

[0126] In step S5, the changes of the considered control signal are then determined, where these changes are based on Figure 5 occurs at times t1 and t2 in the timing diagram.

[0127] Subsequently, in step S6, at least one comparison period [t1, t2] is determined, the comparison period starting from the change of the control signal at time t1 and ending at the next change of the control signal at time t2. Typically, multiple comparison periods are considered.

[0128] In step S7, the comparison period [t1, t2] is then divided into sub-segments according to the parameters retrieved in step S3. DYN Duration and conditioning phase d STAT The duration of the comparison period [t1, t2] is then divided into 0...n regulation phases d DYN , 0...m regulated stages d STAT and 0...1 remaining stages. Figure 5 In the time diagram, the comparison period [t1, t2] consists of three adjustment stages d DYN and two regulated stages d STAT However, this is just an example, that is, the comparison period [t1, t2] may also be composed of sub-segments in a different manner.

[0129] In step S8, univariate statistical characteristics of the raw sensor data, ie the actual signal measured by the sensor and representing an operating variable of the application device, are determined for each of the individual segments. Figure 5 The time diagram of specifically shows that the shaping air flow of the rotary atomizer is measured as an actual signal. However, other actual signals can also be used to calculate the statistical parameters according to the signal set retrieved in step S3. For example, the univariate statistical parameter can be the arithmetic mean, just to give an example.

[0130] In step S9, features are then determined, of which the univariate statistical features calculated in step S8 are components. Other components of the features are the target value and the change in the target value.

[0131] In step S10 , the features are then used to determine the state of the application device, which can be done with simple (manually parameterized) rules or with a machine learning algorithm.

[0132] Figure 6 A flow chart is shown which, in addition to measuring raw sensor data in step S1 and extracting features in step S2 , also shows applying rules to the extracted features in step S3 .

[0133] Figure 7 Shown according to Figure 5 A modification of the timing diagram of FIG. 1, in which two control signals and two actual signals are shown. The comparison period is limited by the jumps of the two control signals.

[0134] The first comparison period is between time t1 and t2. The second comparison period is between time t2 and t3.

[0135] Then, the above-mentioned comparison period may be divided into sub-segments again in the above-mentioned manner.

[0136] Furthermore, statistical parameters are again calculated for each sub-segment, although these parameters may also be multivariate parameters, for example calculated from two actual signals. An example of such a multivariate statistical parameter is the Pearson correlation coefficient.

[0137] The invention is not limited to the preferred embodiments described above. On the contrary, a large number of variants and modifications are possible, which also utilize the concept of the invention and therefore fall within the scope of protection. In particular, the invention also claims the subject matter and features of the dependent claims independent of the claims cited in each case, and in particular also without the features of the independent claims. Therefore, the invention comprises different aspects of the invention that enjoy protection independently of each other.

[0138] Reference numerals list

[0139] 1 Machine

[0140] 2 Controller

[0141] 3 Analysis software

[0142] 4 Analyzing Hardware

[0143] 5 Analysis software

[0144] 6 Visualization Device

[0145] 7 Hardware

[0146] 8 Data storage

[0147] 9 Gateway

[0148] 10 Feature Extraction Module

[0149] 11 Rule Creation Module

[0150] 12 Rule Application Module

[0151] 13 Historical raw data

[0152] 14 Tags

[0153] 15 Features

[0154] 16 New raw data

[0155] 17 Machine Status

[0156] d DYN Adjustment phase

[0157] d STAT Adjusted stage

[0158] [T A , T E ] Observation time window

[0159] [t1, t2] comparison period

Claims

1. A monitoring method (1) for a coating device for applying a coating agent, the monitoring method comprising the following steps: a) determining first raw sensor data (16) representing operating variables of the application device (1), and b) obtaining a first control signal for controlling the coating device (1), wherein the first control signal is a target value for controlling the coating device or is used to set other operating variables of the coating device, Characterized in that the monitoring method comprises the following steps: c) extracting features (15) from the first raw sensor data (16), the features (15) comprising a reduced amount of data compared to the first raw sensor data (16) and serving as a data basis for a machine learning algorithm, with the aid of which the features (15) of the currently determined first raw sensor data (16) are evaluated for determining an operating state of the application device (1), wherein extracting the features (15) comprises the following steps: c1) defining an observation time window ([T A , T E ]), in the observation period window ([T A , T E ]) within a period of time when the first control signal changes, c2) defining the observation period window ([T A , T E ]) within at least one comparison period ([t1, t2]), wherein the at least one comparison period ([t1, t2]) is determined between two consecutive changes of the first control signal, c3) Subdividing the comparison period ([t1, t2]) into a plurality of successive sub-segments (d DYN d STAT , d REST ),as well as c4) is calculated for each sub-segment (d DYN d STAT d REST ), wherein at least one statistical parameter of the first raw sensor data (16) within the sensor data is a component of the feature (15).

2. The monitoring method according to claim 1, characterized in that: The painting device is a painting device for painting motor vehicle body parts.

3. The monitoring method according to claim 1, characterized in that: The monitoring method comprises the following steps: a) determining second raw sensor data (16) which reflect operating variables of the application device (1) which differ from the first raw sensor data (16), b) extracting a feature (15) from the second raw sensor data (16), wherein the feature (15) comprises a reduced amount of data compared to the second raw sensor data (16) and serves as a data basis for a machine learning algorithm, and analyzing the feature (15) of the second raw sensor data (16) with the aid of the machine learning algorithm for determining the operating state of the application device (1).

4. The monitoring method according to claim 3, characterized in that: The monitoring method comprises the following steps for extracting features (15) from the first raw sensor data and the second raw sensor data (16): a) defining an observation time window ([T A , T E ]), b) defining the observation period window ([T A , T E ]) within at least one comparison period ([t1, t2])， c) subdividing the comparison period ([t1, t2]) into a plurality of successive sub-segments (d DYN d STAT d REST ),as well as d) Calculate for each sub-segment (d DYN d STAT d REST ), wherein at least one statistical parameter of the first and second raw sensor data (16) within the sensor data (1), the statistical parameter being a component of the feature (15).

5. The monitoring method according to claim 4, characterized in that: The monitoring method comprises the following steps: a) detecting a second control signal for controlling the application device (1), b) determining the time of the change of the first control signal within the observation period window, c) determining the time of the change of the second control signal within the observation period window, d) Determining the at least one comparison period between two consecutive changes of the two control signals.

6. The monitoring method according to claim 5, characterized in that: The monitoring method comprises the following steps: a) determining an amount of change of the first control signal and / or the second control signal, b) determining a time period since a last change of the first control signal and / or the second control signal, c) where the amount of change and the time period since the last change are part of the features (15).

7. The monitoring method according to any one of claims 4 to 6, characterized in that: The comparison period is divided into at least one of the following subsections: a) zero, one or more initial adjustment phases (d) DYN ), in the initial adjustment phase, the first raw sensor data responds to the control signal step, b) zero, one or more temporally subsequent conditioned phases (d STAT ), the first raw sensor data at least partially responds to the control signal step during the conditioned phase, and c) the subsequent remaining phase in time (d REST ), which lasts until the end of the comparison period.

8. The monitoring method according to claim 7, characterized in that: The monitoring method comprises the following steps: a) determining a time constant of the first raw sensor data (16), and b) determining at least one of the following quantities based on the time constant of the first raw sensor data (16): b1) Adjustment phase in the comparison period (d DYN ), b2) Duration of each adjustment phase (d DYN ), b3) Duration of each regulated phase (d STAT ).

9. The monitoring method according to claim 7, characterized in that: Each adjustment stage (d DYN ) and / or each regulated stage (d STAT )'s duration is determined randomly.

10. The monitoring method according to any one of claims 4 to 6, characterized in that: The observation time window ([T A , T E ]) is one of the following time periods: a) the time period associated with the component, or b) the time period related to the application equipment, c) a time period related to time.

11. The monitoring method according to claim 10, characterized in that: The time period associated with a component is a1) a preparation period of the application equipment, which is used to coat the components for subsequent coating, a2) The coating period of the component is the body cycle.

12. The monitoring method according to claim 10, characterized in that: The time period associated with the application equipment is b1) the duration of the setting operation of the application device (1), b2) the duration of the test operation of the application device (1), b3) the duration of manual operation of the application device (1), b4) the duration of the maintenance operation on the application equipment (1), b5) the duration of the production operation of the application device (1), b6) Time periods for non-component-related automated secondary processes of the equipment (e.g. time-dependent flushing, automatic testing or diagnostic operations, in which system components are also specifically controlled.

13. The monitoring method according to claim 10, characterized in that: The time period associated with time is c1) one hour, one day, one week, one month, one quarter, one year, or c2) The length of the work shift.

14. The monitoring method according to any one of claims 4 to 6, characterized in that: The statistical parameter is one of the following parameters: a) considering only the univariate statistical parameters of the first raw sensor data (16), b) A multivariate statistical parameter which, in addition to the first raw sensor data (16), also takes into account the second raw sensor data (16) and optionally further raw sensor data (16).

15. The monitoring method according to claim 14, characterized in that: The univariate statistical parameters are a1) Arithmetic mean, a2) Geometric mean, a3) median value, a4) Variance, a5) Maximum value, a6)Minimum value.

16. The monitoring method according to claim 14, characterized in that: The multivariate statistical parameter is b1) Pearson correlation coefficient, b2) Rank correlation coefficient.

17. The monitoring method according to any one of claims 4 to 6, characterized in that: The first raw sensor data (16) and / or the second raw sensor data (16) and / or the further raw sensor data (16) reproduce one of the following operating variables of the application device (1), and / or the first control signal and / or the second control signal control one of the following operating variables of the application device (1): a) The speed of the turbine of the rotary atomizer, b) the air pressure of the driving air for driving the turbine of the rotary atomizer, c) coating agent pressure at the coating pressure regulator, d) Filling current of electrostatic coating filling system, e) Filling voltage of electrostatic coating filling system, f) humidity in the coating booth, g) the air pressure of the shaping air used to form the coating agent spray, h) the mass flow rate of the shaping air used to form the coating agent spray, i) air temperature in the coating booth, j) the position of the coating impact point of the application device, k) the speed of the coating impact point of the coating application device, l) valve position of the valve, m) the drive variable of the drive, n) flow rate of the coating agent pump or metering unit, o) a position of a linear conveyor for conveying the part to be coated past the applicator (1), p) flow rate, temperature and pressure of the material, q) The rotation speed of the swirl applicator.

18. The monitoring method according to claim 17, characterized in that: The valve is a coating agent valve, a flushing agent valve, a pulse air valve or a lubricant valve.

19. The monitoring method according to claim 17, characterized in that: The drive is a robotic drive.

20. The monitoring method according to claim 17, characterized in that: The drive variables are position, velocity, acceleration, current, voltage, power, temperature.

21. The monitoring method according to any one of claims 3 to 6, characterized in that: The first raw sensor data (16) and / or the second raw sensor data (16) reflect an operating variable of one of the following components of the application device (1): a) Motor, b) coating robot joints of the robot, c) a drive controller for a robot drive of a coating robot, d) a turbine for driving a rotary atomizer, e) a shaping air controller for controlling the flow of shaping air used to form the atomizer spray, f) Pumps, g) a metering device for metering the coating agent, h) Valves, i) Paint pressure regulator, j) High voltage generator for electrostatic coating agent filling, k) Switches, l) Sensors, m) heater, n) control system, o) electrical fuses, p) Batteries, q) Uninterruptible power supply, r) Uninterrupted signal transmission, s) Transformer, t) Fluid pipelines for transmitting fluids, u) Swirl applicator.

22. The monitoring method according to claim 21, characterized in that: The pump is a coating agent pump.

23. The monitoring method according to claim 21, characterized in that: The valve is a proportional valve.

24. The monitoring method according to claim 21, characterized in that: The fluid is a coating agent, a thickener, an adhesive, a diluent, air or water.

25. The monitoring method according to claim 21, characterized in that: The fluid line is a hose or a pipe.

26. The monitoring method according to any one of claims 1 to 6, characterized in that: The monitoring method comprises the following steps for determining first raw sensor data: a) measuring first raw sensor data (16) by means of a first sensor, or b) Reading first raw sensor data previously measured and stored from a database (16).

27. The monitoring method according to any one of claims 1 to 6, It is characterized in that The monitoring method comprises the following steps: A rule for subsequent analysis and processing of the first raw sensor data is established, wherein the rule is obtained by the following steps: a) Retrieve stored historical raw sensor data (13), b) determining or retrieving a stored tag (14) temporally associated with the historical raw sensor data (13), said tag (14) reflecting the state of the application device (1) when the historical raw sensor data (13) was measured, c) extracting features (15) from the historical raw sensor data (13), and d) Determining the rules by means of a machine learning algorithm by analyzing features (15) determined from historical raw sensor data (13) and associated labels (14).

28. The monitoring method according to any one of claims 1 to 6, It is characterized in that The monitoring method comprises the following steps: A rule is established for subsequent evaluation of the first raw sensor data, wherein the rule is determined by the following steps: a) retrieve stored historical raw sensor data, b) extracting features from historical raw sensor data, c) Determining the rules by means of an unsupervised machine learning algorithm by evaluating features determined from historical raw sensor data without associated labels.

29. The monitoring method according to any one of claims 1 to 6, It is characterized in that The monitoring method comprises the following steps: A rule is created for subsequent evaluation of the first raw sensor data, wherein the rule is determined by the following steps: a) retrieve stored historical raw sensor data, b) extracting features from historical raw sensor data, c) Determining the rules with the aid of human experts by analyzing features determined from historical raw sensor data without associated labels.

30. The monitoring method according to any one of claims 3 to 6, characterized in that: The monitoring method comprises the following steps: Features (15) of currently determined first raw sensor data (16) and / or second raw sensor data (16) are analyzed with the aid of a machine learning algorithm to determine an operating state of the application device (1), the operating state being a fault state of the application device (1).

31. The monitoring method according to claim 30, characterized in that: The operating state is one of the following operating states of the application device (1): a) Wear or defect of paint pressure regulator, b) wear or defects in the mixer which mixes the various components of the coating agent with one another, c) wear or defect of the pump, d) wear or defect of valve, e) wear or defect of heater, f) Wear or defect of the drive motor, g) Electrical contact failure, h) detect air entrapment in the paint or interruption of the paint supply, i) the properties of the operating medium of the application device (1), j) Analytical processing of stress application and motion procedures on machine parts or equipment, k) Check the jetting of the bell, l) detect contamination and / or moisture on the atomizer, m) Detect anomalies in the operating behavior of motors, pumps, pistons, shaping air, turbines, high pressure, flow regulators, n) general detection of anomalies in the sense of significant deviations of the signal curve from the normal curve, o) Predict maintenance intervals.

32. The monitoring method according to claim 31, characterized in that: The pump is a coating agent pump.

33. The monitoring method according to claim 31, characterized in that: The valve is a valve for controlling paint, thick substances, adhesives, thinners, air or water.

34. The monitoring method according to claim 31, characterized in that: The operating medium is air, water, paint, adhesive or thick substance.

35. A coating device (1) for coating a component, comprising: a) at least one applicator for applying a coating agent, b) at least one robot for moving the applicator, c) at least one sensor for measuring first raw sensor data representing an operating variable of the application device (1), and d) a control device (2-5) for controlling the applicator (1), said control device (2-5) being arranged in a single component or distributed over a plurality of components, It is characterized in that e) The control device (2-5) or a separate monitoring device can perform the monitoring method according to any one of claims 1 to 34.

36. The coating device according to claim 35, characterized in that The painting device is a painting device for painting motor vehicle body parts.

37. The coating device according to claim 35, characterized in that The manipulator is a multi-axis coating robot.

Images