DE102017219549A1 - Method for condition monitoring of a production plant - Google Patents

Abstract

The present invention relates to a method for condition monitoring of a manufacturing plant, which performs reproducible recurring processes, in particular a clock-bound manufacturing plant. In order to provide a method for monitoring the condition of a production plant, which is easy to use and automate, it is proposed according to the invention that the active power P (t) of several consecutive recurring processes is registered as a reference data set P _ref (t), and from the reference data set P _ref ( _FIG. t) automatically at least one parameterized reference data record P _{ref, para} (t) is formed, the course of which is compared with the actual course of the active power P _ist (t) of the current process, so that deviations between the parameterized reference data record P _{ref, para} (t ) and the actual course of the active power P _is (t) allow a conclusion on the state of the manufacturing plant.

Description

The present invention relates to a method for condition monitoring of a manufacturing plant, which performs reproducible recurring processes, in particular a clock-bound manufacturing plant.

Condition monitoring of a manufacturing plant is used to detect wear-related errors during the ongoing process before the worst-case faults lead to factory failure or faulty productions. Depending on the type of production facility, different methods are available for this, which are adapted to the workpiece to be machined and to the expected signs of wear.

For example, by means of a vibration monitoring damage and wear on bearings, gears, shafts or similar components can be determined early and reliable by changing the specific frequency spectrum of each component, including the amplitude and the frequency and the accelerations occurring at the components of the system are measured. Deviations of these measured values from specified target values indicate individual signs of wear and tear.

Another known method for monitoring the condition of a production plant is an oil quality measurement, in which wear particles in the lubricant are detected and the viscosity of the lubricant is determined.

Furthermore, in particular in cutting tools, such as turning tools, cutters, wear phenomena can be determined by a thermography, in which the temperature of the manufacturing plant is measured area-wise, because the machining with worn tools generates increased friction and therefore an elevated temperature.

Finally, it is known to estimate the state from the energy consumption or the power consumption of individual drives of a production plant. This method is based on the consideration that any wear on friction points leads to an increase in the friction and thus to an increase of the torque and the power consumption. For example, this method allows a conclusion on the tool wear of a cutting machine tool, because at constant delivery and cutting speed, the current consumption of the main drive is proportional to the cutting force. This increases with increasing tool wear, which allows you to determine the required tool changes based on the current consumption.

With known methods, the condition monitoring of a production plant via the monitoring of the energy data is possible only with the addition of further data, in particular sensor data, which are correlated with the energy data. This is usually the temperature, speed and / or travel speed of components. For example, in DE 199 03 826 A1 discloses a system for monitoring a manufacturing plant in which different process characteristics are recorded and processed from various sensors.

In a similar way also suggests DE 10 2010 043 651 A1 a method for monitoring a plant, in which in addition to energy consumption values and component-specific status data are collected, which are to be evaluated together for condition monitoring.

A disadvantage of the known method is that the methods are each designed for specific applications and on special manufacturing equipment, which is why the application of the method associated with a considerable effort in the adjustment of required parameters. In addition, an extensive sensor technology is required for known methods, which generates a high workload and therefore high costs both during installation and during maintenance. Furthermore, the known methods can not be readily automated due to the inconsistent and system-specific characteristics, which also increases the workload during installation.

It is therefore an object of the present invention to provide a method for monitoring the condition of a manufacturing plant, which is easy to use and automated. This object is achieved by the method according to claim 1. According to the invention, it is provided that the active power of several consecutive recurrent processes is registered as a reference data record, and from the reference data set at least one parameterized reference data record is automatically formed, the course of which is compared with the actual course of the active power of the current process, so that deviations between the parameterized reference data set and the actual course of the active power allow a conclusion on the state of the manufacturing plant. The active power is the product of the current and the voltage applied to the production plant, so that the method is analogous by the measurement of the current is applicable, provided that a constant voltage can be assumed at the manufacturing plant.

In contrast to the known condition monitoring methods, the method according to the invention allows an automatic condition monitoring, which is based exclusively on the measurement of the active power and / or the current during the ongoing process. Further characteristic data by other sensors, in particular measured values of temperatures, speeds, accelerations or the like can be dispensed with in the context of the method according to the invention. In this respect, the method according to the invention can readily be applied to different production systems without having to be adapted to different sensor arrangements. Thus, a simple automation of the process is given.

Preferred embodiments of the present invention are given below and in the subclaims.

It has already been pointed out that the present invention is characterized in particular by the possibility of automatic parameterization of the reference data record. For this purpose, it is provided according to a first preferred embodiment that based on the reference data set, a data analysis, a segmentation, a classification of the segments, a method selection and parameter calculation and a review of the parameterized reference data set is performed. The individual process steps will be discussed below.

The reference data set is preferably recorded over a long period of time, preferably over 24 hours, so that a number of recurring processes are carried out during this time. In the data analysis of the reference data set, several characteristic values are determined, in particular the number of recurring processes, the number of cycles, the average process or cycle time, the minimum process or cycle time, the minimum and maximum active power, the average difference between two consecutive measured values and / or the measurement resolution, ie the accuracy with which the values are measured. The process or cycle time is the period of time that elapses between the start of a bar and the next bar start. Any transport times of a workpiece (standby time) that enters and exits the production line are already included in the cycle time. This is to be distinguished from the process or cycle time, which represents the time period between the beginning of the bar and the end of the bar. The transport times of the workpieces between two bars are therefore not attributable to the cycle time.

On the basis of the determined characteristic values, the segmentation of the reference data set takes place in segments that are to be assigned to a process or cycle time. Each segment begins with the beginning of a bar and ends with the end of the bar. For example, in the case of cycles lasting 55 s and a transport time of the workpiece of 5 s, the (theoretical) assumption of uninterrupted production gives 1,440 segments for a reference data record with a length of 24 hours.

The segments are then iteratively classified by comparing the segments and assigning different classes in case of significant differences. The significant differences may be due to sometimes different operations performed on workpieces in the same manufacturing plant. How many classes ultimately have to be formed depends, in particular, on the process to be monitored and the predefinable tolerances.

The subsequent parameter calculation is done class by class. This means that for parameter calculation, at least one description method is applied class-wise to all reference data of the segments, so that a parameterized reference data record results per class and description method. A description method is a selectable rule with which the measured values of all segments of a class are linked to one another. By way of example, a preferred description method provides for averaging the values of all the measured values of the segments at one time and determining the standard deviation. In this case, the parameterized reference data record contains a chronological progression of the mean values of all measured values including an envelope of the height of three times the standard deviation.

After the automated parameterization of the reference data set, a check is to be carried out as to whether the parameterized reference data record is applicable and whether it describes the reference data record with sufficient accuracy, for which it is first tested on the original reference data record. For this purpose, it is preferably provided that the parameterized reference data sets are applied to the reference data record for checking, wherein the check is concluded positively if at least one parameterized reference data record corresponds to the course of the reference data record at each clock without significant deviations. Accordingly, it is checked whether each section of the reference data set corresponds to the course of at least one parameterized reference data set, wherein any tolerances are adjustable and can be specified for example by the envelope. If the course of the reference data set is not due to the multiple stringing of parameterized reference data records are to be rejected, the parameterized reference data records must be discarded and the parameterization of the reference data record must be repeated with other specifications or parameters.

As soon as a parameterized reference data record is available that provides a sufficiently accurate description of the reference data record, the current production process can be monitored with the parameterized reference data record by comparing the time profile of the energy data of the current process with the parameterized reference data record. If the energy data show significant deviations over time that can not be found in the data of the reference data set, the deviations indicate signs of wear. The time within a cycle in which the deviations are present provides valuable information on the type of wear and which component of the production plant is affected during the fault diagnosis.

• 1 einen schematischen Aufbau einer Vorrichtung zur Durchführung des Verfahrens und
• 2 einen Verfahrensablauf bei der Parametrierung eines Referenzdatensatzes.

A concrete embodiment of the present invention will be explained below with reference to the drawings. Show it:

• 1 a schematic structure of an apparatus for performing the method and
• 2 a procedure in the parameterization of a reference data set.

1 shows a manufacturing plant 1 within a production line of motor vehicles. The illustrated production plant 1 is a time-bound production plant 1 in which repeatable manufacturing processes are performed. To the state of the manufacturing plant 1 To be able to monitor permanently during operation is the production plant 1 via a measuring line 2 with a microcontroller 3 connected to the energy data of the manufacturing plant 1 initially records and stores as a reference data set over a period of 24 hours at a resolution of one reading per second. The parameterization of the reference data set then takes place via a web server on a suitable computer 4 which sends the parameterized reference data sets to the microcontroller after a successful check 3 returns. The monitoring of the current process in the production takes place again on the microcontroller 3 for permanently storing the measured values to a correspondingly established database 5 passes.

The automatable parameterization of the recorded reference dataset is shown schematically in 2 shown. This is from a reference record P _ref (t) which shows the energy data of a clock-linked production plant as a function of time. The recorded reference data set P _ref (t) is first analyzed and different characteristics are determined. For this purpose, the times are first determined at which a clock starts ( T _on ) and at which a bar ends ( T _bn ), which can be assumed that the energy consumption during production - ie during a cycle - is many times higher than during the standby time between two cycles, in which the workpiece moves out of the production line and retracts a new workpiece. Based on the specified times, the number of clocks that took place during the recording of the reference data set is determined. In the illustrated reference data set P _ref (t) are found taking into account the determined values five clocks, which are each assigned to a segment in the segmentation, so that in the embodiment shown, the segments S ₁ , ..., S ₅ result.

The identified segments are then compared with each other and assigned to different classes in case of significant differences. In the embodiment shown, the segments S ₁ . S ₃ and S ₅ assigned to class A and the segments S2 and S4 form class B. The number of different classes is not defined before the parameterization and depends on how much the energy consumption per segment differs from other segments.

In each class, the measured values of all segments are linked with each other via a suitable description method, so that a resulting time course of the energy data including an envelope results, which forms a kind of tolerance range around the temporal course of the energy data. For each class and description method, a parameterized reference data set is created, where in 2 only one parameterized reference data record is shown per class.

The final review of the identified records reveals whether the segments S ₁ , ..., S ₅ are completely described by at least one of the parameterized reference data records or whether deviations are present. In this case, the parameterization would have to be repeated with other values and possibly with a different description method until a suitable parameterized reference data record is available, so that the original reference data record results from a sequence of different parameterized reference data records. These parameterized reference data records can then be used for the continuous monitoring of the production process.

LIST OF REFERENCE NUMBERS

1

manufacturing plant

2

Measurement line

3

microcontroller

4

computer

5

Database

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 19903826 A1 [0007]
• DE 102010043651 A1 [0008]

Claims (8)

Method for condition monitoring of a production plant (1), which carries out reproducibly recurring processes, in particular a clock-bound production plant (1), characterized in that a) the active power P (t) of several consecutive recurring processes is registered as a reference data set P _ref (t), and b) from the reference data record P _ref (t), at least one parameterized reference data record P _{ref, para} (t) is automatically formed, the course of which is compared with the actual course of the active power P _ist (t) of the current process, so that deviations between the parameterized reference data set P _{ref, para} (t) and the actual course of the active power P _is (t) allow a conclusion on the state of the manufacturing plant. Method according to Claim 1 , characterized in that for the automatic parameterization of the reference data set P _ref (t) a data analysis of the reference data record P _ref (t), a segmentation, a classification of the segments, a process selection and parameter calculation as well as a check of the parameterized reference data set P _{ref, para} (t) is carried out. Method according to Claim 2 , characterized in that in the data analysis of the reference data set P _ref (t) a plurality of characteristic values are determined, in particular - the number N of recurring processes, in particular the number N of cycles, - the average process or cycle time T _Z , - the minimum process - or cycle time T _D , - the minimum and maximum active power P _min and P _max , - the average difference between two consecutive measured values and / or - the measured value resolution. Method according to Claim 2 , characterized in that in the segmentation, the portions of the reference data record P _ref (t) are determined, which are assigned to a process or cycle time T _D and thus a segment. Method according to Claim 4 , characterized in that the determined segments are compared with each other and assigned to different classes in case of significant differences. Method according to Claims 5 , characterized in that at least one description method is used for parameter calculation on all reference data of the segments in a class-specific manner, so that a parameterized reference data record P _{ref, para} (t) results per class and description method. Method according to Claim 6 , characterized in that for carrying out a description method an averaging of all measured values of the segments of a class takes place at a time and the standard deviation is determined, so that the parameterized reference data set P _{ref, para} (t) a time course of the mean values of all measured values and an envelope of contains the amount of three times the standard deviation. Method according to one of Claims 1 to 7 , characterized in that the parameterized reference data sets P _{ref, para} (t) are applied to the reference data record P _ref (t) for verification, wherein the check is completed positively if at least one parameterized reference data set P _{ref, para} (t) the course of Reference data set P _ref (t) at each clock without significant deviations.

Images