METHOD FOR TESTING A PRODUCTION PROCESS FOR MANUFACTURING COMPONENTS

Description

[0001] The present invention relates to a method for controlling a production process, wherein components are manufactured by the production process.

[0002] During the production of series components, various quality and process measurements are carried out regularly. These measurements are usually performed randomly. For example, a specific characteristic (e.g., the weight of a sintered component) can be recorded at regular intervals on a specific number of components. Further measurements relate, for example, to other characteristics of the component, such as length, density, parallelism of surfaces, or diameter of the component. Alternatively or additionally, process parameters can also be recorded at regular intervals, such as the pressure curve on a machine during the pressing process in the production of so-called green compacts.

[0003] A purely static determination of inspection intervals fails to address the constantly changing risks that arise in a real process. In stable processes where components are produced with consistently high quality, too many measurements are usually performed. This results in avoidable costs. In unstable processes, however, there is a risk that defects may not be adequately detected due to too few measurements.

[0004] A method for operating a measuring device is known from DE 10 2016 215624 B3.

[0005] EP 2 993 541 A1 is directed to a method for assessing the quality of a component produced by means of an additive manufacturing process.

[0006] EP 0 826 973 A2 is directed to a method for testing a large number of products.

[0007] An automatic optical inspection system and an operating method are known from US 2018 / 276811 A1.

[0008] Based on this, the object of the present invention is to at least alleviate or even solve the problems described with reference to the prior art. In particular, a method for controlling a production process for manufacturing components is to be provided, in which the testing effort is dependent on the stability of the production process.

[0009] To achieve these objects, a method according to the features of patent claim 1 is proposed. Advantageous further developments are the subject of the dependent patent claims. The features listed individually in the patent claims can be combined with one another in a technologically expedient manner and can be supplemented by explanatory facts from the description, whereby further examples of the invention are shown.

[0010] A method for controlling a production process for components, i.e., for manufacturing components, is proposed. At least the components or a manufacturing facility used to manufacture the components have a plurality of measurable characteristics (e.g., a volume, a weight, a density, a parallelism of surfaces, or a diameter of the component; a temperature, a pressure, a power consumption of the manufacturing facility). At least one characteristic is defined as a primary characteristic, and at least one further characteristic is defined as a secondary characteristic. The method comprises at least the following steps: a) Defining a test plan for recording the at least one main characteristic and the at least one secondary characteristic by tests (e.g. by a measurement), wherein at least the at least one main characteristic is measured with (a first value) a first test frequency and the at least one secondary characteristic is measured with a second test frequency; wherein at least one stability criterion is defined for the at least one main characteristic; b) Manufacturing the components and parallel implementation of the test plan to generate test results, wherein (temporarily) exclusively the at least one main characteristic is tested at the first test frequency; c) Evaluating the determined test results; wherein, if at least one test result for the at least one main characteristic violates the stability criterion, d) Continuing implementation of the test plan, wherein at least the at least one secondary characteristic is also tested.

[0011] The present proposal is to utilize correlations between the various characteristics of a component and / or a production facility. For this purpose, a single quality and / or process characteristic (or a small group) is defined as the main characteristic. This main characteristic can be measured, in particular, automatically and preferably for each produced component. The remaining characteristics (secondary characteristics) are measured with a significantly lower test frequency (i.e., not for all components); ideally, rarely or not at all.

[0012] In particular, only one main characteristic is defined and included in the test plan.

[0013] From the temporal measured value history of the main characteristic (or a small group of main characteristics) and from correlations between all characteristics, predictions regarding the quality status of a measured component can be derived. If changes in the distribution of the main characteristic(s) lead to an increased probability of a possible tolerance violation in one or more of the other characteristics of the component, additional measurements (and, if necessary, process corrections) are initiated (automatically) for these characteristics. Process stability is monitored exclusively using the main characteristic.

[0014] In particular, the use of IoT (Internet of Things) technologies makes it possible in modern manufacturing facilities to automatically record and store specific characteristics and their test results for each individual component produced. These characteristics can correspond to either quality or process characteristics. Examples include a weight (of the component) or a press force (of the manufacturing facility). This results in a significantly higher measurement rate (e.g., 15 measurements per minute, corresponding to a production speed of 15 strokes) compared to a traditional test plan (e.g., 5 measurements every two hours). This makes it possible, in particular, to record trends for characteristics across preferably 100% of all produced components, without the need for manual ("expensive") measurements of the characteristics (e.g., by removing the components and performing a measurement in a laboratory outside the manufacturing facility, or similar).

[0015] A production facility can, for example, comprise a press, a sintering furnace, or a complete production plant with a press and a sintering furnace. Furthermore, a production facility can also comprise, for example, a hall in which a press and / or a sintering furnace are located. Characteristics that can also be recorded include, for example, the ambient air temperature, air humidity, or air composition (proportion of protective gas / oxygen, etc.).

[0016] In particular, it can be assumed that a minimum number of key characteristics exists that directly correlate with the quality of the produced components. Furthermore, it can be assumed that if this minimum number of key characteristics is in a stable equilibrium (e.g., no negative trend is apparent), no further tests are necessary to monitor quality.

[0017] If trend changes of main characteristics (e.g. violation of limit values, drifting downwards or upwards, oscillating within the limit values, atypical constellation of characteristics, dispersion, ...) lead to an increased probability of a tolerance violation (or a violation of the stability criterion) with regard to one or more characteristics (one could also say: if a negative prognosis is made), additional measurements should (automatically) be initiated for the affected characteristics.

[0018] The characteristics concerned are in particular the characteristics covered by the forecast, whereby the forecast is created by observing at least one main characteristic.

[0019] Any necessary process corrections can be made, in particular, based on the overall test results (i.e., both primary and secondary characteristics). This can include not only the test results from current component production, but also historical data from previous production runs.

[0020] A production process involves the manufacture of a component, e.g., the sintering of a component. This involves, for example, providing a powdered material, pressing it into a green compact, and then subjecting it to a sintering (i.e., thermal) treatment. Machining may be performed if necessary. Within the scope of the production process, a large number of components (e.g., at least 100, 1,000, or several tens of thousands) are manufactured, ideally identical to one another.

[0021] A main characteristic to be recorded could be, for example, the weight of a component produced using sintering technology. Another main characteristic or secondary characteristic could be, for example, a length, a density, a parallelism of surfaces, or a diameter of the component. Furthermore, a burr or the appearance of a component's surface, e.g., matte or glossy, etc., can also be defined as a feature. Likewise, a feature of a production facility, e.g., an applied pressure, e.g., pressing force (as a main characteristic) or a power consumption (e.g., as a secondary characteristic), can be recorded.

[0022] A first value of the first test frequency can, for example, represent a baseline value for a newly started production process for the component(s). This value can, for example, comprise a test on one component every 1,000 components or a test of a sample of several components (e.g., five components manufactured directly one after the other) every 1,000 components, or something similar. Alternatively or simultaneously, the first test frequency can be set depending on a period of the production process. In this case, for example, a characteristic of a production facility can be recorded, or a test (on one or more components) can be carried out every hour of the production process. If necessary, the number of components that were produced during the period or that underwent a specified processing step can be taken into account.

[0023] The test results can be determined from the assessment of a single value (the test result of the characteristic being tested) or from multiple values. The evaluation or assessment of multiple values ​​can be based on mathematical methods, e.g., averaging and / or the standard deviation within a sample.

[0024] A stability criterion can be defined as a defined deviation from a target value for the respective characteristic. In particular, the stability criterion does not have to correspond to a permissible tolerance for the target value (a value outside the tolerance would mean, for example, that the component is not OK, i.e., defective). In particular, the tolerance is broader than the stability criterion. Accordingly, a component would not be defective if the first characteristic violates the stability criterion (and thus may still be within the tolerance).

[0025] In particular, it may also be provided that a violation of the stability criterion depends on a method used, for example, to evaluate a sample of components checked during a test.

[0026] For example, a number can be defined as a subset of the components in a sample (e.g., the sample comprises five components) above which a violation of the stability criterion is detected. Therefore, a minimum number of components in a sample (e.g., at least two components in a sample of five components) must exhibit a certain minimum deviation from a nominal value of the first characteristic, at which point a violation of the stability criterion is detected.

[0027] Violation of the stability criterion makes process instability particularly apparent, so that a possible impending non-compliance with the tolerance can be prevented by corrective intervention in the production process.

[0028] The stability criterion can, for example, be based on a measure of change in a test result. In particular, if a trend or a sudden, possibly sporadic, change in the test results emerges, a violation of the stability criterion can be determined.

[0029] In particular, the production of the components and the implementation of the test plan, i.e. the performance of measurements of the first characteristic, are carried out in parallel.

[0030] In particular, the first test frequency comprises a test of at least 50% of the components (i.e. the relevant feature of the component or the manufacturing facility is checked for at least 50% of all components), preferably at least 75%, particularly preferably 100%.

[0031] In particular, the second test frequency deviates from the first test frequency, wherein the second test frequency is lower or even significantly lower than the first test frequency. Preferably, the second test frequency comprises, in particular, a maximum of 50% (preferably a maximum of 20%, particularly preferably a maximum of 5%) of a number of tests performed as a result of the first test frequency being performed. In particular, only the at least one main characteristic is temporarily tested, while the at least one secondary characteristic is tested at a significantly lower second test frequency (continuously and thereby superimposing the first test frequency).

[0032] Preferably, the second test frequency comprises a test of 0% of the components and a test of at least one secondary characteristic is only started if a stability criterion of the at least one main characteristic has been violated.

[0033] In particular, at least one (preferably only one) main characteristic is recorded during one or more process steps in the first test frequency. If it remains within predefined limits for a predefined period of time (i.e., the stability criterion is not violated), no further tests or measurements are performed. If the limits of the stability criterion are violated, all characteristics are recorded and tested according to the test plan.

[0034] In particular, the individual main characteristic is, for example, the weight of a component. In particular, at least one main characteristic is checked automatically during the production of the component. In the case of the weight of a component, the relevant number of components can be weighed. In the case of pressure / compression force, for example, the pressure applied by the production facility for the relevant number of components can be recorded.

[0035] In particular, changes in the variability or distribution properties of the test results of at least one main characteristic can be used as a trigger for carrying out additional or all measurements according to the test plan (additional).

[0036] In particular, at least a plurality of minor characteristics is present, whereby the violation of a stability criterion of a major characteristic requires the testing of only a selection of the minor characteristics. In particular, not all of the minor characteristics identified in the test plan are tested, but only a selection of the minor characteristics. In particular, further minor characteristics can be tested successively, or the selection can be changed.

[0037] In particular, there are a number of main characteristics (e.g. weight and surface colour of the component), whereby the violation of a stability criterion of a main characteristic (e.g. weight) requires the examination of a first selection of the secondary characteristics (e.g. length, width of the component; pressure of the production facility) and the violation of a stability criterion of another main characteristic (e.g. surface colour) requires the examination of a second selection of the secondary characteristics (e.g. composition of the starting materials for the component, temperature in the production facility, pressure of the production facility).

[0038] In particular, the violation of a stability criterion includes at least one of the following conditions: the test results follow a trend (continuously increasing or decreasing); the test results fluctuate within an interval that exceeds limits defined by the stability criterion; the test results of at least different main characteristics deviate from a predefined correlation of the main characteristics (e.g., weight and volume, or height, length, width; or weight and pressure are interdependent).

[0039] In particular, if test results of the main characteristic(s) change (e.g., if a trend is present, if there is a sudden change, or if the intervals are too long) a review of additional (or all) characteristics specified in the test plan can be initiated. In particular, the relevant test frequencies will only be increased for those characteristics whose test results demonstrate the specified properties.

[0040] In particular, in the case of changing test results (e.g., in the case of a trend, a sudden change, or excessively large intervals) of the main characteristic(s), only a selection of characteristics (i.e., not all characteristics from the test plan) can be tested. The selection of characteristics can be determined, in particular, by experts. If necessary, the selection can be made anew each time or, alternatively, stored in the test plan as a fixed selection.

[0041] At the beginning of a component manufacturing process, for example, the weight of the produced components could be continuously recorded and considered a key characteristic. In a subsequent step, additional key characteristics could be added or replaced by the previously used key characteristics. Possible additional key characteristics could include, for example, a pressing force or a pressure applied to the component by the manufacturing equipment, or even individual dimensions of the produced components.

[0042] In particular, at least the first test frequency is changed, whereby (either) the first test frequency is increased if at least one test result for the at least one main characteristic violates the stability criterion; or the first test frequency is reduced if a certain number of consecutive test results for at least the at least one main characteristic are in accordance with the stability criterion.

[0043] In particular, the first test frequency (which is assigned to the at least one main characteristic or all main characteristics) is changed, i.e., the first value of the first test frequency is set to a different second value of the first test frequency. The first test frequency is increased if, for example, at least one test result for the at least one main characteristic violates the stability criterion. On the other hand, the first test frequency is decreased if all test results are in line with the stability criterion.

[0044] It is therefore suggested here that if the production process is stable (and the test results for at least one main characteristic do not show any major deviation, i.e. the test results are in line with the stability criterion), the first test frequency can be reduced.

[0045] If a possible instability of the production process is detected (e.g. at least one test result for at least one main characteristic shows a larger deviation, thus the test result violates the stability criterion), the first test frequency can be increased.

[0046] For example, the expert will use a so-called control limit, a tolerance limit, or a stability limit to define a stability criterion. Thus, if the test results for the characteristic to be tested lie within the defined limits, the test frequency is gradually reduced.

[0047] Conversely, the test frequency is increased as soon as the test results of the characteristic to be tested lie outside defined limits, since this limit violation indicates a higher quality risk for the component.

[0048] Under stable conditions of the production process, measurements of characteristics to be tested can be reduced to a minimum while still guaranteeing a high quality of the produced components.

[0049] In particular, based on computer-aided measurement data acquisition and recording, test results can be continuously analyzed and assessed to determine whether the production process is stable or whether interventions in the production process are necessary.

[0050] Based on the measurements performed and the resulting test results, it is particularly possible to identify a trend in the change of a characteristic (e.g., weight or height). If the test results remain within defined limits over a defined period of time or across a certain number of samples (meeting the stability criterion), it can be concluded that the process for the characteristic being tested is stable, and the test frequency can therefore be reduced. This reduction can also be achieved gradually, and ideally, the test frequency can be set to a defined minimum.

[0051] However, if one or more test results during a test violate at least one stability criterion, this indicates an unstable production process and the test frequency is increased or additional characteristics (main characteristics and / or secondary characteristics) are checked, thus increasing the production process monitoring. At the same time, intervention in the production process can be carried out to restore process stability, including the rejection of suspected defective components. The higher test frequency or the testing of additional characteristics ensures that the production process is monitored more closely for a defined period of time. Only when the test results meet at least one stability criterion over a defined period of time or across a certain number of samples or components can the test frequency and / or the number of monitored characteristics be (gradually) reduced again.

[0052] In particular, at least main characteristics (and possibly additional secondary characteristics) are linked to one another by a model, whereby test results of at least one main characteristic are used to infer a change in other main characteristics or secondary characteristics, taking the model into account. A test of at least one of these main characteristics or secondary characteristics is initiated as soon as a violation of a stability criterion associated with this characteristic is predicted.

[0053] For example, a continuous increase in the weight of a component can be used as the basis for a forecast. Based on this forecast, a future violation of a stability criterion could be predicted. Based on this forecast, tests for additional characteristics could be initiated, allowing early, corrective intervention in a manufacturing process (possibly before a stability criterion is violated).

[0054] In particular, sensitivity and dependency analyses can be performed. Based on these, both the main characteristics and their influence on the remaining (secondary) characteristics can be identified. These analyses are also intended to determine the testing frequency with which the main characteristics must be recorded, whether high-frequency (e.g., 100% of the components) or perhaps only through periodic measurements with fixed time and / or production intervals. In particular, derived main characteristics are provided, which are not directly measurable but can be calculated from several measured characteristics. One example is, for example, the density of a component, calculated from the measured weight and volume (which in turn is derived, for example, from the measured component height).

[0055] In particular, a model is created that can estimate the distribution of the values ​​of the remaining characteristics from the values ​​of the main characteristics. The model is based, in particular, on the sensitivity and dependency analyses mentioned above. Measurements of characteristics are initiated automatically if an increased probability of a predicted tolerance violation or violation of a stability criterion for one or more characteristics is calculated, or if the prediction interval (measures the uncertainty) of an estimated characteristic becomes too large. The test results provided by the automatically initiated measurements are used, in particular, to validate the model and / or improve it online (in real time).

[0056] In particular, the model is continuously validated and modified based on the tests carried out and the test results.

[0057] A validation state of the model, expressed in particular in the binary distinction (grading) "good" or "bad", can be linked to the generally known term model quality in such a way that a model quality determined as not lying within a defined range is equivalent to the state of a violation of a stability criterion.

[0058] Automatically initiated measurements or tests are used not only to validate the model, but also to extend it. For example, measured features can be given greater confidence than those that are only estimated. The features measured as a result of automatic initiation can thus temporarily become key features, which in turn can be used to initiate further measurements or replace existing ones.

[0059] In particular, the model for selecting the main or secondary characteristic to be additionally tested takes into account at least one of the following factors: Costs of the additional testing; time required for the additional testing; availability of the required measuring equipment.

[0060] In particular, when measurements or tests are initiated automatically, the framework conditions of these measurements are (also) modeled in the model, with the goal of always using the most cost-effective measurements. For example, the effort, costs, duration, utilization of the measurement system, measurement reliability, etc. are modeled and taken into account when selecting the measurements.

[0061] In particular, the additional measurements or tests initiated are specifically evaluated. Multiple initiations can, for example, indicate systematic or "incredibly occurring" errors in processes and / or machines / tools. These errors must be identified and subsequently eliminated. Dynamic adjustments to system maintenance intervals based on the frequency of characteristic measurements are also conceivable.

[0062] In particular, the main characteristics and their dependencies (on other characteristics) can be newly identified for each new product or process. However, to reduce this effort, classes of products / processes with similar behavior / requirements can be identified. For this purpose, existing measured values ​​(e.g., from quality, process, and machine data) can be evaluated, for example, based on pattern similarity in the deviations.

[0063] Furthermore, dependencies can be established between these patterns and the component properties to be produced (e.g., size, weight, geometry, density, etc.). By comparing a newly produced component with components already produced, a suitable base model class (which defines the recommended tests) can be suggested for the new component.

[0064] In particular, the basic models to be developed are increasingly generic. Once established, they are particularly easy to adapt to various new products and / or processes. For this purpose, more measurements can be performed at the start of a new production run. The model parameters and the associated confidence intervals can be adjusted, particularly automatically. As soon as a previously defined stability criterion is reached, the model can be activated (and used for the process), with measurements or tests subsequently being performed only as needed.

[0065] The stability criterion could, for example, be calculated from the main features and the confidence intervals of the remaining features. A concrete example of the "good case" (i.e., the case of a functioning model): The main features behave "unobtrusively" and the confidence intervals are within predefined limits for a predefined period of time.

[0066] In particular, the test plan for a component to be manufactured is created at least partially automatically during the manufacture of the component by identifying at least one main characteristic and one secondary characteristic based on test results of characteristics.

[0067] In particular, in order to validate the test plan, in addition to testing the identified at least one main characteristic, a plurality of other characteristics can be tested periodically.

[0068] In particular, in addition to the test results at the end of a production run, the model parameters, the behavior of the parameters and features during the learning process, and the number and type of (automatically) initiated measurements or tests, etc., are saved. This information can be automatically incorporated into the learned model during the next production run, thus gradually improving its performance.

[0069] In particular, based on the evaluation of the (automatically) initiated measurements or tests across all products and processes, the need for measurement automation can be derived.

[0070] In particular, the production of the component comprises a plurality of process steps, wherein a feature is present and tested after a first process step and wherein a further (measurable) feature is present after a second process step following the first process step.

[0071] A data processing device is further proposed, comprising a processor configured to carry out the described method.

[0072] As a precaution, it should be noted that the numerals used here ("first", "second", "third", ...) primarily serve (only) to distinguish between several similar objects, quantities, or processes, and therefore do not necessarily dictate any interdependence and / or sequence of these objects, quantities, or processes. Should a dependence and / or sequence be required, this is explicitly stated here or will be obvious to the person skilled in the art upon studying the specifically described embodiment.

Claims

1. Method for controlling a production process for components, wherein at least the components or a manufacturing device used for producing the components have a multiplicity of metrologically detectable features, at least one feature being defined as a main feature and at least one further feature being defined as a secondary feature; having at least the following steps: a) establishing a test plan for detecting at least one main feature and at least one secondary feature by tests, the at least one main feature being measured with a first test frequency and the at least one secondary feature being measured with a second test frequency; wherein at least one stability criterion is defined for the at least one main feature and the at least one secondary feature; wherein the stability criterion is established as a defined deviation from a setpoint value for the respective feature; wherein a component is not being considered to be defective if a test result of the respective feature violates the stability criterion but is still within a tolerance defined für that feature but is only considered defective once the test result is outside the tolerance; b) producing the components and in parallel carrying out the test plan in order to generate test results, only the at least one main feature being tested with the first test frequency; c) evaluating the test results which have been determined; and, if at least one test result for the at least one main feature violates the stability criterion, d) continuing to carry out the test plan, the at least one secondary feature at least also being tested automatically; wherein by means of the violation of the stability criterion, a process instability of the production process is identified and then the at least one secondary feature is tested and the production process is therefore monitored more closely, wherein in parallel, an intervention in the production process is carried out in order to restore the process stability, so that a possible impending noncompliance with the tolerance is prevented; wherein, if the test results satisfy the stability criterion over a defined period of time or over a particular number of random samples or components, the test frequency for at least one secondary feature is reduced; wherein the second test frequency is lower than the first test frequency.

2. Method as claimed in claim 1, wherein in controlling a production process for components the first test frequency comprises testing of each component.

3. Method as claimed in one of the preceding claims, wherein in controlling a production process for components at least the at least one main feature is tested in an automated fashion during the production of the component.

4. Method as claimed in one of the preceding claims, wherein there are at least a multiplicity of secondary features, the violation of the stability criterion of the at least one main feature entailing the testing only of a selection of the secondary features.

5. Method as claimed in claim 4, wherein there are a multiplicity of main features, the violation of a stability criterion of one main feature entailing the testing of a first selection of the secondary features and the violation of a stability criterion of another main feature entailing the testing of a second selection of the secondary features.

6. Method as claimed in one of the preceding claims, wherein the violation of a stability criterion comprises at least one of the following states: • the test results follow a trend; • the test results fluctuate within an interval which exceeds limit values defined by the stability criterion; • the test results of at least various main features deviate from a predefined correlation of the main features.

7. Method as claimed in one of the preceding claims, wherein at least the first test frequency is varied, • the first test frequency being increased if at least one test result for the at least one main feature violates the stability criterion; or • the first test frequency being reduced if a particular number of successive test results at least of the at least one main feature are in accordance with the stability criterion.

8. Method as claimed in one of the preceding claims, wherein at least main features are linked to one another by a model, a variation of other main features or secondary features being inferred with the aid of test results of at least one main feature by taking the model into account; wherein testing of at least one of these main features or secondary features is initialized as soon as the violation of a stability criterion assigned to this feature is prognosed.

9. Method as claimed in claim 8, wherein the model is validated and varied continuously with the aid of the tests carried out and the test results.

10. Method as claimed in one of the preceding claims 8 and 9, wherein at least one of the following factors is taken into account by the model for the selection of the main feature or secondary feature additionally to be tested: • cost of the additional test; • time expenditure of the additional test; • availability of the required measuring device.

11. Method as claimed in one of the preceding claims, wherein in controlling a production process for components the test plan for a component to be produced is compiled at least partially in an automated fashion during the production of the component by identifying at least one main feature and one secondary feature with the aid of test results of features.

12. Method as claimed in claim 11, wherein in addition to the testing of the at least one main feature identified, a multiplicity of other features are periodically tested in order to validate the test plan.