CPC G06F 16/2477 (2019.01) [G05B 19/4183 (2013.01); G06F 16/2379 (2019.01); G06F 16/24556 (2019.01)] | 12 Claims |
1. A computer-implemented method for obtaining parameters to be used by a control module of a computer, wherein the control module is adapted to control technical equipment by processing batch-run data from the technical equipment, the method comprising:
receiving, from technical equipment that has performed a production process, a reference plurality of multi-variate reference time series, wherein the multi-variate reference time series are based on time points of a batch-specific batch-run time interval and comprises data values from sources that are related to the technical equipment, wherein each multi-variate reference time series comprises:
a first time series from a first source, with data for first measurement values,
a second time series from a second source, with data for second measurement values, and
a third time series from a third source, with data describing particular manufacturing operations during the batch-specific batch-run time interval;
splitting, for each multi-variate reference time series of the plurality separately, the batch-specific batch-run time interval into a plurality of phases, the phases being consecutive time intervals, by determining transitions between the particular manufacturing operations, and dividing the first time series into particular phase-specific first partial series and the second time series into particular phase-specific second partial series;
processing, for each phase separately, and for the phase-specific partial series in combination, the phase-specific partial series to differentiate the plurality-related phase-specific time series into relevant partial time series or non-relevant partial time series according to relevance rules;
storing relevance information as the parameters, wherein storing the relevance information is executed by providing a parameter matrix with binary elements that selectively allow or block the control module of the computer to process the batch-run data from the technical equipment;
receiving a plurality of further multi-variate time series, from batch-runs for which the quality indicators are already available;
providing a set of parameter variations of the parameter matrix by toggling parameters that allow to elements that block;
obtaining for each parameter variation, and for each further multi-variate time series, quality indicators by processing the multi-variate time series; and
determining differences of the obtained quality indicators and taking the variation for which the differences are minimal as selected parameters.
|