Master pattern generation method based on control program analysis and training method for cycle analysis model
The present disclosure discloses a master pattern generation method which is a major pattern in a repeated cycle by analyzing programmable logic controller (PLC) logic, and a method for training a model that may analyze an error of a cycle using the generated master pattern. The master pattern generation method and the training method for a cycle analysis model according to the present disclosure are different from the related art in that the methods are a technology of processing a machine control language (low-level language) that is difficult for humans to analyze and converting the machine control language into an analyzable language (high-level language), i.e., a machine language processing (MLP)-based technology that analyzes the executed machine language (a language that controls a machine) with a computer and can be understood by humans.
1. A master pattern generation method from a ladder logic of a programmable logic controller (PLC), the method comprising:
generating, by a processor, a master pattern to be compared with contacts in the ladder logic of the PLC to determine whether each cycle is normal or abnormal, wherein
the generating the master pattern includes:
generating a relationship between the contacts included in the ladder logic;
generating bars for each cycle from log data of the PLC and links for each cycle;
calculating a statistic of the bars and links included in a plurality of cycles; and
generating h master pattern based on the calculated statistic,
the determination of whether each cycle is normal or abnormal includes:
comparing the master pattern with input data to generate output data indicating whether there are errors for each bar and link, and
detecting, with an artificial neural network trained using the input data and the output data, at which contact and link the error occurs with respect to data of a new cycle, and correcting the error, and
the generating of the relationship between the contacts includes:
removing a manual column from the ladder logic and expanding a path by searching for a step in which contact A that maintains an OFF state as usual and then changes to an ON state is used as an output contact of another step.
2. The method of claim 1 , wherein the relationship between the contacts is Include, Includable, Exclude, and Excludable.
3. The method of claim 2 , wherein, in the generating of the relationship between the contacts, when two different contacts each have the same contact as Include and Exclude, the relationship between the two different contacts is generated as Exclude.
4. The method of claim 1 , wherein
the relationship between the contacts is Include, Includable, Exclude, and Excludable,
in the generating of the link, when the relationship between the two contacts is Include or Includable, a link connecting starting points of each bar corresponding to the two contacts is generated, and
when the relationship between the two contacts is Exclude or Excludable, a link connecting an ending point of a FROM bar and a starting point of a TO bar among the two contacts is generated.
5. The method of claim 1 , wherein the generating of the link includes removing a bar or a link whose frequency of occurrence within each cycle is less than a preset minimum occurrence rate.
6. The method of claim 1 , wherein the generating of the link includes removing a link whose duration is outside of a preset duration range.
7. The method of claim 1 , wherein
the statistic of the bars is an average start time of the bars, an average duration of the bars, and a standard deviation of durations of the bars, and
the statistic of the links is an average duration of the links and a standard deviation of durations of the links.
8. The method of claim 1 , wherein the calculating of the statistic includes calculating the statistic of the bars and the statistic of the links included in the plurality of cycles corresponding to conditions established by a user.
9. The method of claim 8 , further comprising, after the generating of the master pattern, extracting and treeizing common elements of the master pattern for each condition.
10. A method of training a cycle analysis model of a programmable logic controller (PLC) to detect an error in the PLC, the method comprising:
generating, by a processor, a master pattern to be compared with contacts in a ladder logic of the PLC to determine whether each cycle is normal or abnormal, wherein the generating the master pattern includes:
generating a relationship between the contacts included in the ladder logic;
generating bars for each cycle from log data of the PLC and links for each cycle;
calculating a statistic of the bars and links included in a plurality of cycles; and
generating the master pattern based on the calculated statistic;
generating input data for start times of all the bars, durations of all the bars, and durations of all the bars included in each cycle by using h log data of h PLC;
comparing the input data with the master pattern to generate output data indicating whether there are errors for each bar and link;
training an artificial neural network with a supervised learning algorithm using the input data and the output data;
inputting data of a new cycle into the trained artificial neural network and detecting at which contact and link the error occurs; and
correcting the error in the PLC based on a detecting result.
11. The method of claim 10 , wherein the start time of each bar in the input data is a positive value for a relative time interval between a start time of each cycle and the start time of each bar.
12. The method of claim 11 , wherein a start time for a bar that does not occur in the input data is a negative value.
13. The method of claim 10 , wherein the output data has a first value when a contact or a link of each cycle is normal compared to the master pattern, and has a second value when the contact or link of each cycle is abnormal compared to the master pattern.