KR-20260062474-A - Multi-channel lost data restoration method using AI

KR20260062474AKR 20260062474 AKR20260062474 AKR 20260062474AKR-20260062474-A

Abstract

The present invention relates to a method for restoring lost multi-channel data using artificial intelligence. The present invention relates to a technology that, even if data in any channel among the multiple channels constituting multi-channel data is lost, can restore the lost data to a state close to that of data without loss by using an artificial intelligence model.

Inventors

서호건
이재준
유용균

Assignees

한국원자력연구원

Dates

Publication Date: 20260507
Application Date: 20241029

Claims (11)

A multi-channel lost data recovery method using artificial intelligence in which each step is performed by a computational processing means, An initial processing step (S100) of inputting multi-channel data into a pre-trained AI-based restoration model and receiving restored data for each channel; A repetitive processing step (S200) of re-inputting the data output by the above restoration model into the above restoration model and receiving the restored data for each channel again; A group storage step (S300) for storing restored data for each channel output by each restoration count according to the above restoration model as a single group; A change calculation step (S400) that calculates a restoration change value for the data input and output to the restoration model for each channel based on each restoration count, using the data input by the initial processing step (S100) and the data stored for each restoration count in the group storage step (S300); A determination step (S500) for determining whether the result from the initial processing step (S100) to the change calculation step (S400) corresponds to a preset termination condition; and A final restoration step (S600) in which, depending on the result of the judgment step (S500) above, if the termination condition is met, the data input into the restoration model in the last restoration cycle is set as the final restoration signal; A multi-channel lost data recovery method using artificial intelligence, including
In Article 1, The above change calculation step (S400) is A multi-channel lost data recovery method using artificial intelligence, which calculates a recovery change value for each channel regarding the data input to and output to the recovery model for each channel based on the number of recovery attempts, using a preset data change evaluation method.
In Paragraph 2, The above change calculation step (S400) is An average calculation step (S410) for calculating the average of the calculated restoration change values for each restoration iteration; A multi-channel lost data recovery method using artificial intelligence, further including
In Paragraph 3, The above judgment step (S500) is A first determination step (S510) for determining whether the number of restorations by the above restoration model is greater than or equal to a preset maximum number of restorations; Includes, A multi-channel lost data recovery method using artificial intelligence, wherein if the result of the judgment in the first judgment step (S510) is greater than or equal to a preset maximum number of recovery attempts, it is determined that the above termination condition applies.
In Paragraph 4, The above judgment step (S500) is If the result of the judgment by the first judgment step (S510) is less than the preset maximum number of restorations, a second judgment step (S520) determines whether it corresponds to a preset restoration stop condition by using the average value for each number of restorations calculated by the average calculation step (S410); Includes more, A multi-channel lost data recovery method using artificial intelligence, wherein if the result of the judgment in the second judgment step (S520) above corresponds to a preset recovery stop condition, it is determined that it corresponds to the termination condition.
In Paragraph 5, The above restoration suspension conditions are A multi-channel lost data recovery method using artificial intelligence, wherein the value of the average calculation step (S410) for the last recovery cycle is compared with the value of the average calculation step (S410) for the cycle immediately preceding the last recovery cycle, and the value of the average calculation step (S410) for the last recovery cycle is greater.
In Paragraph 5, The above-mentioned multi-channel lost data recovery method using artificial intelligence Based on the result of the judgment step (S500) above, if the termination condition is not met, the restoration number is increased by +1, and A data synthesis step (S700) for synthesizing the data input into the above-mentioned restoration model and the data output by the above-mentioned restoration model in the very last restoration cycle; Includes more, A multi-channel lost data recovery method using artificial intelligence, wherein the data synthesized by the above data synthesis step (S700) is replaced with the data re-entered into the recovery model by the above iteration processing step (S200) to perform the operation.
In Paragraph 4, The above judgment step (S500) is If the result of the judgment by the first judgment step (S510) is less than the preset maximum number of restorations, a third judgment step (S530) determines whether it corresponds to a preset restoration stop condition by using the restoration change value calculated for each restoration count and each channel calculated by the change calculation step (S400); Includes more, A method for recovering multi-channel lost data using artificial intelligence, wherein if the result of the judgment in the third judgment step (S530) above corresponds to a preset condition for stopping recovery, the recovery status for the corresponding channel is set to False, and the corresponding channel is determined to correspond to the termination condition.
In Paragraph 8, The above final restoration step (S600) is A method for recovering multi-channel lost data using artificial intelligence, wherein if the result of the judgment in the third judgment step (S530) above determines whether the recovery of all channels based on the multi-channel data is set to False, the data input to the recovery model in the last recovery cycle is set as the final recovery signal.
In Paragraph 8, The above restoration suspension conditions are A multi-channel lost data recovery method using artificial intelligence, wherein the recovery change value of a corresponding channel calculated by the change calculation step (S400) for the last recovery cycle is compared with the recovery change value of a corresponding channel calculated by the change calculation step (S400) for the cycle immediately preceding the last recovery cycle, and the recovery change value of a corresponding channel calculated by the change calculation step (S400) for the last recovery cycle is greater.
In Paragraph 8, The above-mentioned multi-channel lost data recovery method using artificial intelligence If, as a result of the judgment in the third judgment step (S530) above, the restoration status for all channels based on the multi-channel data is not set to False, the restoration number is increased by +1, and A data synthesis step (S800) for each channel where the restoration status is not False, synthesizing the data input to the restoration model in the last restoration cycle and the data output by the restoration model; Includes more, A multi-channel lost data recovery method using artificial intelligence, wherein the data synthesized by the above data synthesis step (S700) is replaced with the data re-entered into the recovery model by the above iteration processing step (S200) to perform the operation.

Description

Multi-channel lost data restoration method using AI The present invention relates to a method for restoring multi-channel lost data using artificial intelligence, and more specifically, to a method for restoring multi-channel lost data using artificial intelligence that can restore the corresponding lost data when there is data loss in at least one channel of multi-channel data by utilizing a recursive utilization technique of an artificial intelligence model. For any system operating based on multi-channel data (including single values and continuous signals), it is common practice to design it under the premise that the input multi-channel data is in a normal state—that is, under a normal multi-channel input situation—in order to ensure proper operation. Even when utilizing artificial intelligence models for this purpose, model training is generally conducted under the condition that all channels are valid data, and the process of utilizing the trained model also operates under the condition that all multi-channel data input to the model is valid. In other words, for both systems and artificial intelligence models that utilize multi-channel data as input, normal operation is guaranteed only when data from all channels is valid. In other words, there are stability and reliability issues where it is difficult to expect normal operation of the system or artificial intelligence model if lost data is input due to an invalid state (addition of noise, significant amplification or attenuation, loss, etc.) occurring for any reason in at least one channel of the multi-channel data. In particular, for artificial intelligence models trained on the premise that all channels are valid, there is a risk of producing significantly distorted output if data lost in some channels is input. Accordingly, restoring data from lost channels among multichannel data to maximize the validity of data from all channels is an essential element for reasonably expecting the utility of systems or models operating based on multichannel data. In this regard, Korean published patent No. 10-2023-0171704 ("System and method for diagnosing anomalies in multivariate data based on an autoencoder model") discloses a technology for diagnosing anomalies in multivariate data by training an autoencoder model, restoring data using representative values in the latent space of the autoencoder model, and calculating the loss with respect to test data as an outlier. Figure 1 is an example diagram showing the output result when data for all channels of the multi-channel data is valid in an artificial intelligence model that performs operations based on input multi-channel data. Figure 2 is an example diagram showing the output result in the case where data for at least one channel among the multi-channel data is lost in an artificial intelligence model that performs operations according to input multi-channel data. FIG. 3 is an example diagram showing the output result applied after restoring lost data for at least one channel among the multi-channel data in an artificial intelligence model that performs operations according to input multi-channel data through a multi-channel lost data restoration method using artificial intelligence according to an embodiment of the present invention. FIG. 4 is an example diagram showing data input to a restoration model and data output by a multi-channel lost data restoration method using artificial intelligence according to an embodiment of the present invention. FIGS. 5 and 6 are flowcharts illustrating a method for recovering multi-channel lost data using artificial intelligence according to each embodiment of the present invention. Hereinafter, a multi-channel lost data recovery method using artificial intelligence according to the present invention, having the configuration as described above, will be explained in detail with reference to the attached drawings. The drawings presented below are provided as examples to ensure that the concept of the present invention is sufficiently conveyed to those skilled in the art. Accordingly, the present invention is not limited to the drawings presented below and may be embodied in other forms. In addition, throughout the specification, the same reference numerals indicate the same components. Unless otherwise defined, technical and scientific terms used herein have the meaning commonly understood by those skilled in the art to which this invention pertains, and descriptions of known functions and configurations that could unnecessarily obscure the essence of the invention are omitted in the following description and accompanying drawings. Furthermore, a system refers to a set of components, including devices, mechanisms, and means, that are organized and interact regularly to perform necessary functions. As described above, since a system or artificial intelligence model that utilizes multi-channel data as input is designed on the premise that the data of all channels constitu