Time-series data diagnosing/compressing method
A predicted-failure-evidence diagnosing section for equipment that does not depend on the equipment and does not require knowledge of the equipment is provided. On the basis of a result of a predicted-abnormality-evidence diagnosis carried out by this section on time-series data gathered from the equipment, an allowable error used for compressing the gathered data can be set and managed in order to compress the data if the result of the diagnosis is normal or to restrict the compression of the data during a period in which an evidence of a predicted abnormality is detected. Thus, the amount of data stored in a memory can be reduced.
1. A method implemented by a computer for diagnosing and compressing a time-series data, comprising:
carrying out a predicted abnormality evidence diagnosis on time-series data gathered from one or more sensors provided on equipment;
managing a compression strategy for the data on the basis of a result of the diagnosis and thereby compressing the data with data compressing means;
after compressing the time-series data with data compressing means through execution of the predicted abnormality evidence diagnosis on the data and through the management of a compression strategy for the data on the basis of a result of the diagnosis, decompressing the compressed data;
carry out a predicted abnormality evidence diagnosis on the decompresses data;
comparing a result of the diagnosis carried out on the decompressed data with a result of the diagnosis carried out earlier on the uncompressed data in order to find an error which is the difference between the two results; and
storing the compressed data in a storage if the error is not greater than a reference value, or storing the uncompressed data in the storage if the error is greater than the reference value.
2. The time-series data diagnosing and compressing method according to claim 1 , wherein the degree of diremption of the gathered time-series data from a cluster is used as a result of the predicted abnormality evidence diagnosis, the cluster being classified based on vector quantization by making use of a group of time-series data gathered in advance from the sensors at normal times.
3. The time-series data diagnosing and compressing method according to claim 1 , wherein, in the management of a compression strategy for the data, an allowable error of the data is computed from a result of the predicted abnormality evidence diagnosis.
4. The time-series data diagnosing and compressing method according to claim 3 , wherein, in the management of a compression strategy for the data, as an allowable error computed for the time-series data, the same allowable error is set for all of one or more pieces of sensor data composing the time-series data.
5. The time-series data diagnosing and compressing method according to claim 4 , wherein, as allowable errors computed for the time-series data, allowable errors each computed independently for one of the sensors are set for one or more pieces of sensor data composing the time-series data.
6. The time-series data diagnosing and compressing method according to claim 3 , wherein, in the management of a compression strategy for the data, an output allowable error of the data is stored in a storage.
7. The time-series data diagnosing and compressing method according to claim 1 , wherein the data compressing means carries out irreversible data compression of thinning data satisfying a specified allowable-error range and of leaving data departed from an allowable error.
8. The time-series data diagnosing and compressing method according to claim 7 , wherein the data compressing means carries out multi-stage compression by further executing reversible compression after execution of the irreversible data compression.