# US 12,181,865 B2 · Operating state classification system, and operating state classification method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,181,865 B2
- Document ID: US12181865B2
- Status: Granted
- Kind: B2
- Application number: 17/801,922
- Filed: 2020-07-29
- Published: 2023-03-23
- Granted: 2024-12-31
- Earliest claimed: 2020-02-27
- Assignees: Hitachi, Ltd.
- Applicants: Hitachi, Ltd.
- Inventors: Kohji Maki
- Web page: https://patentgrep.com/patent/US12181865B2

## CPC

- G05B 23/0213: {Modular or universal configuration of the monitoring system, e.g. monitoring system having modules that may be combined to build monitoring program; monitoring system that can be applied to legacy systems; adaptable monitoring system; using different communication protocols}
- G05B 23/0254: {based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks}
- G05B 23/0216: {Human interface functionality, e.g. monitoring system providing help to the user in the selection of tests or in its configuration}
- G05B 23/024: {Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks}
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/09: Supervised learning
- G06N 3/091: Active learning
- G06N 3/098: Distributed learning, e.g. federated learning
- G06N 3/063: using electronic means
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The objective of the present invention is to realize an operating state classification system having a classification accuracy that continuously improves through additional learning, and which is appropriately protected from unauthorized duplication of a classification function. In this operating state classification system, in which an edge device and a server are connected by means of a communication network, and which inputs sensor data into a neural network and outputs a state label: the edge device includes a first storage unit which stores an upstream side of the neural network, and a dimensionality reduction unit which inputs the sensor data into the upstream side of the neural network and outputs intermediate data; and the server includes a second storage unit which stores a downstream side of the neural network, an inference executing unit which inputs the intermediate data into the downstream side of the neural network and outputs the state label, and a learning unit which updates the downstream side of the neural network by means of additional learning.

## Claims

### Claim 1

An operating state classification system having an edge device and a server connected via a communication network, which inputs sensor data to a neural network and outputs a state label,
wherein the edge device includes:
a first storage device which stores a pre-stage side of the neural network, and
a dimension compression system which inputs the sensor data to the pre-stage side of the neural network and outputs intermediate data, and
wherein the server includes:
a second storage device which stores a post-stage side of the neural network,
an inference execution system which inputs the intermediate data to the post-stage side of the neural network and outputs the state label, and
a learning system which updates the post-stage side of the neural network by additional learning;
wherein the pre-stage side of the neural network stored in the edge device is an encoder which is a pre-state portion of a self-encoder learnt for each state to be classified, and
wherein the post-stage side of the neural network stored in the server is a state classifier which inputs intermediate data output by the encoder and outputs the state label.

### Claim 2 (depends on 1)

The operating state classification system according to claim 1, wherein the learning system performs additional learning of the post-stage side of the neural network, based on the intermediate data stored in the server and the correct state label associated with the intermediate data.

### Claim 3 (depends on 2)

The operating state classification system according to claim 2, wherein when the probability of the state determined by the inference execution system to be the most probable is higher than a preset value, the correct state label is a state label output by the neural network.

### Claim 4 (depends on 2)

The operating state classification system according to claim 2, wherein when the probability of the state determined by the inference execution system to be the most probable is lower than a preset value, the correct state label is a state label input by a human being.

### Claim 5 (depends on 1)

The operating state classification system according to claim 1, wherein a decoder which is a post-stage portion of the self-encoder is also stored in the server.

### Claim 6 (depends on 1)

The operating state classification system according to claim 1, wherein the edge device is an industrial controller or a power conversion device.

### Claim 7

An operating state classification method comprising:
a learning step of learning a neural network inputting sensor data and outputting a state label on a server;
a first storing step of storing a pre-stage side of the neural network in an edge device, wherein the pre-stage side of the neural network stored in the edge device is an encoder which is a pre-state portion of a self-encoder learnt for each state to be classified;
a second storing step of storing a post-stage side of the neural network in the server, and wherein the post-stage side of the neural network stored in the server is a state classifier which inputs intermediate data output by the encoder and outputs the state label;
a third storing step of storing a decoder which is a post-stage portion of the self-encoder is also stored in the server; and
an additional learning step of additionally learning the post-stage side of the neutral network on the server.

### Claim 8 (depends on 7)

The operating state classification method according to claim 7, further including a step of storing additional learning data in which when the probability of a state determined by the neural network to be the most probable is higher than a preset value, the state label output by the neural network is given to intermediate data output by the pre-stage side of the neural network,
wherein in the additional learning step, the post-stage side of the neutral network is additionally learnt using the additional learning data.

### Claim 9 (depends on 7)

The operating state classification method according to claim 7, further including a step of storing additional learning data in which when the probability of the state determined by the neural network to be the most probable is lower than the preset value, a correct state label input by a human being is given to the intermediate data output by the pre-stage side of the neutral network,
wherein in the additional learning step, the post-stage side of the neural network is additionally learnt using the additional learning data.