Manufacturing Data Characterization Using DRL and Coupled Feature Encoding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing characterization methods for discrete manufacturing industry data struggle with mixed data composed of continuous and discrete data, leading to information loss due to discretization and an inability to adapt to dynamic and complex environments, as they ignore the relation between continuous and discrete features.

Innovation Solution

A characterization method based on deep reinforcement learning that collects discrete manufacturing industry data, creates a spatio-temporal database, divides data into discrete and continuous features, and uses a data coupling coding network to convert coding vectors into characterization vectors, employing cluster evaluation indexes as dynamic rewards to update neural network parameters and optimize data characterization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If continuous features are discretized using traditional methods, then data can be processed by discrete algorithms, but information loss occurs and relationships between continuous and discrete features are ignored

Engineering Contradiction:

Improveadaptability to discrete algorithmsVSAvoidinformation loss during discretization

Core Design Contradiction:

Adaptability or versatilityVSLoss of information

Solution Approach 1:

The patent introduces a correlation matrix as an intermediary structure that captures relationships between continuous and discrete features. This matrix serves as a mediator that preserves the coupling relations during the discretization process, allowing continuous features to be converted to discrete form while maintaining their relational structure through the correlation matrix representation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms continuous feature parameters into discrete parameters through a systematic conversion process. By changing the parameter representation from continuous to discrete while using the correlation matrix to guide the transformation, the method maintains the essential relationships between features while making the data suitable for discrete algorithms.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If existing deep learning algorithms are used for static discrete manufacturing environments, then they can process data effectively, but they cannot adapt to complex and dynamic discrete industry manufacturing problems

Engineering Contradiction:

Improvedata processing effectivenessVSAvoidadaptability to dynamic environments

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent transforms static deep learning algorithms into dynamic ones by introducing a dynamic reward mechanism. The reward function adapts to changing manufacturing environments by incorporating temporal correlations and dynamic feature relationships, allowing the algorithm to adjust its learning process in real-time to complex and dynamic discrete manufacturing problems.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements a feedback mechanism through the dynamic reward function that continuously monitors performance and adjusts the learning process. The reward signal provides feedback about the quality of predictions in dynamic environments, enabling the algorithm to learn from interactions and adapt to changing manufacturing conditions.

Inventive Principle:

Principle #23Feedback

3Ease of manufacture

If artificial experience is used to set reward values in deep reinforcement learning, then implementation is straightforward, but optimal data characterization decision-making cannot be provided for discrete industry manufacturing systems

Engineering Contradiction:

Improveease of implementationVSAvoidcharacterization decision accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent enables the system to automatically determine optimal reward values through self-service mechanisms. The algorithm learns the appropriate reward structure by analyzing the manufacturing data and problem characteristics itself, rather than relying on manual configuration. This allows the system to provide optimal characterization decisions while maintaining ease of implementation through automated parameter determination.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20240210924A1Characterization method based on deep reinforcement learning for discrete manufacturing industry data

Publication Date: 2024.06.27 NANJING UNIV OF POSTS & TELECOMM

AI summary

Disclosed is a characterization method based on deep reinforcement learning for discrete manufacturing industry data. The method includes: collecting discrete manufacturing industry data, and creating a spatio-temporal database; dividing the discrete manufacturing industry data into a discrete feature and a continuous feature, creating a data coupling coding network, converting a coding vector in the coding network into a characterization vector, and creating a data characterization model; quantitatively characterizing discrimination of a data category by means of cluster evaluation indexes; and using weights of cluster evaluation indexes of different dimensions as dynamic rewards, creating a deep reinforcement learning model, and updating a neural network parameter of deep reinforcement learning through characterization of an interactive relation between a model and a discrete manufacturing decision-making analysis system.