Real-Time Production Control for Residence Time Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing real-time production control methods struggle to efficiently manage residence time constraints in complex production systems, such as semiconductor manufacturing, due to the large state space and diverse structures of these constraints.
Innovation Solution
A simulation-based real-time control method using a Markov Decision Process (MDP) model with a feature-based approximate architecture is proposed. This method reduces the state space complexity by extracting features and approximating a lookahead function, allowing for real-time control of production systems with various classes of residence time constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional real-time production control methods are used, then the system can handle simple production scenarios, but it fails to efficiently manage complex production systems with large state space and diverse residence time constraints
Solution Approach 1:
The patent segments the complex production system into multiple workstations and buffers, each with locally defined state variables. The global state space is decomposed into local state spaces at each workstation, allowing independent analysis and control. This segmentation reduces the computational burden of managing the overall large state space while maintaining the ability to handle diverse residence time constraints across different system segments.
Solution Approach 2:
The patent transforms the control problem by changing parameters from traditional production control variables to residence time-based parameters. By defining residence time constraints for different buffer segments and using these as primary control parameters, the system can efficiently manage diverse constraint types without exponentially increasing state space complexity. The parameter transformation enables the use of specialized algorithms optimized for residence time management.
2Productivity
If production rate is maximized without constraints, then output increases, but scrap rate increases due to excessive intermediate parts in buffers
Solution Approach 1:
The patent implements feedback control by continuously monitoring the number of intermediate parts in each buffer and adjusting workstation production rates accordingly. When buffer occupancy approaches residence time limits, the system provides feedback signals to upstream workstations to reduce production, preventing excessive accumulation. This closed-loop feedback mechanism dynamically balances production rate maximization with scrap rate minimization by adapting control actions to real-time system state.
Solution Approach 2:
The patent applies preliminary action by proactively controlling workstation production rates before buffers become overloaded with intermediate parts. By predicting potential residence time violations based on current buffer occupancy trends, the system takes preventive control actions to maintain parts within acceptable residence time windows, thereby avoiding scrap generation while sustaining high production rates.
3Loss of substance
If strict residence time constraints are enforced, then scrap rate decreases, but production rate decreases due to frequent machine idle time
Solution Approach 1:
The patent applies dynamics by making residence time constraints adaptive rather than static. The system dynamically adjusts effective residence time limits based on real-time conditions such as downstream workstation availability, buffer capacities, and part priority levels. This dynamic constraint enforcement allows the system to maintain low scrap rates by being strict when necessary while permitting higher production rates when system conditions allow, optimizing the trade-off between quality and productivity.
Solution Approach 2:
The patent implements partial action by applying strict residence time control only to critical buffers and part types where scrap prevention is most important, while allowing more flexible control in less critical areas. This selective enforcement approach ensures that scrap rate is minimized for high-risk items without unnecessarily constraining overall production rate, achieving effective control with reduced impact on productivity.
Data Source
AI summary
A computer-implemented model is trained or otherwise configured for real-time production control. The model incorporates a plurality of classes of residence time control and optimizes produce performance by managing an associated machine's behavior according to real-time system states. The model can formulate a Markov decision process and feature-extraction method and feature-based approximation architecture to reduce state space of the model. Simulation is employed in training to estimate parameters of feature-based approximate architecture to obtain a lookahead function of the Markov decision process.