Deep Learning Thresholding for Predictive Factory Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manufacturing processes face challenges in consistently meeting design specifications due to the limitations of traditional process controllers, which rely on static algorithms and fail to optimize key performance indicators (KPIs) in real-time, leading to inefficiencies and waste.
Innovation Solution
A manufacturing system incorporating a deep learning processor that simulates manufacturing processes to generate expected values and target values for KPIs, allowing for dynamic adjustments to processing parameters to optimize performance, using a combination of confidence scores, specification limits, and Statistical Process Control (SPC) rules to guide changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional process controllers with static algorithms are used, then device complexity is reduced, but manufacturing precision and productivity deteriorate due to inability to optimize KPIs in real-time
Solution Approach 1:
The patent transforms static process controllers into dynamic systems by implementing continuous simulation and prediction capabilities. The system dynamically adjusts processing parameters based on real-time predictions of quality outcomes, enabling adaptive optimization rather than fixed control logic. This resolves the contradiction by making the controller complex enough to predict and adjust for precision while maintaining operational simplicity through automated decision-making.
Solution Approach 2:
The system performs preliminary simulation and prediction of quality outcomes before actual processing occurs. By using predictive models to forecast quality metrics based on proposed parameter changes, the system pre-evaluates potential outcomes and selects optimal adjustments beforehand. This preliminary action ensures design specification compliance is achieved proactively rather than reactively, improving manufacturing precision without requiring overly complex real-time control algorithms.
2Device complexity
If traditional process controllers with static algorithms are used, then device complexity is reduced, but productivity deteriorates due to inability to optimize key performance indicators
Solution Approach 1:
The system implements dynamic optimization by continuously simulating different processing scenarios and predicting their impact on key performance indicators. Rather than relying on fixed control algorithms, the system adapts its control strategy in real-time based on predicted outcomes, enabling continuous improvement of productivity metrics while managing controller complexity through automated predictive modeling.
Solution Approach 2:
The patent establishes a closed-loop feedback system where simulation results and predicted quality outcomes feed back into parameter adjustment decisions. The system continuously monitors actual processing results, compares them with predictions, and uses this feedback to refine future predictions and adjustments. This feedback mechanism enables continuous productivity optimization without requiring exponentially increasing controller complexity, as the system learns and improves over time.
3Manufacturing precision
If deep learning simulation is implemented for real-time optimization, then manufacturing precision and productivity improve, but device complexity increases
Solution Approach 1:
The patent introduces an intermediary simulation layer that acts as a bridge between simple process control and complex deep learning models. Rather than directly implementing complex deep learning in the controller, the system uses a simulation environment that predicts quality outcomes, allowing the controller to remain relatively simple while still achieving high manufacturing precision through informed decision-making based on simulation results.
4Productivity
If deep learning simulation is implemented for real-time optimization, then productivity improves, but device complexity increases
Solution Approach 1:
The simulation system serves as an intermediary that handles the computational complexity of deep learning predictions, allowing the actual production control system to remain simpler. By separating the complex predictive modeling from the execution control, the system achieves high productivity through optimized parameter selection while managing overall system complexity through this architectural division.
Solution Approach 2:
The system performs preliminary simulations and predictions before actual production adjustments, evaluating multiple scenarios in advance to identify optimal parameter settings. This preliminary action allows complex computational work to be done beforehand, enabling rapid decision-making during actual production with simpler real-time control logic, thus improving productivity without proportionally increasing real-time system complexity.
Data Source
AI summary
A deep learning process receives desired process values associated with the one or more process stations. The deep learning processor receives desired target values for one or more key performance indicators of the manufacturing process. The deep learning processor simulates the manufacturing process to generate expected process values and expected target values for the one or more key performance indicators to optimize the one or more key performance indicators. The simulating includes generating a proposed state change of at least one processing parameter of the initial set of processing parameters. The deep learning processor determines that expected process values and the expected target values are within an acceptable limit of the desired process values and the desired target values. Based on the determining, the deep learning processes causes a change to the initial set of processing parameters based on the proposed state change.