Machining Workflow Prediction Using Clustered Process Progressions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In manufacturing processes, particularly in CNC machining, the time required to plan and execute tasks is substantial due to the need for defining multiple steps, tools, and parameters, and there is a lack of standardization across users and over time, influenced by factors like machine wear and operating conditions.
Innovation Solution
A system comprising a processor and memory that uses machine learning to determine similarity metrics and clusters from previous tasks, creating inference models to predict process progressions for new tasks based on similarity, and adjusts predictions based on user preferences and machine capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual planning processes are used to program machines for manufacturing tasks, then flexibility and adaptability to specific task requirements are maintained, but substantial time is required for defining steps, tools, and parameters
Solution Approach 1:
The system creates copies of previously executed task characteristics and process progressions, storing them in a database. When a new task arrives, the system retrieves and reuses similar past processes through similarity matching, eliminating the need to manually plan from scratch and significantly reducing planning time while maintaining proven effective processes
Solution Approach 2:
The system performs preliminary actions by pre-processing and encoding task characteristics, pre-calculating similarity metrics, and pre-clustering process progressions during idle periods or after task completion. This preparation work is done in advance so that when new tasks need planning, the system can quickly retrieve and adapt pre-processed information rather than computing everything in real-time
2Stability of the object's composition
If standardized processes are implemented across users and time, then consistency is improved, but the ability to capture changes in user behavior due to machine wear, tooling wear, and operating conditions is reduced
Solution Approach 1:
The system implements dynamic process planning where the standardization level adapts based on task similarity. For highly similar recurring tasks, the system enforces strict standardization by reusing exact process progressions from the database. For tasks with notable differences or when machine/tool wear is detected, the system allows greater flexibility to modify processes, thus maintaining both consistency where needed and adaptability where required
Solution Approach 2:
The system incorporates feedback mechanisms where actual task execution results, machine performance data, and tool wear indicators are continuously monitored and fed back into the database. This feedback loop allows the system to learn from real-world performance and automatically adjust future process recommendations, capturing changes in user behavior and equipment conditions while maintaining overall process standardization
3Measurement precision
If machine learning models are trained on all available task data, then prediction accuracy is improved, but computational resources and training time are increased
Solution Approach 1:
The system segments the task database into multiple clusters based on task characteristics and process progression similarities. Instead of training a single model on all data, separate inference models are trained for each cluster. This segmentation reduces the computational burden of training while maintaining high prediction accuracy for each specific task type, as each model focuses on a narrower, more homogeneous dataset
Solution Approach 2:
The system performs preliminary clustering and organization of task data before model training. By pre-processing and structuring the data into meaningful clusters in advance, the training process becomes more efficient as models only need to learn patterns within each cluster rather than from the entire heterogeneous dataset, reducing both training time and computational resources required
Data Source
AI summary
Systems, devices, and methods for executing a model preparation component for training a system and a model query component for querying the system are disclosed. In the preparation stage, the system determines a set of process clusters based on a set of similarity metrics for task characteristics using a set of encoded task characteristics and associated process progressions. The system may also create a set of inference models using the determined set of process clusters. In the query stage, the system may identify an encoded task characteristic and a process progression candidate by matching the encoded task characteristics with the closest similarity metric. The system may then determine a process progression prediction based on at least one of the process progression candidate using the similarity metric and/or calculating a new process progression by an inference calculator using the created set of inference models.