Predictive Process Modeling for Industrial Parameter Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current industrial engineering practices rely on spreadsheet-based modeling and physical simulations, which lack predictive capabilities and are not easily maintainable, leading to inefficiencies in data management and optimization across industrial processes, particularly in applications like roll grinding where performance data is often lost when key personnel leave.
Innovation Solution
A multivariate statistical approach is implemented to create a predictive modeling tool that performs data analysis and generates models to optimize industrial processes, allowing for continuous improvement and recommendation of machine parameters, products, and process settings using a web-based interface without requiring extensive coding changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If spreadsheet-based modeling is used to track industrial performance data, then data collection and reporting are simple to implement, but predictive capability and optimization ability are insufficient
Solution Approach 1:
The patent replaces traditional spreadsheet-based mechanical data tracking systems with an automated electronic data collection and predictive modeling system. Sensors and controllers automatically capture process data, while machine learning algorithms generate predictive models, eliminating manual spreadsheet operations and enabling sophisticated predictive analytics without requiring expert intervention.
Solution Approach 2:
The system enables self-service through automated data collection from process sensors, automatic model training using historical data, and autonomous generation of optimization recommendations. The predictive modeling system continuously learns from new data and improves its predictions without requiring manual intervention from application engineers or data scientists.
2Measurement precision
If physical simulations with fine-tuned formula coefficients are used to describe current knowledge, then key performance indicators can be well described, but predictive capability for new information is limited
Solution Approach 1:
The patent implements dynamic predictive models that continuously adapt to new information. Unlike static physical simulations with fixed coefficients, the machine learning models are continuously trained on new process data, allowing them to evolve and improve their predictions over time. The system dynamically updates model parameters based on incoming data streams, enabling both accurate description of current knowledge and prediction of new outcomes.
Solution Approach 2:
The system performs preliminary actions by continuously training predictive models on historical data before new processes or products are introduced. This pre-learning phase builds a knowledge base that enables the system to make informed predictions about new scenarios, reducing the need for extensive physical simulations and expert analysis when encountering new situations.
3Ease of operation
If physical models are maintained by formula and coefficient experts, then knowledge can be spread to a wider group, but the models require continuous expert motivation to maintain
Solution Approach 1:
The system eliminates the need for expert maintenance through self-service automation. The predictive modeling platform automatically collects data, trains models, validates predictions, and updates parameters without requiring formula experts or data scientists. This automation reduces maintenance complexity while preserving and spreading organizational knowledge through the persistent digital models that outlive individual employees.
Solution Approach 2:
The patent creates digital copies of expert knowledge embedded in machine learning models. Instead of relying on human experts to maintain and transmit knowledge, the system captures expert insights in algorithmic form that can be replicated and distributed across the organization. These digital knowledge artifacts maintain consistency and can be updated centrally, then automatically propagated to all users without requiring expert intervention at each location.
4Reliability
If three to five years are needed to become an experienced Industrial Application Engineer, then deep process knowledge is gained, but performance information is lost when key personnel leave
Solution Approach 1:
The system creates digital copies of experienced engineer knowledge embedded in predictive models and algorithms. These models capture the nuanced understanding that takes years to develop, encoding it in machine-readable form that persists independently of individual employees. When personnel leave, the organizational knowledge remains intact in the system, eliminating the risk of knowledge loss while maintaining the depth of expertise through continuous model training on historical data.
Solution Approach 2:
The system performs preliminary knowledge capture by continuously learning from all process data, including cases handled by experienced engineers. This ongoing learning process builds a comprehensive knowledge base before any personnel changes occur, ensuring that institutional knowledge is preserved and made accessible to all employees regardless of their individual experience levels.
5Reliability
If statistical machine learning applications are used for industrial applications, then predictive capability and continuous optimization are enabled, but the systems require sophisticated data models and algorithms
Solution Approach 1:
The patent implements self-service automated machine learning that eliminates the need for data scientists to manually build and maintain complex models. The system automatically performs data preprocessing, feature engineering, model selection, training, and validation using pre-configured algorithms tailored to industrial applications. This automation reduces system complexity from the user perspective while maintaining sophisticated predictive capabilities through continuous automated model improvement.
Data Source
AI summary
A system and method of simulating and optimizing industrial and other processes includes a computer that performs multivariate analysis of input variables and output variables to generate a data model of the operation of the process. For industrial applications, the input variables include process variables and the output variables include result variables from the operation of the industrial process. The data model determines contributions to changes in the output or result variables by the respective input or process variables and is provided to a predictive algorithm to identify parameter values for input or process variables expected to have a most significant impact on the output or result variables during performance of the process. The outputs of the predictive algorithm are parameter values that are provided as input or process variables to the industrial process for simulation or performance optimization or product recommendations/optimizations.