If you’re unfamiliar with Factory Harmonizer, let’s begin with a quick brief. Factory Harmonizer is an intelligent software solution designed for operators and engineers in the process industry. Factory Harmonizer brings together the best of human expertise, machinery, and automated machine learning to stabilize the production process for better productivity.
With that introduction complete, we can move on to today’s topic: pilots.
An Introduction to Offline Pilots
SimAnalytics’ Offline Pilot is a structured 3–4-month initiative aimed at exploring the latent possibilities of Factory Harmonizer using the customer's own historical production data.
While the insights generated during pilots can be used for immediate production improvements, they also validate and demonstrate the value potential of data-driven methods in improving the customer’s processes. In practice, this implies using the historical data obtained from the customer to develop and test models that identify unique patterns, anomalies, and optimization opportunities. To ensure customer satisfaction, each project is assigned a dedicated project manager and a project team of process experts, data engineers and machine learning specialists.
Phase One: Target Setting and Data Collection
The first phase of the pilot involves finalizing the targets discussed during the sales process and collecting the necessary data. This phase is crucial as it establishes the groundwork for all subsequent activities. The objective is to define clear goals i.e. to understand what challenges the customer wants to solve using the tool, this can range from improving the yield, and increasing throughput, to enhancing product quality. The goals for the offline project are set to match these goals. For example, if the customer wants to improve their yield, what information needs to become clear from the results of the offline project for the investment to be valuable to the customer?
Once the goals have been set a workshop is held to understand what data points/measurements are relevant for the model building. Thereafter, comprehensive historical data from the customer's production processes (usually a year of data for each data point) is gathered by the customer and delivered to SimAnalytics’ Delivery Team for processing. This data serves as the foundation for developing models and identifying optimization opportunities.
Key activities in this phase include:
- Establishing benchmarks for current and future performance.
- Collecting historical data to ensure a robust dataset for analysis.
Benefits for the Customer:
- Helps the customer define and clarify what specific challenges they want to address using Factory Harmonizer.
Time Required from the Customer:
- 4-12 hours (can be longer in certain cases due to the customer’s automation system capabilities)
Phase Two: Data Health Check
Before any models can be built, the first step in any successful data-driven initiative is ensuring the integrity and quality of the data itself. During this phase, a comprehensive Data Health Check report is generated for the customer. This report scrutinizes the historical data for completeness, accuracy, and consistency. By identifying gaps and inconsistencies, this phase ensures that the subsequent analyses and models are built on a solid foundation. SimAnalytics’ machine learning specialists produce the Data Health Check and provides the customer with a comprehensive understanding of the quality of their production data, which can hence also be used by the customer for internal development efforts.
Key activities in this phase include:
- Checking historical data for completeness, accuracy, and consistency.
- Generating a comprehensive Data Health Check report.
Benefits for the Customer:
- Provides the customer with a comprehensive report on the quality of their production data.
- The Data Health Check can be used by the customer for their own internal data quality improvement efforts.
Time Required from the Customer:
- None
Phase Three: Model Building
Once the data's health is validated and tags are selected accordingly, the next phase involves SimAnalytics’ Delivery Team putting their best efforts into building the most accurate models they can using the historical data. The type of model depends on the module(s) in question. For the FH Harmony module, the models are as the name suggests, harmony models, whereas for FH Cycle, these modules are target windows showcasing optimal operation parameters. For FH Soft sensors, the outcome is key figures, such as model prediction accuracy. Whilst the aim of the models is for them to be put into online use, the information provided by the models can prove to be very useful for continuous improvement efforts.
The activities in this phase include:
- Developing models that predict process behaviour.
- Continuous testing and tweaking of models to best help solve the customers’ challenges.
Benefits for the Customer:
- The models developed provide insights that can help the customer optimize their production processes.
Time Required from the Customer:
- None
Phase Four: Analysis, Evaluation, and Iterations
The final phase involves analyzing the results, evaluating the models, and making necessary iterations in close collaboration with the customer. This phase ensures that the insights generated are accurate and actionable.
At this stage, once the models are approved, the customer receives training and access to their own Factory Harmonizer environment with their own historical data. This allows the customer to explore the benefits of the tool and get a feel for using the tool for their use case.
Key activities include:
- Iterations of models based on customer feedback to ensure model accuracy
- Evaluating the findings to determine how the insights can be utilized to adjust processes.
- Conducting workshops to facilitate knowledge transfer and practical application of the findings.
- Setting up the customers’ own Factory Harmonizer environment.
Benefits for the Customer:
- The customer receives training and access to their own Factory Harmonizer environment, enabling them to explore and leverage the tool's capabilities as well as the insights that it provides..
Time Required from the Customer:
- 1-2 days of which the majority for own testing
Demonstrating Value and Driving Improvement
The culmination of the offline pilot is a comprehensive demonstration of the value potential of data-driven methods. By the end of the three to four months, the customer will have a clear understanding of how their data can be transformed into a strategic asset. The insights generated not only offer immediate opportunities for improving production but also lay the groundwork for a sustained, data-driven approach to process optimization.
To learn more about how your production data can be the key to solving your challenges, read more through one of the blog posts below, or book a demo with one of our experts.
