“I don’t spend my time pontificating about high-concept things; I spend my time solving engineering and manufacturing problems.”
Elon Musk, CEO and product architect of Tesla, Inc.
If finding the Golden Batch in your production run is a big challenge, repeating it every time is a bigger problem.
The golden batch is an ideal production run, which works as a template of gold standards. All future mass-production lines would have to measure up to it for quality control. You find a golden batch when all the stages of production get perfected, from design to the process workflow and finally to your customer approval. It may take several trials and revisions and thus a lot of time and effort gets invested here, but the golden batch is a manufacturer’s pursuit. Not only does it act as a benchmark of optimum quality, production efficiency, cost-effective processes achieved with minimal wastage, but also proves that you can reliably supply what the customer has ordered.
However, several components continue to influence the production line. Raw materials, equipment, labor, space, and costs are the usual suspects. You would need to keep a keen eye out to see how these elements influence each other. Then, for a perfect quality production run, you would need to measure the variables carefully, regulate, and change the production processes accordingly. Thus, the problems of the manufacturer do not end with the finding of a single golden batch. You face the challenge of consistently delivering uniformly excellent yields from every production line thereafter.
Despite all your efforts, there is a high error rate causing quality issues? There are many likely reasons for that.
Hurdles in Repeating the Golden Batch
- The R&D Golden batch does not consider process variables, e.g., raw material variations, asset performance deterioration, ambient conditions, etc.
- Multi-step and often complex processes require multiple variable analysis over long and varying batch cycle times (BCT). This evaluation is difficult and requires significant manual effort.
- Most of this analysis is offline and dated and may not be most relevant to current production runs.
- Operators do not get the intimation of alerts and recommendations in real-time, and lack of timely guidance hampers the production run.
- Subject-Matter Experts can detect where and when the processes deviate, but it is tough to get to the exact root cause of parameter variations.
When there are thousands of parameters interacting and varying simultaneously, the impact of inadequacies directly affects the output. Fluctuating yields across batches, sub-optimal production, quality issues, and rejected batches can create instability. The business may suffer severely, too.
The good news is that it is not too late. You could still consider implementing AI in your manufacturing processes to improve operational decision-making and productivity.
AI for real-time data analytics is the simplest solution to hitting big wins by identifying the Golden Batch and the best manufacturing process that will repeat it consistently.
AI applications can show you every aspect of your manufacturing batch or production process, using multivariate data analysis techniques (MVDA). Simplifying the identification of the best range of parameters to guarantee optimized production consistently. It also monitors operations, creates simulations, and offers real-time guidance for easy implementation on the shop floor. Integration of hardware with the Golden Batch recipe enables end-to-end digitalization.
The 5 Easy Steps to Getting a Golden Production Run:
Step 1- Feature Generation
Feature engineering is extracting the most relevant variables and refining features from raw data, which will better represent the underlying problem to the predictive modeling algorithms. That results in improved model accuracy on unseen data.
According to a survey in Forbes, data scientists spend 80% of their time on data preparation [1]. But with the Golden Production Run AI App, acquiring integrated data gets sped up and simplified.
It helps achieve optimal outcomes by:
- Feature Engineering based on long Batch Cycle Time
- Integration of a) Compositions b) In-Process Parameters c) Yield d) Production Quantity e) Quality Results f) Asset Downtimes and Maintenance
- Automated algorithmic data quality checks to ensure processing of good and relevant data
- Handling complexities arising from time dimensions and complex process interactions in terms of data preparation
Step 2- Identifying the Golden Thread
The Golden Thread is a term used to describe a performance model, which aligns your manufacturing objectives with the accurate measures of success. Identifying the golden thread means finding that perfect combination of raw material composition and process parameters that will guarantee the best output in quantity and quality every time.
The Golden Production Run AI App establishes a relationship between various factors that influence the production line and assesses the Process Memory-based algorithms to derive the golden thread. It enables statistical pattern identification and gives real-time insights into correlations between varying parameters and the production output.
Step 3- Generation of Parameter Scorecards
A scorecard is the visual representation of the manufacturing objectives. It tracks and presents the success metrics based on your needs and the time frame you select while comparing the different values.
The Golden Production Run AI App generates a simple yet effective parameter scorecard. It helps in the real-time identification and rating of prime contributors, causing variations in output metrics. When the production run output is not up to the mark, the scorecard identifies the combination of maximum impact parameters and the magnitude of fluctuations in terms of significance.
Step 4- Validate the Golden Thread
Subject Matter Experts (SMEs) can fine-tune the performance model when they zero in on the golden production parameters. SMEs use their extensive industry experience to confirm the parameter relevance and associated importance with output metrics while validating the golden thread.
Step 5- Enjoy the Golden Production
And Voila! You will have successfully implemented a Golden Production Model that delivers consistent optimal capacity yield of desired quality. Run time predictions, alerts for production leadership teams, and the real-time guidance system for operators create an immediate impact on the shop floor and the final product.
By utilizing the real-time Process Memory algorithms and the real-world expertise of SMEs, the Golden Production Run AI app delivers consistent production runs excelling in quantity, quality, and efficiency.
What is the Golden Production ROI?
- 8-10% improvement in production
- 3-5% improvement in yield
- 2-4% reduction in rejected batches
- 1-2% reduction in batch cycle times
- 3-5% improvement in quality parameters
Read the Full Story of How a Large Pharmaceutical Company in Europe Improved their Yield…
The sub-process, which needed examination, was the API process for a blockbuster drug manufactured by a large pharmaceutical company in Europe. It generated a pivot product that gets crystallized in 2 salts. The production process involved reactions across various process steps and the crystallization process. A higher yield became desirable from each batch run because the manufacturing process had a long batch-cycle time. The crystallization process got repeated multiple times to ensure increased output. The recovery of ethanol was also a side step, and it got reused in the key chemical process of manufacturing the drug.
The Business Problem
The sub-process selected for this implementation had a problem of fluctuating yield between various batches; yields would fluctuate between 8-10% over the different production runs. The production operator’s team could see which area the yield was getting dropped, but they had a limited understanding of the process parameters and the reasons for the fluctuation. They, therefore, could not get to the root cause of the problem. Since the element was disturbing the production process of a blockbuster drug, low yield meant ineffective utilization of the production plant capacity. In summary, the process deviation was causing-
a) Cost impact in terms of quantity of raw material used and
b) Production capacity wastage due to low yield.
How Was the Problem Solved?
After initial process understanding, the parameters got mapped at the Micro and Macro levels. At each chemical step, the process got broken down into multiple technical steps using fuzzy algorithms. After acquiring the break-up for all the batches, the process memory algorithm got run — both at macro and micro levels. Data harmonization and data quality checks across batch times provided real-time monitoring and comparisons. The parameters got identified, which kept varying locally and globally, causing the fluctuating yield. Simultaneously, optimal process parameters for the best output also got recommended in real-time. Subsequently, the production team verified all the parameters, and controlled and changed only those parameters that got selected for action. When provided with accurate recommendations and simplified solutions, the customer noticed a significant impact on business operations. Using Neewee’s AI App, the pharmaceutical company improved its yield by 2-4%.
The Bodhee® Golden Production Run AI App had successfully identified that the quantity of water and ethanol used during one of the technical steps was the hidden cause of the hurdle in repeating the golden batches.
Reference:
[1] Press, G. (2016, March 23). Cleaning Big Data: Most Time-Consuming, Least Enjoyable Data Science Task, Survey Says. Forbes. https://www.forbes.com/sites/gilpress/2016/03/23/data-preparation-most-time-consuming-least-enjoyable-data-science-task-survey-says/?sh=369279086f63
