By Praveen Hewage, Alumex PLC, and Craig Werner, Werner Extrusion Solutions LLC.
Artificial Intelligence (AI) has the potential to reshape the modern world in profound ways, using computer-based systems to augment human intelligence with capabilities, such as perception, learning, reasoning, problem-solving, and decision-making. AI learns from data and makes decisions based on those learnings. It can analyze large amounts of data, automate processes, make predictions, enhance decision-making, and revolutionize industries leading to improved efficiencies and outcomes. It can uplift mankind’s quality of life, and for the aluminum industry, it can improve manufacturing processes, such as die correction. It is important to note that AI is not intended to replace human workers, but to serve as an aid, supporting their work during manufacturing and maintenance tasks.
This article will utilize die correction as a short phrase to synthesize the understanding of existing equipment, processes, tooling, and material flow, which together affect the metallurgical performance and critical quality aspects of shape control, surface finish, and related attributes. “Die correction” is not intended to simply mean the mechanical aspects of modifying tools to improve them, but to include all aspects of using training, experience, situational analysis, and past trials and results to achieve improved results.
An Overview of AI
AI is used for diagnosis, personalized treatment plans, and drug discoveries in health care; automating fraud detection, risk assessment, and investment strategies in finance; enabling precision farming and crop monitoring in agriculture; and driving autonomous vehicles and optimizing logistics in transportation. It is being utilized in customer service, e-commerce, energy, education, and many other sectors, revolutionizing processes, improving outcomes, and unlocking new possibilities for businesses and society at large.
In manufacturing, AI-powered systems are used for predictive maintenance, quality control, supply chain optimization, and inventory management. Robotics and automation powered by AI are enhancing productivity and reducing human error. AI enables the development of smart factories with real-time data analysis, enabling better data-driven decisions and higher levels of operational excellence. AI in manufacturing identifies patterns by analyzing vast amounts of production, sensor, and quality control data and correlates this data using machine learning algorithms to detect patterns, trends, and anomalies in the data, enabling manufacturers to identify potential issues and optimize processes.
The Importance of Tooling in the Extrusion Market
Global aluminum demand in the transportation, aerospace, construction, packaging, and consumer goods markets is projected to grow from $163.5 billion in 2020 to $258.8 billion by 2028—driven by increasing demands for aluminum’s light weight, strength, sustainability, and other properties. Technological advancements in aluminum processing make it a promising and dynamic industry. The global aluminum extrusion industry offers versatile and customizable solutions for many applications and is projected to grow from $65.6 billion in 2020 (representing 40% of the aluminum industry) to $100.8 billion by 2028. Again, this is due to the increasing demand for sustainable materials that are able to fulfill decarbonizing requirements and support lightweighting.
Extrusion tooling knowledge is of paramount importance in the aluminum extrusion industry, directly affecting the product quality and process efficiency. A properly designed, manufactured, and maintained die ensures precise and consistent extrusion of profiles, resulting in high-quality products that meet customer specifications and industry standards. Die failures can have significant cost and delivery delay impacts, requiring costly scrap, rework, and downtime. Improper understanding of the root cause and correction can result in inconsistent dimensions and surface finishes on the extruded profiles and even more scrap, delays, and costs. Failed dies may require expensive repairs or replacements and lead to customer complaints and potential loss of business opportunities. Delays and poor-quality extruded products can lead to customer dissatisfaction, product recalls, and damage to the reputation of the extrusion company, resulting in severe financial effects.
Ensuring proper die analysis and correction is crucial in the aluminum extrusion industry to minimize the cost impacts of die failures, enhance productivity, maintain high product quality, and uphold customer satisfaction. Proper die maintenance, regular inspections, and most importantly skilled die technicians are essential to prevent die failures and ensure smooth and efficient extrusion operations.
Die Technician Challenges
Developing skilled die technician capabilities is one of the biggest challenges faced by the extrusion industry. The main factors behind this challenge can be identified as follows.
Complexity
The complex nature of die correction requires a high level of technical expertise and experience. This involves analyzing the intricate geometry of the die to understand effective improvements and the application of precise adjustments to achieve the desired profile dimensions, surface finishes, metallurgical performance, productivity, and yield. This requires a thorough understanding of the extrusion process, metallurgy, and material behavior, as well as proficiency in using specialized precision hand tools and machining equipment.
Limited Availability of Training Programs
There are limited formal training programs specifically focused on improving die technician capabilities. Most of the knowledge and skills related to die correction are acquired through on-the-job training and hands-on experience, making it difficult to improve these skills efficiently and effectively.
Lack of Standardized Processes
Die correction techniques vary depending on the specific extrusion profile, process, equipment, material, and die design. There is no one-size-fits-all approach to die correction, which too often requires a trial-and-error approach to achieve optimal results. This lack of standardized processes makes it challenging to develop consistent and standardized training programs for die technicians.
Limited Exposure to Advanced Technologies
Die correction can involve the use of advanced technologies such as computer-aided design (CAD), computer-aided manufacturing (CAM), scanning technologies, finite element analysis (FEA) simulation software, and precision measurement tools. Not all extrusion companies have access to or utilize these advanced technologies, which limits the exposure of technicians to these tools and hinders their skill development.
Retention and Turnover of Skilled Technicians
The retirement of experienced die experts continues to have a significant impact on the knowledge transfer of critical skills to die technicians, resulting in the loss of institutional knowledge, reduced average skill levels, delays in problem-solving, and the potential for costly mistakes. The expertise and experience of retired experts may not be easily transferable to the next generation of technicians, leading to challenges in maintaining efficient and effective die design and correction processes and likely suboptimal extrusion results.
Traditionally, proactive methods in these situations would include documenting institutional knowledge, implementing comprehensive training programs, facilitating mentorship and knowledge transfer, providing opportunities for job rotation and cross-training, and retaining skilled workers through incentives or flexible retirement plans. These methods can help capture and preserve the expertise of retiring experts, transfer their knowledge to younger technicians, and ensure a smooth transition of skills and knowledge, mitigating the impact of expert retirement on die correction processes and maintaining efficient operations. Due to the dwindling number of experts, these traditional means will likely not be sufficient or efficient enough to fill the need.
AI Applied to Aluminum Extrusion
Driving automobiles has been controlled by humans for more than a century. However, society is moving toward autonomous transportation, driven by the systems that process big data from image recognition systems and other sources and aided by machine learning and neural networks. If AI systems can replace such a complex, skill-based task as driving why not use AI to ensure and improve the most skill-oriented task—die correction troubleshooting and decision-making?
This accomplishment will require a huge concentrated effort from many in this industry working together. Capturing extrusion tooling trouble shooting knowledge and die correction intelligence in such a system will then offer tremendous advantages to help fill the upcoming knowledge gap—ensuring that these core technical skills in the aluminum extrusion industry aren’t lost.
By working towards developing an AI platform for die correction intelligence, the industry can capture these critical skills and insights, revolutionizing the future of die correction in the aluminum extrusion industry, enabling data-driven decision-making, enhancing die correction intelligence, and supporting even more futuristic knowledge management (Figure 1). AI algorithms can analyze vast amounts of diversified data related to machinery, process parameters, and product quality details, as well as aluminum flow behaviors to identify patterns and correlations, leading to more informed and accurate decision-making. AI-powered knowledge management can capture and transfer expertise, ensuring continuity in the correction process despite retirements or turnover. Collectively, these advancements can greatly enhance the efficiency and effectiveness of die correction, leading to revolutionary improvements in the aluminum extrusion industry.
Die design and correction are complex tasks that require a deep understanding of various factors, such as material properties, process parameters, and production constraints. Skilled die correctors possess valuable tacit knowledge that guides their decision-making process, which includes their experience and intuition. Capturing and understanding this thought process is essential for developing an AI platform that can effectively emulate their decision-making and provide accurate recommendations for die correction tasks.
The die correction process is context-dependent. Different die designs, extrusion types, and production environments may require different approaches to achieve optimal results. Skilled die correctors take into account the specific context of each analysis and correction task and apply their expertise accordingly. Understanding their thought process allows the AI platform to comprehend the nuances and complexities of the die correction process, enabling it to generate contextually relevant and effective solutions.
Analyzing the Die Correction Problem
Before making any decisions or physical alterations, skilled die technicians analyze the problem and understand the entire context, taking into account a number of factors. The correction plan to be developed must be based on the factors collected during the analysis of prior steps, connecting all the threads and using expertise to identify root causes and identify possible correction options. This analysis can be defined as follows.
Equipment Conditions
Equipment capabilities need to be understood, including unit pressure, tool stack, press alignment, thermal alignment of the process, downstream equipment conditions, etc. For example, a perfect profile shape can be deflected by physical contact with downstream equipment, such as worn-out roller sleeves.
Material Properties
Alloys and casting/homogenization affect material properties, including flow stress, differing thermal shrinkages, surface imperfections, etc.
Production Parameters
Extrusion velocity affects flow stress, metallurgical performance, and shape control. For example, higher speeds can cause tearing, while reduced velocity can cause portions of the profile to run slower versus other portions, and so on. This directly affects the profile quality issues. Similarly, billet temperature, puller tension and other production parameters can modify results.
Die Design and Conditions
Die design is one of the most critical factors in extrusion, as it determines the aluminum flow to every region of the profile and affects the relative velocity of every region of the profile. Differing velocities result in extrusion anomalies. Other critical die parameters include die bearing lengths and tapers joining differing lengths, undercuts, flatness (choke or relief of bearings), die wear, etc.
Nose Piece Velocities and Features
The emerging nose piece provides important information about the results from the prior four topics, by showing the flow of the alloy. By analyzing the nose piece velocities and features, die technicians can correlate the resultant profile quality issues with all other factors to determine how metal flow must be modified to improve the results.
Quality of the Extrusion Profile
The extrudate may have quality issues, such as dimensional or shape distortions, surface quality issues, or metallurgical performance variations, which are determined later. Die technicians must then use the insights gained from this data to develop the best strategy to drive improvements.
Die Correction Thought Process
Skilled die technicians follow a sequential thought process and develop a correction plan based on factors collected during the analysis phase. This thought process connects all the threads and uses their expertise to identify likely root causes and possible correction options.
Correlate the Data Collected During Analysis with Profile Defects
In this phase, die technicians carefully interconnect the equipment, process parameters, and features of the nose piece with the quality issues and can often understand the key factors involved and the resultant changes required in the aluminum flow pattern to correct the quality issues of the profile. They identify the appropriate root cause(s) that lead to resultant flow imbalances and develop a plan to proceed. For example, slower aluminum flow in some regions may lead to concavity in the profile, which can be mitigated by alterations to sink in recesses, bearing lengths, choke/relief, or other means to modify flow.
Identify Possible Correction Options
After processing the collected data, die technicians identify possible correction options. For example, they may consider options, such as leveling the bearings, reducing the bearing lengths by EDM, increasing the sink in recess depth or area by manual or CNC milling, or adding additional bearing relief by manual filing to increase the flow and rectify the profile quality issue. In some cases, die technicians may consider increasing or decreasing the flow in some regions to achieve better results.
Identify Most Appropriate Correction Option
Die technicians then analyze the advantages and disadvantages of various options and choose the most appropriate correction method that will sustain the correction, while also taking into consideration the available resources and their current availability. For instance, to increase aluminum flow, the technician considers bearing length adjustment as the most sustainable correction. However, at that moment, EDMs are occupied, and there is a long queue. Therefore, the technician examines the remaining options, such as adding relief or increasing sink pocket depth by machining. Since the die is new and all other bearing lands are still flat, the technicians may opt for manual milling. If there are no design features to increase flow or if the die has heavy usage and most areas have bearing imperfections, the technicians may consider bearing relief through manual filing.
Quantify the Correction
After finalizing the most appropriate correction method in the above phase, the technicians quantify the amount of correction by utilizing their previous experience and considering the current situation. If they want to run the die at the same speed as the previous run, they may need to increase the recess depth by 0.07 mm. However, if they feel that they can increase the press velocity due to a lower exit temperature during the previous run, they may consider that factor and increase the depth of the recess by 0.1 mm to alter the effect of flow imbalance at the upcoming production condition.
After finalizing the correction option and quantification, die technicians execute the correction. For certain manual corrections, the die technicians’ hand skills are critical. An AI system may help with the data collection, analysis, and even recommendation of correction options, but the industry still depends on human skills to execute certain aspects of the correction process. What AI can provide is a greatly reduced training time coupled with improved decision processes to guide the technicians to the appropriate solutions.
Evaluate the Results of the Die Correction Process
Next die technicians keep records on die correction details and follow the testing cycle previously discussed until the die is able to produce proper extrudate that will fulfill customer requirements at appropriate productivity and scrap levels. Recording the process will also provide a base of knowledge that will enable continuous improvement of the AI systems and the die correction process. The analysis phase and development of a correction plan discussed here is the most crucial aspect of die correction intelligence. It ensures that this critical intelligence is not lost when technicians retire and is instead transferred to new technicians.
Making AI for Die Correction a Reality
When thinking about an AI platform for die correction intelligence, there should be a very strong input module for the acquisition of input details of various equipment and process conditions, tooling design, and resulting dimensional/shape control and metallurgical aspects. Appropriate numerical input data systems must be employed to collect key process parameters, such as temperatures, speeds, alignment, scanners, and optical character recognition (OCR) to identify die design techniques and resulting tool geometry. Image processing capabilities to identify profile distortions are also needed, as well as an innovative capability to capture profile nose piece velocity details.
To input the correction thought process in an AI platform there must be a comprehensive methodology to archive all the information digitally and store it for further processing coupled with a strong categorization method to correlate all the input and resulting details. Once this categorization, quantification, and data storage is accomplished, AI can be implemented using a self-learning neural network to emulate the thought processes of skilled die technicians. Finally, there should be an innovative output module that is able to provide enhanced correction suggestions and recommendations.
Implementing such a platform may not be an easy task due to various issues including the lack of standardization, copyright-related issues, and difficulties in aligning die correction know-how to develop self-learning algorithms across those in the industry working together to create this AI solution. If the industry can align and develop such a system, it will help to both preserve and enhance die correction intelligence through improved processing capabilities, likely in an exponential fashion. The increased use of self-learning algorithms will lead to ever more accurate predictions, recommendations, and knowledge storage and transfer.
The AI system developed will by definition rely on extensive data related to the extrusion equipment, process, tooling, and resultant dimensional and metallurgical performance. It can serve as an excellent methodology to store, analyze, and monitor die-related performance data, allowing the recall of previous production runs and results at any time.
Editor’s Note: This article is the first in a series that will discuss the implementation of AI to enhance die correction intelligence. In the forthcoming articles, the authors plan to discuss the AI system’s features and challenges in more detail, as well as the possibilities of overcoming these challenges and establishing such a system.
This article first appeared in the June 2023 issue of Light Metal Age. To receive the current issue, please subscribe.