Production Planning
Assembly Line
Machine learning-based dispatching for a wet clean station in semiconductor manufacturing
The concept of cyber manufacturing has become a critical element in semiconductor fabrication environments, where automation and systemization are integral, for addressing the growing complexity of processes and facilitating predictive capabilities through data integration. This study deals with the dispatching problem to minimize makespan at a wet clean station in semiconductor fabrication using artificial intelligence-enabled manufacturing control techniques. The wet clean station is comprised of sequential chemical and rinsing baths for cleaning wafer lots and multiple robot arms for lot handling. In the station, wafer lots are sequentially immersed in several baths for cleaning to eliminate residual contaminants and stains that cause defects on wafer surfaces. The station can process various types of products, and the specific order of immersion differs depending on the product type. Unlike typical dispatching problems, the information required for dispatching, such as processing times and sequences inside the station, is not available. The only available data are historical logs that record when each lot enters and leaves the station. However, even when cleaning the same product type, the duration that lots spend in the station may vary based on the combination of product types being cleaned simultaneously and the settings of the station. Thus, using the time records, this study proposes a dispatching method based on machine learning models (multiple linear regression, deep neural network, and convolutional neural network). The proposed algorithms were evaluated and verified by comparing them with CPLEX solving a mixed integer programming and dispatching methods used in a semiconductor fab in Korea. Through this experiment, we observed that the proposed models can provide dispatching solutions that are practical and effective in a rapidly changing production setting. These models have the potential to enhance the capacity of a wet clean station and will contribute to artificial intelligence-based manufacturing system control.
A novel fine-grained assembly sequence planning method based on knowledge graph and deep reinforcement learning
In the assembly sequence planning (ASP) of aviation products, recalibration of components or sufficient space to assemble subsequent components are critical factors for ensuring product quality. To address this need, a fine-grained ASP (FASP) is defined to take assembly operations as units to plan sequences. Lots of operations have complex sequence constraints that are attended unequally in the FASP. A method based on knowledge graph (KG) and deep reinforcement learning is proposed to plan assembly operations. Firstly, continuous and discrete procedures are defined, and a quantitative characterization method is presented to deduce complex constraints objectively. Then, a dynamic KG is designed to establish and update the information model mainly composed of constraints. Finally, a labeled degree centrality algorithm (LDCA) considers edge labels to minimize the number of assembly tool changes and assembly direction changes for sequences. An improved deep Q-network (IDQN) introduces a convolutional layer to extract local features of technical requirements for planning procedures more efficiently. A helicopter structure assembly is used to verify the effectiveness of the proposed method. The improved algorithms have better performance in solving speed, sequence quality, and convergence speed than ordinary ASP methods, respectively. The fine-grained assembly sequence is more reasonable and feasible by comparing it with the ordinary sequence.
Why Small Fab And Assembly Houses Are Thriving
High-volume products get more than their fair share of attention in the semiconductor world, but most chips don’t fit into that category. While a few huge fabs and offshore assembly and test (OSAT) houses process enormous volumes of chips, small fabs and packaging lines serve for lower volumes, specialized technology, and prototyping.
Some companies, such as Analog Devices, are willing to split production between their own specialized lines and commercial foundries. When the device involves mainstream technology, the company outsources it. “If you look at Analog Devices, which is well known for small volume and high mix, a big chunk of their business is now outsourced to foundries like TSMC,” said Brosnihan. “They keep the more specialty stuff in-house.” That provides Analog Devices with more flexibility for the mainstream parts, sidestepping the need to build a fab and then keep it full.
NPI lots almost always are built in the same fab as the higher follow-on volumes for the re-qualification reasons mentioned above. One pair of Canadian companies has a unique model, however. A small firm called C2MI (Centre de Collaboration MiQro Innovation) handles prototyping and low-volume production. If volumes ramp beyond what it can manage, production transfers to Teledyne Dalsa, which intentionally keeps its equipment as close as possible to that used at C2MI to alleviate the requalification burden. These aren’t separate arms of one large company. “They are independent organizations that have a partnership,” said Brosnihan.
Tapping AI for Leaner, Greener Semiconductor Fab Operations
Flexciton’s software combines two conventional approaches to production scheduling: heuristics and mathematical optimization. “We begin with a relatively simple rule-based algorithm to find a good starting point for our scheduling problem,” Flexciton CEO Jamie Potter told EE Times Europe. “This initial solution is what we call a ‘warm start.’ Then we use mixed-integer linear programming to further optimize the problem from that starting point. However, this problem is often too complex for computers to solve within a practical time frame.
Flexciton uses a cloud-based hybrid scheduling model whose two key components are a global scheduler and a toolset scheduler. The global scheduler focuses on the broader objective of optimizing the overall manufacturing process at the fab level. It considers the big picture and aims to make decisions that maximize efficiency and productivity across the fab. The toolset scheduler works at a more granular level, focusing on the efficient operation of individual machines or tools within the fab. Its main goal is to ensure that wafers move through the fab machines smoothly and efficiently. It also strives to follow the priority ranking that was determined by the global scheduler.
To ensure that even the slightest changes in a fab element’s operation do not hamper the manufacturing process, Flexciton’s software runs its calculations every 10 minutes for the global scheduler and every three minutes for the toolset scheduler.
🧠🗓️ Explainable production planning under partial observability in high-precision manufacturing
Conceptually, high-precision manufacturing is a sequence of production and measurement steps, where both kinds of steps require to use non-deterministic models to represent production and measurement tolerances. This paper demonstrates how to effectively represent these manufacturing processes as Partially Observable Markov Decision Processes (POMDP) and derive an offline strategy with state-of-the-art Monte Carlo Tree Search (MCTS) approaches. In doing so, we face two challenges: a continuous observation space and explainability requirements from the side of the process engineers. As a result, we find that a tradeoff between the quantitative performance of the solution and its explainability is required. In a nutshell, the paper elucidates the entire process of explainable production planning: We design and validate a white-box simulation from expert knowledge, examine state-of-the-art POMDP solvers, and discuss our results from both the perspective of machine learning research and as an illustration for high-precision manufacturing practitioners.
Using ML For Improved Fab Scheduling
The exact number of available tools for each step varies as tools are taken offline for maintenance or repairs. Some steps, like diffusion furnaces, consolidate multiple lots into large batches. Some sequences, like photoresist processing, must adhere to stringent time constraints. Lithography cells must match wafers with the appropriate reticles. Lot priorities change continuously. Even the time needed for an individual process step may change, as run-to-run control systems adjust recipe times for optimal results.
At the fab level, machine learning can support improved cycle time prediction and capacity planning. At the process cell or cluster tool level, it can inform WIP scheduling decisions. In between, it can facilitate better load balancing and order dispatching. As a first step, though, all of these applications need accurate models of the fab environment, which is a difficult problem.
The GlobalFoundries group demonstrated the effectiveness of neural network methods for time constraint tunnel dispatching. The relationship between input parameters and cycle time is complex and non-linear. As discussed above, machine learning methods are especially useful in situations like this, where statistical data is available but exact modeling is difficult.
The setup matrix for production optimization
A setup matrix is a powerful tool for detailed optimization of production for recurring lots of similar or identical products. The challenge is to determine the transition times from one product to the next with sufficient precision and to provide the data for a planning system.
The Benefits of Production Stabilization and the Sorcery of the Product Wheel
Volatile demand is everywhere, and companies facing it typically choose between two options: One, attempt to meet demand as it arises (chasing the volatility). Two, maintain a certain inventory level as a buffer from volatility. Of course, there are situations where option one is viable, but option two is the one that most companies take. Still, pursuing the benefits of production stabilization, even in this environment, is worth the effort.
The product wheel is a framework for consuming capacity by making specific products – on a particular asset, in fixed quantities – over a defined time horizon. Therefore, the ability to populate the wheels with products that can conform to smooth production is essential. Determining which products work with this strategy and which don’t is an analytical effort requiring product segmentation, statistical forecasting, replenishment policy selection, and inventory parameter development.
The Next Revolution: Industry 4.0 in the Intelligent Enterprise
Which companies benefit from being able to automatically control the entire supply chain through machines and sensors?
“Auto-control” management means saving effort in manual processes along the entire supply chain and realizing the full potential of intelligent machines and sensors. Businesses in Europe in particular are creating opportunities here – their strength is traditionally more in customer-centric manufacturing, rather than mass production. However, standard products also benefit from flexibility. The global crisis of supply and logistics poses challenges for every manufacturer. Only those who dynamically parameterize production to deploy alternative materials and processes at the push of a button will win the global race for capacity and resources.
Part Level Demand Forecasting at Scale
The challenges of demand forecasting include ensuring the right granularity, timeliness, and fidelity of forecasts. Due to limitations in computing capability and the lack of know-how, forecasting is often performed at an aggregated level, reducing fidelity.
In this blog, we demonstrate how our Solution Accelerator for Part Level Demand Forecasting helps your organization to forecast at the part level, rather than at the aggregate level using the Databricks Lakehouse Platform. Part-level demand forecasting is especially important in discrete manufacturing where manufacturers are at the mercy of their supply chain. This is due to the fact that constituent parts of a discrete manufactured product (e.g. cars) are dependent on components provided by third-party original equipment manufacturers (OEMs). The goal is to map the forecasted demand values for each SKU to quantities of the raw materials (the input of the production line) that are needed to produce the associated finished product (the output of the production line).
Automotive – Production Schedule
Hercules is a system used by the Volvo Car Corporation (Volvo) for planning car production. Its main output is the monthly Master Production Schedule (MPS) – a detailed plan on how many cars are to be produced per market, factory and week during the following 13-15 months. Migrating such a comprehensive and vital system from an APL2 Mainframe to a Dyalog Microsoft Windows platform is no small task. Working with experienced and expert APL programmers from Aplensia in Gothenburg together with Mahindra Satyam, the project was completed in record time, with no disruption to the users or the operations, saving a substantial amount in the process. Peter Simonsson from Aplensia takes us through the process of the successful Hercules migration project, which was completed in the autumn of 2012.