Streamlining supply chain planning with AI and automation

Modern manufacturing plants are designed to produce a wide range of different products on multiple production lines. This flexibility makes production planning a complex problem. Additionally, scheduling challenges increase exponentially as constraints are added. This complexity implies that such problems cannot be solved using traditional manual methods. Instead, they require intelligent AI approaches to find an optimal solution. 

Supply chain planners need a schedule that matches production targets, respects operational and technical constraints, and achieves set objectives, such as maximising throughput or minimising costs. 

A key consideration is the cost of alternating between the production of goods on the same line, due to equipment reconfiguration, line cleaning, etc. This is often the biggest issue facing production planning. One solution favours long product runs, but decreases operational flexibility while raising inventory holding costs. Shorter runs increase flexibility, decrease inventory cost and increase supply chain resilience, but increase changeover costs. 

So, how short should each run be, and in what sequence? What happens if multiple lines can make the same product, or if there are multiple lines to make each product?

Why a traditional static product wheel approach doesn’t work

Product wheels are one solution to the production planning problem and have been applied to a wide variety of industries including FMCG, food, and chemicals. A static product wheel has the following features:

• Regularly repeating, fixed production sequence, optimised to minimize changeover cost / time. Product groups are always made in the same sequence.
• The time for one complete cycle is relatively fixed. Overall wheel time is divided among products based on relative demand. Higher-volume products have longer duration “spokes”; lower-demand products have shorter duration spokes.

Product wheels address three issues facing operations managers:

• Minimise production costs while meeting demand. 
• Keep the production plan stable.
• Schedule different product groups on the same line.

These constraints are “hard” or “soft”. Hard constraints are rules that cannot be broken. Soft constraints may be flexed (e.g. safety/max stock levels). 

However traditional static product wheels are not a panacea:

• They work best when demand is predictable and non-seasonal. If demand varies, then the fixed pattern becomes inflexible and demand may not be met.
• It does not take into account unexpected downtime or delays in supply of raw materials.
• Product wheels are suitable for a stable portfolio of products. If the product mix changes, i.e. new products introduced or old products retired, then the product wheel must be redesigned.
• Designing the optimal product wheel requires experience and judgment to achieve a usable product sequence on a shared resource. It requires accurate demand forecasts and changeover estimates. It should be seen as a heuristic approach, rather than an optimisation technique.

A modern solution to food production planning

The Replan production planning solution takes the best parts of the product wheel (e.g. managing competing constraints), while relaxing the requirements for fixed patterns, and uses AI to rapidly replan production and generate dynamic production sequences, as demand or constraints change. This means that the solution is flexible and can easily handle changes. Our solution can:

• Handle complex group planning calculations, while your existing systems continue to deliver core planning functionality.
• Work with any planning platform (SAP, o9, Kinaxis, Oracle, Infor).
• Be delivered as a SAAS solution or as a fully managed service.

Our cloud-based platform augments existing ERP supply chain solutions enabling production data (product groups, line capacity, product run rate, changeover cost and demand forecast) to be ingested from your ERP via APIs. Importantly, Replan has integrations with IBP & APO and can use SAP Cloud Platform Integration (CPI) Data Services. 

By taking in all the relevant constraints and using advanced algorithms and AI to rapidly deliver an optimal solution, Replan redefines production planning. The software supports three families of optimisation algorithms: exhaustive search, construction heuristics and metaheuristics. Under the hood, it combines sophisticated AI optimisation algorithms (such as tabu search, simulated annealing, late acceptance and other metaheuristics) with an efficient score calculation and other constraint solving techniques. 

We have translated the production planning business requirements into relationships and constraints to deliver an efficient supply chain planning solution. This vastly simplifies how a planner interacts with the software, so they don’t need any sophisticated AI training.

Other features of the Replan solution include:

• Multi-line planning including constraints for product groups on dedicated lines, off-loading excess production to other product lines, etc.
• Multi-level (stage) group planning. This is where there are multiple stages in production, e.g., mixing and packaging, which have different resources with different changeover implications.
• Co-production, i.e., two products that must be made simultaneously on a production line.
• The Replan solution has automated notifications and end-to-end process analytics, comparing multiple scenarios against safety and max stock thresholds.

The Replan production planning solution can return the following outputs to planning engine:

• Fixed Cycle – Fixed Group Quantity: Group production plan is loaded as a fixed plan onto Group Products. Use together with group planning functionality.  Constrained planning engine will manage the mix of the SKUs whilst still keeping the group volume and sequence fixed.
• Fixed Cycle – Flexible Quantity: Use Planning Calendar – Apply planning Calendar to every SKU in the group. Constrained planning engine will respect group sequence but has flexibility to respond to changes in demand.

Unconstrained Planning Engine: Adjust location product inventory target proportionally based on group stock projection.  Planning engine heuristics / MRP run will plan with new targets in line with group sequence and available capacity.