Optimising the bill of materials is a constant work in progress. Because of articles that supplier sometimes changes, sometimes evolves, sometimes dismisses. On the other hand, you also want to cut costs and adapt some of your items to more standardised ones.

If we look at this process more in the abstract, it comes out that it is a typical application of a machine learning pattern. Let’s see how the implementation could help the production.

The process

The bill of material is only part of the process of production. The production capacity, the degree of parallelism, the shared lines of production, the workforce, and machinery maintenance are crucial information for optimising the production flow.

The historical information, in addition to the possible item substitutions or adaptation, based on the production parameters, allow the machine learning algorithm to create simulations connected to goals: cost reduction, faster production, waste reduction, quality elevation and faster order fulfilment (reduction of dependency between departments).

Self-production or purchasing?

The machine learning algorithm gives further help in decisions. It calculates from all the known information whether it makes sense to produce an item or better purchase it. And the other way round. In fact, in some cases, self-produced parts are worth the effort.

The algorithm can calculate the break-even point where the ROI (Return of Investment) becomes profitable. In some cases, an investment also brings subsequent opportunities to share benefits (i.e. new employees, new machinery capable of producing other items).

Guided decisions

When the orders are enough to initiate a productive process, the ML can suggest whether to overproduce items, stock them, or stop production after reaching the required amount. It gives the metrics to choose the quantity to stock.

On the other side, when the ordered items are not enough to begin the flow, the algorithm provides the calculated information regarding the costs and benefits for all the possible decisions: start the production regardless, purchase items, or wait for other orders (and delay the current ones).

Conclusions

In all the phases of this area, AI offers help in decision making. The parameterised metrics allow the operator to give a measurable value to decisions. What before was decided by the gut, now AI gives it quantifiable values of reference.