Features and innovations
Feb 2024
5 min.
Features and innovations
Summary
"Warehouse Healing" defragments the warehouse and reduces travel times in the warehouse for man and machine through better positioning of items. This is based on the intelligent analysis of relevant data such as stock levels, topology, movement data and shopping baskets - the data basis is provided by S&P's Warehouse Management System SuPCIS-L8. The goal is to analyze patterns in the order history and to generate stock transfer suggestions to minimize picking paths. Special emphasis is placed on the shortest possible time-to-value.
Algorithms are used to identify and determine relocation and exchange proposals with the greatest effect, which are assigned the highest priority with a view to implementation. Based on artificial intelligence and simulations of changed model parameters, the result is constantly adjusted to the changed circumstances in order to minimize the sum of the stock removal costs. After just a few hundred stock transfers, the labor-intensive process of picking is optimized in this way. An example of this is a multi-story shelf system, where the use of "Warehouse Healing" can ideally reduce walking time by up to 40%. The resulting increases in performance, optimal utilization of workflows, and sensible resource planning in the logistics centers ensure efficient and sustainable (intra)logistics at this point!
Introduction
Studies show that walking time during picking in typical distribution centers accounts for a significant portion of the total order processing time. If picking is done manually, this can amount to more than 50% of the total time. Increasing the efficiency of intralogistics processes, especially in the area of labor-intensive order picking, is therefore in the interest of a sustainable design of future flows of goods.
As a result of rapidly changing product portfolios and changes in customer purchasing behavior, warehouse allocation that is initially optimized for efficient picking can quickly lose its effect. Once fast-moving items can become slow-moving items within a few months, blocking valuable storage space for items that are new to the trend. Instead of subjecting the warehouse to a complete reorganization at irregular intervals, which interrupts operations, a few interspersed stock transfers during the operational warehouse business can be used to react to changes in ordering behavior and retain the advantages of chaotic warehousing.
Warehouse healing is a concurrent process that reduces the average stock removal distance by selectively transferring selected stock within a storage area or across storage areas. In contrast to a complete warehouse reorganization, only a few stock transfers per week are necessary, since only stock transfers with the greatest path cost benefit in each case need to be acted out. The stock transfer proposals are not only based on concepts such as storage according to ABC classification or a simple turnover orientation, but also take into account patterns in the ordering behavior of customers. These patterns are made accessible by a shopping cart analysis and converted to a handy neighborhood of articles, which show a strong shopping cart relationship, in the warehouse.
Motivation
Due to the day-to-day activity in a warehouse, the lack of ongoing bin evaluation (which items should be stored where) makes distributions like the one in the following diagram likely (Figure 1, left).
The warehouse heatmap shows an example of a warehouse with 3 aisles, whose shelves have been placed with three levels for a better overview. The order frequency (movement) per storage compartment (squares) is represented by the color saturation of the red tone. The spatial distribution of this article attribute does not follow any regularity in case of a chaotic storage strategy.
Articles can be related to each other. Articles that are frequently part of a goods order are considered to have an affinity to each other. Since article affinity is not taken into account in the normal case of chaotic warehousing when assigning an article to a storage bin, the spatial distribution of affine articles among each other follows chance (see exemplary selection of affine articles in Figure 1 left).
Various studies have shown that storing affine items in adjacent storage bins leads to an advantage when picking these items. In addition, the ergonomics at the workstation are improved by the increased ease of grasping related shopping cart items.
The strong fragmentation of a warehouse caused by chaotic putaway with regard to the item property order frequency and shopping cart relationship leads to significantly longer retrieval paths and an enormous effort when picking the purchase orders.
The aim of warehouse healing is to achieve both a favorable position of the items with regard to the transfer location (picking location, put location, packing location, etc. - marked with "I/O" in Figure 1) and a neighborhood that is conducive to picking, in order to noticeably reduce the picking effort in this way.
A - strongly idealized - example allocation as a result of using this rearrangement module can be seen in Figure 1 on the right.
The Warehouse Management System SuPCIS-L8 provides the necessary data ad hoc
A number of pieces of information are needed to evaluate a relocation proposal:
This information is retrieved from the productive warehouse management software SuPCIS-L8 database and, where possible, already prepared at database level for the shopping cart analysis. Of decisive advantage is that SuPCIS-L8 inherently creates a detailed history of all inventory movements - a fertile breeding ground for various applications around data mining and a prerequisite for any system that requires extensive training data.
Thus, the warehouse management software SuPCIS-L8 also keeps all necessary information available for warehouse healing, so that there are no further dependencies on other installations. The customer can concentrate solely on the operational running of the warehouse and the WMS provides the necessary data.
Warehouse query segmentation powerd by Data Science
The formulation of a suitable mathematical model enables the evaluation of stock transfer options. A stock transfer option results as an exchange of two stock placements in the considered storage area. A stock placement, together with a physical coordinate system designed for the storage area under consideration, is to be understood as a position in space.
By swapping two stock placements in pairs, there is a counter movement for each movement. If these stock transfer options were carried out aimlessly, i.e. unvalued, this would not change the situation fragmentation (see Figure 1 left). Therefore, the evaluation of each option is necessary. For this purpose, a so-called score is defined, which enables a cross-storage area evaluation of all putaways. The usefulness of a pairwise article exchange can subsequently be calculated via the change in the score conveyed by this.
The combination of retrospective analysis of past orders (shopping cart analysis) and mathematical modeling of the order picking process makes it possible to implement a system that learns from past data and derives suggested actions in the form of favorable stock transfers for future order fulfillments.
Rapid benefit analysis through proof of value
In order not to stress interested parties for Warehouse Healing with too abstract concepts, S&P offers the evaluation of the expected benefit for a concrete warehouse within the scope of a small pre-project. With the Proof-of-Value the often high hurdle in the run-up to a project approval is reduced.
The goal of a proof-of-value project culminates in the statement: The application of warehouse healing in warehouse area X can reduce the retrieval paths by Y % on average.
After determining to which storage areas the healing is to be applied, the picking model is created. For this purpose, semi-automatic floor plan information is evaluated, a coordinate system of all storage compartments is created and linked to the inventory data from the WMS. Furthermore, all peculiarities of the customer are taken into account: Which aids, e.g. picking carts, are used? In which way can the shelves in the warehouse be run? What is the picking strategy, is multi-order picking used? To name just a few aspects.
The WMS also provides the comprehensive history of all inventory movements in the warehouse being analyzed. A shopping cart analysis reveals long-term and short-term patterns in past order activity and incorporates them into the customer-specific score function.
The result of such a benefit analysis with customer data from practice is shown, among others, in Figure 2. The L-shaped manual picking warehouse of a technical wholesaler is shown in plan view. Deviating from figure 1, only the representation of the first level is necessary, since picking is only done from this shelf level. The coloring of the shelves corresponds to the picking frequency of the items stored there. Chaotic storage can be recognized by the lack of order: Frequently picked items and their dark red shelves can be seen both near the transfer location I/O and far away from it.
The same applies to the exemplary selection of an article and its green-shaded shelf as well as its affine shopping cart items - stored in the green-shaded shelves. Picking this frequently occurring shopping cart thus requires walking through a large part of the warehouse and thus makes a considerable contribution to the stock removal effort of future order jobs.
The blue and green tracks show the best 300 healing suggestions that Warehouse Healing has identified in a first run and reduced the outsourcing path with maximum benefit. The customer decides for himself if and which of these suggestions will be sprinkled into the daily operational business at which time. In a further illustration, we will see that of these 300 suggestions, the first 100 are particularly worthwhile and the rest can be disregarded.
In order to extract the maximum benefit and thus the greatest possible reduction of the retrieval path for each individual stock transfer, a parameter analysis is performed as part of the proof-of-value. Here, the degrees of freedom in the parameterization of the underlying model are used to achieve the best fit of the warehouse healing module to the warehouse. The result of such an analysis results in diagrams as shown in Figure 3 on the left.
Real customer data here shows how a particular model parameterization causes the stock-out path curve to point downward most steeply in relation to the number of healing transfers performed.
The healing module maximizes the benefit when the consideration of order frequency and shopping cart induced item affinity is in the ratio 1:3.4. In this configuration, the application of the healing module promises the reduction of the stock-out distance according to the green curve in Figure 3 on the right. The diagram also contains, with the red curve, the answer to the question of the advantage of including market basket analysis over simple stock transfer based on inventory turnover. If the module learns from the historical order entries, a walkway reduction of 15% can already be achieved after 38 stock transfers, whereas 90 stock transfers are necessary if the shopping cart specification in the history is ignored.
Conversely, warehouse healing achieves a 20% reduction in the distance traveled after just 90 stock transfers, which is never possible with a purely stock transfer-oriented approach. It is also clear that from 100 stock transfers onwards, a saturation of the stock transfer advantage occurs, according to which it is sufficient to implement a maximum of 100 stock transfers.
Weighing the effort and resources available for the relocations, the proof-of-value concludes with the proposition:
The application of Warehouse Healing in the ML1 storage area can reduce stock-out distances by an average of 20-25%.
Do you want to reduce your travel times with the help of AI and would like an evaluation of the expected benefits for your warehouse? Feel free to contact us!
