Session: 02-03-01: Optimization

Paper Number: 144702

144702 - Implementing Knowledge- Based Engineering in the Design of the Electrical Enclosures. 

Electrical enclosures are devices which are used to control the flow of electrical power from main electrical lines into end user applications while providing protection from electrical faults and failures. The design of enclosures requires adhering to design standards such the IEC 61439 or other country dependent design codes for other considerations, thus such designs could vary depending on the manufacturer of the electrical enclosures. While adhering to the required design standards and rules, a design engineer must depend on several factors which includes the total cost to construct the enclosure, the parts inventory available, and the time constraint required to design the enclosure (the more parts that could be used from the inventory, the less time required to design and construct the enclosure). With increased number of enclosure manufacturers and varying customer's requirements. It is necessary to develop a system that could read customer requirements then decide on an assembly that minimizes the total cost of construction while still abiding by the required standards and codes. 

In this paper, it is assumed that customer's requirements are presented in schematic diagrams known as single line diagrams. Such diagrams show all the required electrical components, their rating, and the number of polls. Single line diagrams (SLD as they will be called) show the direction of flow of electrical power (top to bottom or bottom to top).  The approach will be limited to low voltage enclosures which use circuit breakers mainly in addition to supporting components such as lighting and metering.

The SLD is read using a DarkNET53 based Neural network trained on reading electrical symbols associated with the low voltage switchgear. The neural network was able to determine the associated component of the symbol and its orientation (top to bottom or bottom to top) as well as the number of components required. Text is read and analyzed using a CRAFT- based text analyzer where keywords are used to associate properties like rating and polarity to the detected symbols. The required components are then grouped together in groups depending on similar rating and polarity. Each group is arranged into subassemblies that are arranged as per a construction code. 

The approach is kept simple in approach by using the principle of parametric drawing to create sheet metal parts for the construction of the enclosure. This principle is used so that parts from 3 categories (which are structural parts, supporting parts, and external doors and covers) are constructed by providing the required external dimensions and material algorithm.

enclosures are assembled using a bottom-up approach where the algorithm starts by the grouping subassemblies of components together and arranging them so that the components are freely accessible and follow the minimum distances required by the aforementioned standards and codes. This is followed by ensuring that the overall sizes and thus cost is minimized for all the rest of the structural components. Stochastic optimization algorithms such as the particle-swarm optimization algorithm, were used to optimize the size of the enclosure. 

The algorithm was trained on a set of ~600 SLDs that were randomly generated in such that no two are identical. In addition, it was test ran using another set for validating the data. The corrected algorithm was later tested using two single line diagrams where the algorithm was able to construct a full enclosure in about 1 minute. 

Presenting Author: Mohammad Elzein Kuwait University

Presenting Author Biography: Mohammad Elzein is a Mechanical Engineer who received his BS and MS in Mechanical engineering in Kuwait university, and previously worked in the construction of low-voltage switchgear for 3 years. His research topics include knowledge-based engineering and parametric drawing.

Authors:

Mohammad Elzein Kuwait University
Raed I. Bourisli Kuwait University