	US 12,436,511 B2
	Method and system for semi-automatic completion of an engineering project
Marcel Hildebrandt, Munich (DE); Serghei Mogoreanu, Munich (DE); Mitchell Joblin, Surrey (CA); Martin Ringsquandl, Raubling (DE); and Chandra Sekhar Akella, Madison, AL (US)
Assigned to Siemens Aktiengesellschaft, Munich (DE)
Filed by Siemens Aktiengesellschaft, Munich (DE)
Filed on Feb. 23, 2023, as Appl. No. 18/113,224.
Claims priority of application No. 22159221 (EP), filed on Feb. 28, 2022.
Prior Publication US 2023/0273573 A1, Aug. 31, 2023
Int. Cl. G05B 13/02 (2006.01)

CPC G05B 13/027 (2013.01)

10 Claims

1. A computer implemented method for semi-automatic completion of an engineering project, comprising the following operations, wherein the operations are performed by components, and wherein the components are software components executed by one or more processors and/or hardware components:

storing, in a database,

a set of items, I, with each item, i∈I, having technical attributes, x_i, and with each item representing a module, and with each module being a hardware module and/or software module that can be used in an engineering project of a first user, u₁,

a current state of the engineering project, with the current state including items that the first user, u₁, has already selected, and

a user-item interaction graph

representing each user, u, of a set of users, U, and each item, i, of the set of items, I, with a node, and

representing observed interactions between users and items by edges between the respective nodes,

producing, by a feature encoder retrieving the technical attributes, x_i, of each item, i, from the database, an initial item embedding, e_i⁰, for each item, i, which is a latent representation of the respective item, i, in latent space, with all of the initial item embeddings, e_i⁰, forming an initial item embedding matrix, I₀,

processing, by a graph neural network operating on the user-item interaction graph,

an initial user embedding for the first user, e_u_₁⁰,

a first user neighborhood, N_u_₁, and

the initial item embedding matrix, I⁰,

as input to compute an updated user embedding for the first user, e_u_₁^L,

with the initial user embedding of the first user e_u_₁⁰being a latent representation of the first user, u₁, in latent space,

with the first user neighborhood, N_u_₁, being a set of items that the first user, u₁, has interacted with,

processing, by the graph neural network operating on the user-item interaction graph,

each initial item embedding, e_i⁰,

a neighborhood for each item, N_i, and

an initial user embedding matrix, U⁰,

as input to compute updated item embeddings, e_i^L,

with the neighborhood for each item, N_i, being a set of users that have interacted with the respective item, i,

with the initial user embedding matrix, U⁰, being formed by initial user embeddings, e_u⁰, for each user, u, which are latent representations of the respective user, u, in latent space,

processing, by a decoder mapping, the updated user embedding for the first user, e_u_₁^L, and the updated item embeddings, e_i^L, as input in order to compute a recommendation score s_u_₁_,ifor each item, i, indicating a likelihood that the first user, u₁, needs the respective item, i, for completing his engineering project,

forming a set of items by selecting items with the highest recommendation score s_u_₁_,i,

outputting, by a user interface, the set of items to the first user, u₁,

recognizing, by the user interface, a user interaction selecting an item from the set of items, and

automatically completing the engineering project by adding the selected item to the current state of the engineering project in the database.