New Specification for Additive Manufacturing-General Principles-Registration of Process-Monitoring and Quality-Control Data
1. Scope
Increasingly, Laser-based Powder Bed Fusion of Metals (PBF-LB/M) Additive Manufacturing (AM) fabrication machines are being instrumented with various types of nondestructive sensors. Typically, each sensor is designed to collect only one type of measurement dataset and in a unique coordinate system. Consequently, effectively monitoring PBF-LB/M fabrication processes and qualifying the resulting AM metal parts require the use of all the sensor datasets simultaneously. This, in turn, requires multi-modal dataset registration, which includes data alignment. Registration is needed to extract and characterize process signatures, which are needed to control variations in powder spreading, melt-pool geometry, thermal stability, layer integrity, and part quality. Registration of these datasets covers meta data, data cleaning, data correction (due to distortion), data identification, and data alignment. Terms (e.g., data, metadata, dataset, data registration, coordinates, and transformation) and definitions related to data registration are also within the scope. The following types of data are out of scope: cost, production time, and personnel. Industrial AM users are just now beginning to understand the importance of multi-sensor, dataset registration in controlling their processes and qualifying their parts. This proposed standard practice comprises actions that users need take to register those datasets and store them in a repository. In general, each dataset is associated with specific functions in the product lifecycle. This standard practice focuses on two of those functions: build and inspect. The datasets associated with those functions include sensor data, scan commands, photographic images, thermal images, video clips, X-ray computed tomography 2D images and 3D models, CAD models, CMM point clouds, acoustic signals, accelerations, build information, part property data, material microstructural data, mechanical testing data, build data, equipment data, and production management data. As for aligning, both temporally and spatially aligning different data are required. Spatial alignment is a process that converts sensor data from its original, local coordinate system to another coordinate system. Alignment is needed so that different data sets can be compared and fused together correctly.
Keywords
additive manufacturing data; data alignment; data modeling; data registration; data set
Rationale
The measurement devices used for monitoring laser powder bed fusion (LPBF) processes and inspecting the resulting AM parts have been increasing, in both number and types, in recent years. The variety and volume of the data collected by such devices have increased exponentially. Each measurement device collects a unique data type, including text, values, images, and videos. A common, open method to collect, process and organize these different AM data types is needed so that their uses can be identified for downstream applications, including qualifications, certifications, and analytics. More specifically, this standard focuses on functions that 1) monitor and control LPBF processes, 2) predict the material properties of the final part, and 3) qualify other well-known part-quality requirements. The reason for that focus is simple: these functions require registered data sets to be executed accurately. The specific data objects that need registration include in-situ photogrammetry and thermography and ex-situ X-ray computed tomography (XCT). These data sets are generated from a variety of sources, including melt-pool images, scan paths, layer images, and XCT three-dimensional (3D) models. The potential users of this new standard include AM technologies users, sensor developers, researchers, and software developers. The benefits of registering data include 1) accessing validated data with known time, locations, and approvals, 2) data alignment and fusion for process monitoring and control, 3) detecting defects traceable to process, material, equipment parameters, 4) AM part qualification, and 5) understanding AM process for developing predictive models. There are three potential impacts of the proposed standard. First, more software tools for data analytics and AM product lifecycle engineering can be used easily. Second, more robust validation and certification processes for aerospace and medical industries will be available. Third, cost savings in data management, curation, and access will increase.
