Manufacturing Bits: Feb. 4

Non-targeted analysis
Using a technology called machine learning, the Southwest Research Institute has introduced a software tool that detects known and unknown chemical components in food, air and drugs.

It detects compounds in products we are exposed to every day using both machine learning and metrology techniques. A subset of artificial intelligence (AI), machine learning uses advanced algorithms in systems to recognize patterns in data as well as to learn and make predictions about the information.

Southwest Research Institute, meanwhile, combined machine learning and a technology called gas chromatography mass spectrometry. Mass spectrometry identifies known and unknown chemical components in a sample.

Traditionally, using mass spectrometry, the industry sometimes makes use of a targeted analysis approach to detect chemical components. The tool will screen the chemical components against a library of known compounds, but this method sometimes falls short and doesn’t cover the entire chemical space.

This in turn is driving the need for non-targeted analysis (NTA) methods, which reveal any chemical compound in a sample.

That’s where Southwest Research Institute’s technology fits in. The software tool, called Floodlight, uses deep learning algorithms to perform NTA of datasets from mass spectrometry.

Floodlight enables the rapid characterization of chemicals in an array of consumer products, such as food, medicine, packaging and toys. “Known chemicals in a particular sample are relatively easy to find and quantify,” said Kristin Favela, an analytical chemist at Southwest Research Institute. “NTA is a different story. Through an extensive, multiyear NTA program we discovered only about 20% of identified chemicals were included on the consumer product list. The remaining 80% were previously unidentified in these products.

“Consider the mysterious health effects we are seeing from e-cigarettes right now,” Favela said. “Currently, the medical community does not know what’s causing these serious problems, but based on our research, there are likely unknown chemicals present in any given formula.”

Keith Pickens from Southwest Research Institute added: “The key to Floodlight is artificial intelligence and machine learning algorithms that enable advanced analysis of chemistry big data. Floodlight is a sophisticated software tool that can make sense of the vast amounts of data NTA generates.”

AI crystallography
The University of California at San Diego has developed a machine learning algorithm that could make it less labor-intensive to determine the crystal structures of various materials and molecules.

Using this method, researchers are able to identify the crystal structures in materials with at least 95% accuracy. Applications include alloys, proteins, pharmaceuticals, among others.

Researchers from UC San Diego combined machine learning with a conventional scanning electron microscope (SEM). Using the SEM, researchers collected electron backscattered diffraction patterns.

Typically, a SEM uses commercial software to collect diffraction patterns. This in turn can determine crystal orientations, residual stress or strain and other data in a material. But at times, this process is complex and not always accurate.

By combining a SEM with a deep neural network, researchers can analyze diffraction patterns to determine the crystal lattices in structures with an accuracy greater than 95%.

Finding icebergs
The University of Sheffield has developed a machine learning model to predict the number of icebergs that will drift into shipping lanes.

In 2020, the model predicts that a relatively low number of icebergs will drift into shipping regions in the north-west Atlantic. The model predicts that between 479 and 1,015 icebergs will reach waters south of 48°N, compared with 1,515 observed last year, according to University of Sheffield. This is the area of greatest risk to shipping between Europe and north-east North America.

Grant Bigg, a professor at the University of Sheffield, said: “We have issued seasonal ice forecasts to the IIP (International Ice Patrol) since 2018, but this year is the first time we have combined the original control system model with two artificial intelligence approaches to specific aspects of the forecast. The agreement in all three approaches gives us the confidence to release the forecast for low iceberg numbers publicly this year – but it is worth remembering that this is just a forecast of iceberg conditions, not a guarantee, and that collisions between ships and icebergs do occur even in low ice years.”

John Wardman, a senior science specialist at AXA, said: “The impact of sea level rise on coastal exposure and a potential increase in Arctic shipping activity will require a greater number and diversity of risk transfer solutions through the use of re/insurance products and other ‘soft’ mitigation strategies. The insurance industry is keeping a keen eye on the Arctic, and this model is an important tool in helping the industry identify how or when the melting Greenland Ice Sheet will directly impact the market.”

The model was funded by AXA, an insurance firm.