Data science, which is the fourth paradigm in natural science, is now becoming a major technique to investigate natural science. The importance of data science in Tribology is large since, in general, Tribological data contain many aspects and it is complex phenomena. In this article we describe the paradigm in historical and social point of view, then explain the basic method of analysis, and show some topic using data science analysis. Not only for tribo-material development but also for many application in Tribology, data science will boost their future growth.

In this article, a newly developed universal neural network potential (NNP), enabling accelerated molecular simulations, and its application to tribological phenomena were reported. The NNP is based on a graph neural network theory and an original dataset including a huge number of the first-principles calculation results. Two application examples using this method were introduced. One was exploring geometry for a fundamental molecular adsorption of long-chain fatty acid, which is difficult to deal with the conventional first-principles calculations. This new technique successfully observed that coverage of fatty acids on metal surface largely influence on the adsorption structures, i.e., orientation angle and intermolecular interaction of long-chain alkyl group. The second application was a molecular dynamics simulation for understanding influence of base oil structures on adsorption behavior of an oiliness additive. The adsorption time of an oiliness additive in three types of base oil (normal paraffin, isoparaffin, and naphthene) was measured. It was found that an oiliness additive in isoparaffin phase showed the earliest adsorption time among the tested base oils. This study unveiled the adsorption mechanisms of the oiliness additive in the base oils: an oiliness additive can smoothly pass inside the isoparaffin-derived oil film with sparse structure and easily reaches metal surface. We concluded that a rapid tribo-material discovery, as well as analysis on molecular mechanism of tribological phenomena, is expected to be achieved by adapting universal NNP to molecular simulations.

In recent years, artificial intelligence (AI) technologies have been developed dramatically, leading to AI applications and material design in many industrial fields. In the field of tribology, several attempts have been conducted to estimate the friction coefficient and wear, while AI has not been as widely used as in other fields. In the present report, three topics are presented: friction coefficient estimation technology from optical images of friction surfaces using deep learning methods, wear estimation technology, and a novel functionality to avoid contact with damaged positions on the mating surface by combining morphing surfaces and their AI-based control. In AI-based estimation techniques, many engineers and researchers are faced with the issues of interpreting estimation results. The field of interpretable AI has been also developed drastically. On the other hand, it is difficult for engineers and researchers who do not specialize in AI to keep up with the latest advances in the AI technologies. Therefore, collaboration research projects with AI specialists are becoming increasingly important.

In industrial field, computer-aided engineering such as CFD (computational fluid dynamics) simulation becomes essential to the optimal design of components and assemblies. Due to the increasing computing power and the improvement of simulation accuracy, Big Data are created during the design process. These Big Data can be utilized by machine learning for the future design. The author adopted machine learning for the design of surface textured mechanical seals. In order to design a surface texture and fulfill requirements such as friction coefficient and leakage, hydrodynamic lubrication calculation is used. The shape is optimized based on genetic algorithm for each operating condition. In this optimization process, huge amount of simulation data are created. Firstly, DNN (deep neural network) has been adopted to model the correlation between the shape parameters (input) such as texture length and depth, and evaluation values such as load carrying capacity (output). For validation of the DNN model, shape optimization based on genetic algorithm is performed where hydrodynamic lubrication calculation is replaced by DNN model. Secondly, in order to create widely applicable model, a U-Net convolutional neural network architecture is adopted to estimate the pressure distribution (output) from the mesh data of surface texture (input). Finally, recent trends in utilization of machine learning for general CFD are discussed.

Machine equipment usually comprises many mechanical elements that can fail because of functional deterioration and friction. For tribo-elements like plane bearings, it is extremely important to diagnose the abnormal conditions and prevent such parts from breakdown caused by wear. However, diagnosing tribo-elements requires expensive diagnostic equipment and expertise. A cost- and time- effective system that can detect the signs of breakdown during equipment operation by using machine learning to identify abnormalities are required. In this paper, I will focus on the tools and ideas used in the research conducted by the author's research group on technology related to the condition monitoring of sliding surfaces of plain bearings using machine learning. In particular, I describe the application of AI to friction testing and analytical ferrography. It also outlines AI utilization cases and trends in the field of maintenance tribology.

At the cutting process scene, operators usually change cutting tool based on experience and expertise. Therefore, even if true tool life is not reached, operators may change the cutting tools, which hinders cutting cost reduction. If the cutting tool state (cutting tool wear) in cutting process can be visualized, operators can change the cutting tool at the optimum timing regardless of operator skill, which leads to cutting cost reduction. This paper describes methods for estimation of the cutting tool state by analyzing acceleration data acquired in milling process, which methods use machine learning as a fundamental technology for real-time monitoring of its tool state in process.

This paper reports a novel method to measure the very small amount of wear that occurred at roughness peaks of the underlying base steel exposed on the surface covered with a soft material film removed by sliding. It is impossible to detect the minute wear of the exposed roughness peaks by a known method so far. Image data with three-dimensional height information obtained by a laser microscope is used to detect the position and height of roughness peaks of the underlying base steel exposed on the surface. This method makes it possible to track the minute wear progress using the mean height of all the detected roughness peaks or the wear of detected specific roughness peaks. It was also reported that this method has generality to be applied to untreated steel surfaces. It was possible to detect tiny distinctions of wear by this method. Then it was clarified that the wear of steel exposed on the MnP-treated surface is greater than that of the untreated steel, especially it was remarkable in the initial stage of sliding.

A containment seal (CS) can be applied to a flashing hydrocarbon pump used in the oil refinery and petrochemical industry as a sealing device to prevent environmental pollution. CS has great difficulty in design to ensure the contradictory functions of fluid-tight sealing and sealing surface wear resistance, which are required in both conditions of dry-running in gas and wet-running in liquid. In our previous report, a new CS design concept “CSAM”(Containment Seal by Additive Manufacturing) was developed utilizing additive manufacturing to introduce fluid pressure into sealing surfaces through slots and internal cavity formed in sealing ring. In this report, a new design index and design guideline were considered to control opening force and pressure distribution on sealing surfaces by slot arrangement in both gas and liquid conditions for CSAM. A series of numerical analyses and test results were consistent with the theoretical analysis and case study by optimization analysis according to suggested design index and design guideline for CSAM. Thus we conclude that the novel CSAM can be applicable as CS with the functions of sealing and wear resistance in both gas and liquid conditions.