SBIR Phase I: A Blockchain-Driven, Distributed Memory, Computational Platform for Industrial Analytics

这个小企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是为需要低延迟分析洞察力的工业企业提供一种新颖的预测分析技术，也就是说，那些无法负担几乎所有基于云的分析解决方案中广泛存在的那种延迟洞察力的公司。大型制造和能源企业由分布在不同地理位置的多个工厂组成。这些工厂包含使用传感器和物联网（IoT）设备监控的关键流程和设备。预测分析是数字化转型的关键组成部分，其目标是分析大量工业数据，从而产生见解，以提高工业运营效率，优化流程，并降低关键设备和机械的生命周期成本。当今市场上的主流解决方案依赖于使用云来整合数据、执行分析和提取有价值的见解。这个过程对许多需要立即了解其关键操作的工业过程造成了严重的延迟。这个过程也不适合那些已经加强了安全和隐私协议的公司（比如核电站和国防制造业）。该公司旨在使公司能够在不需要从其位置移动数据的情况下对地理分布的数据孤岛进行预测分析，从而减少决策延迟。SBIR第一阶段项目建议开发一种技术堆栈，利用区块链在不同位置的数据孤岛上训练先进的分析算法和机器学习模型，而不依赖于云或任何企业服务器。具体来说，该项目的目标是区块链智能合约与分布式内存编程框架（如消息传递接口（MPI））的创新集成。这种集成提出了几个有趣的研究挑战，其中一个挑战要求为区块链设计功能类似Map-Reduce框架的基本原语。第二个研究部分围绕着为流行的机器学习算法和人工智能（AI）模型开发新的算法分解方案，以促进使用区块链以分散的方式进行训练。如果成功，这项技术可能会比云更快地提供分析见解。它还可以显著降低与实现和部署工业分析相关的成本和人工。数据科学团队将获得修改和重新部署算法的敏捷性和灵活性，而无需重新构建和重新设计数据管道基础设施到集中式服务器。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to provide a novel, predictive analytics technology for industrial enterprises that require low-latency analytics insights, i.e., companies that cannot afford the kind of delayed insights that are widely present in almost all Cloud-based analytics solutions. Large manufacturing and energy enterprises are comprised of multiple plants sites spread across different geographical locations. These plants contain critical processes and equipment that are monitored using sensors and Internet of Things (IoT) devices. Predictive analytics, a key component in digital transformation, targets the analysis of large volumes of industrial data to generate insights to improve the efficiency of industrial operations, optimize processes, and reduce the life cycle costs of critical equipment and machinery. The prevailing solutions in today’s market rely on using the Cloud to consolidate data, perform analytics, and extract valuable insights. This process creates significant delays for many industrial processes that require immediate insights into their critical operations. The process is also not suitable for companies that have heightened security and privacy protocols (such as nuclear plants and defense manufacturing). The company seeks to enable companies to conduct predictive analytics on geographically distributed data silos without the need to move data from its location, thus reducing decision latency. This SBIR Phase I project proposes to develop a technology stack that leverages the blockchain to train advanced analytic algorithms and Machine Learning models across data silos in different locations without relying on the Cloud or any corporate server. Specifically, the project targets the innovative integration of the blockchain Smart Contracts with distributed memory programming frameworks like the Message Passing Interface (MPI). This integration raises several interesting research challenges one of which requires designing basic primitives similar in functionality to a Map-Reduce framework for the blockchain. The second research component revolves around the development of novel algorithmic decomposition schemes for popular Machine Learning algorithms and artificial intelligence (AI) models to facilitate their training in a decentralized manner using the blockchain. If successful, this technology may provide analytic insights faster than the Cloud. It may also significantly reduce the cost and labor associated with implementing and deploying industrial analytics. Data science teams will be provided with the agility and flexibility to modify and redeploy algorithms without having to rebuild and redesign data pipeline infrastructure to a centralized server.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.