IS220PPROS1B Classification and challenges of industrial big data

In fact, there are three characteristics of industrial data:

The first feature is multimodal. In the past, it was simple and crude to divide data into structured data, semi-structured data, and unstructured data, but this is not the case in industrial enterprises. Much of the unstructured engineering data IS220PPROS1B you see today looks like it’s in a different format, but it’s not the same when you open it up. The use efficiency of unstructured data depends on the degree of structure, and only structure can be used efficiently.

The second feature is high throughput. A lot of equipment is not shut down, all the data is 7*24 hours continuous generation, the amount is very large;

The third characteristic is strong correlation. In different parts of the industry, data correlation follows different rules rather than simple aggregation.

IS220PPROS1B Therefore, the characteristics of industrial big data itself bring a lot of challenges. In addition to the challenge of data acquisition, there is also the challenge of data analysis and application.

IS220PPROS1B

IS220PPROS1B The biggest limitation here is causation, that is, data-driven approaches can only tell us about correlation, not causation. For example, Taobao recommends goods, only know to recommend related goods, but do not care about the cause and effect of this matter – why the user is such a person. But this is not feasible in industry, especially in control, so the model needs to be analyzed and validated for a long time.

There are white box models and gray box models in the industrial field, and the white box model is the industrial mechanism, and the enterprise will design the process, product structure and process according to the industrial mechanism, which is the first step. When they are designed, there will be a large number of uncertainties in operation, and the elimination of these uncertainties depends on the experience of experts and craftsmen, so that the entire process of production becomes more stable and efficient, which is the gray box state. A data model that no longer analyses and understands the mechanism and the knowledge itself is a black box model.

IS220PPROS1B The essence of industrial big data and industrial intelligence is to quantify these experiences and knowledge, dig out the tacit knowledge that you have in your heart and mouth, or try to find the statistical relationship through data methods, and then return it to the artisans for analysis.

Industry is industry, its existence time is longer than information time, accumulation is more than information, and big data and artificial intelligence technology only brings small changes to the industry, trying to help it to eliminate uncertainty.