Conventional industrial production systems have quality management related problems as following: Products are not checked by a total inspection but by a spot inspection, in which only randomly selected samples from a production lot are inspected and hence it is impossible to completely prevent some defective products in mass-production from being shipped. By contrast, individual parts and components are inspected unnecessarily and excessively, after each process, which causes a high expense for production. Here the functions of the products are evaluated indirectly despite the well-known fact that quality should be evaluated with products' essential functions, i.e. how they work. Instead, they are evaluated by too many criteria such as size, weight, appearance and so on functions. The check itself is based on the thought of "removing defects." Everytime a new defect is found, new inspection items are specified or the standards made stricter to eliminate the defect. However, since, this method only follows the defects and there are quite many modes of defects, it can not eliminate unknown defects. These problems are interfering with the reduction of the defect rate and the production cost. Conventionally, the quality control method mentioned above has been a mainstream. However, the lot management basically can not detect and eliminate all the defects. So there remain certain defects in a production lot size often million pieces. The rate of management cost increases in such mass production with a cheap unit price. To overcome the problems mentioned above, the authors suggest a new production system as follows: At first, we evaluate the final products directly and totally. When the quality of final products are ensured, the intermediate inspection will be not necessary. In that case, we only have to assure the quality of the unfinished products not by measuring themselves but by monitoring the production process. When the production machine works normally, the product must be fine. Here, the products should be selected with a point of view of 'passing the fine products.' This method can save the inspection cost without decreasing the quality of the products. To determine whether the machine works normally or not, we apply MT (Mahalanobis-Taguchi) system. In order to know that the product is fine from the monitoring datum, the product should correspond to the process datum by one-to-one. For example, an injection molding process has to be a single molding. A process datum of a multi molding includes several data of several parts. If an defective datum is found in a multi molding process, we cannot find which one is defective. Also, two or more defective products in a lot could cancel each other. In that case, the defective products can not be detected. Besides, for measuring the production process data with low noise, conventional production machines are too big. The big machines makes much larger value of process data with a big noise which can hide the process that we need to know. Therefore the production machines have to be as small as possible. By synthesizing small production machines with single production into an assembly line, a consecutive production line without intermediate inspection can be constructed. In this research, in order to realize such production line, the authors develop small production machines that can monitor the process. Specifically we develop a small injection molding machine and a small press machine. Using the small production machines, we achieve one-to-one correspondence of machined products and process data. Besides, we confirm that the evaluation of products is possible by monitoring the process data.
展开▼
