I wonder how the Partitioned search final merge to one solution

[李陈赞]

Hi,anyone can help?I come across a problem when I use the partitioned search.I implement a partitioner class to  split the origin problem into multiple small problem,and it worked,

each partition seems get a fine partition solutiion,but finally when the partition search phase ended,it went into the local search phase,things goes wrong,the score turn into much worse,seems like it go into the origin.I wonder why the optaplaner does't offer a interface method for merge sub partition into one big problem,and what the merge strategy optaplaner use to merge currently?I want to figure out the process.